CNN_networks_on_MNIST_dataset.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "CNN_networks_on_MNIST_dataset.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.1"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "RbdhMBcUpJzD",
        "outputId": "6b144f62-6413-495c-c81c-79d0d9de0b16",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 101
        }
      },
      "source": [
        "from __future__ import print_function\n",
        "import keras\n",
        "from keras.datasets import mnist\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten, Activation\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras import backend as K\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "\n",
        "(X_train,y_train),(X_test,y_test)=mnist.load_data()\n",
        "print('X_train shape: ',X_train.shape)\n",
        "print('Y_train shape: ',y_train.shape)\n",
        "print('-'*50)\n",
        "print('X_test shape: ',X_test.shape)\n",
        "print('y_test shape: ',y_test.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train shape:  (60000, 28, 28)\n",
            "Y_train shape:  (60000,)\n",
            "--------------------------------------------------\n",
            "X_test shape:  (10000, 28, 28)\n",
            "y_test shape:  (10000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "JMAm2ACaqc94",
        "outputId": "574abcf4-5f0e-4684-cdeb-fbadd7fd78c6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "source": [
        "# input image dimensions\n",
        "img_rows, img_cols = 28, 28\n",
        "if K.image_data_format() == 'channels_first':\n",
        "    X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)\n",
        "    X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)\n",
        "    input_shape = (1, img_rows, img_cols)\n",
        "else:\n",
        "    X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)\n",
        "    X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)\n",
        "    input_shape = (img_rows, img_cols, 1)\n",
        "\n",
        "X_train = X_train.astype('float32')\n",
        "X_test = X_test.astype('float32')\n",
        "X_train /= 255\n",
        "X_test /= 255\n",
        "print('X_train shape:', X_train.shape)\n",
        "print(X_train.shape[0], 'train samples')\n",
        "print(X_test.shape[0], 'test samples')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train shape: (60000, 28, 28, 1)\n",
            "60000 train samples\n",
            "10000 test samples\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "1lpx_ET6qus7",
        "outputId": "3347e015-7368-4024-8b22-80c44e6992f4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "#Some initialization\n",
        "batch_size = 128\n",
        "num_classes = 10\n",
        "epochs = 12\n",
        "# convert class vectors to binary class matrices\n",
        "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
        "\n",
        "print(y_train[0])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "8ALB4RqZnb4K"
      },
      "source": [
        "# CNN WITH 3 LAYERS CONSIDERING RELU AS A ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "AbDKIYoUqb6V"
      },
      "source": [
        "###  USING BN+DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "0L4ErwlzqxyD",
        "outputId": "8784d476-b3e8-4421-d466-a46931e57c49",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 554
        }
      },
      "source": [
        "from keras.layers.normalization import BatchNormalization\n",
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d-->BN-->Pooling-->FC-->Dropout-->Softmax\n",
        "#We will use Relu as Activation function and Adam as Optimizer\n",
        "#We will also use BatchNormalization and Dropout at suitable places\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_7\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_19 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_19 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_20 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_20 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_21 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 3, 3, 64)          12        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_21 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_7 (Flatten)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_13 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_7 (Activation)    (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_14 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,166\n",
            "Trainable params: 94,160\n",
            "Non-trainable params: 6\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Vwrsj26Y4kKT",
        "outputId": "b2109d26-7acb-4f32-da57-07d61cdd7f48",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 5s 77us/step - loss: 0.2481 - acc: 0.9224 - val_loss: 0.0739 - val_acc: 0.9775\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0740 - acc: 0.9768 - val_loss: 0.0527 - val_acc: 0.9833\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0519 - acc: 0.9836 - val_loss: 0.0571 - val_acc: 0.9843\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0409 - acc: 0.9866 - val_loss: 0.0571 - val_acc: 0.9830\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0331 - acc: 0.9896 - val_loss: 0.0468 - val_acc: 0.9872\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 57us/step - loss: 0.0305 - acc: 0.9901 - val_loss: 0.0459 - val_acc: 0.9868\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0244 - acc: 0.9922 - val_loss: 0.0521 - val_acc: 0.9854\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0215 - acc: 0.9932 - val_loss: 0.0417 - val_acc: 0.9888\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0195 - acc: 0.9937 - val_loss: 0.0384 - val_acc: 0.9905\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0149 - acc: 0.9952 - val_loss: 0.0565 - val_acc: 0.9851\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 57us/step - loss: 0.0150 - acc: 0.9950 - val_loss: 0.0554 - val_acc: 0.9880\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 4s 59us/step - loss: 0.0136 - acc: 0.9954 - val_loss: 0.0668 - val_acc: 0.9860\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "BCsiP5N45uYx",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "HvR0U5okA0TT",
        "outputId": "eec7a323-7695-4972-d7ec-5a44faa106d0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.06679211560489139\n",
            "Test accuracy: 0.986\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUwknjIEjhdwAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "OhgkmR8bqb6c"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "WiUudpgTqb6d",
        "outputId": "dfe2b162-243b-4418-afb9-cd7f37461e57",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d---->Pooling-->FC-->Dropout-->Softmax\n",
        "#We will use Relu as Activation function and Adam as Optimizer\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_8\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_22 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_22 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_23 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_23 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_24 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_24 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_8 (Flatten)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_15 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_8 (Activation)    (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_7 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_16 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,154\n",
            "Trainable params: 94,154\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "eltWlRchqb6e",
        "outputId": "d7ceef84-5fd4-45ad-d57a-dc036fb5cd9e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 4s 71us/step - loss: 0.3654 - acc: 0.8838 - val_loss: 0.1055 - val_acc: 0.9682\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.1023 - acc: 0.9691 - val_loss: 0.0695 - val_acc: 0.9804\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0728 - acc: 0.9786 - val_loss: 0.0539 - val_acc: 0.9841\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0581 - acc: 0.9823 - val_loss: 0.0513 - val_acc: 0.9842\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.0471 - acc: 0.9852 - val_loss: 0.0430 - val_acc: 0.9871\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 52us/step - loss: 0.0409 - acc: 0.9869 - val_loss: 0.0480 - val_acc: 0.9865\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 52us/step - loss: 0.0370 - acc: 0.9886 - val_loss: 0.0459 - val_acc: 0.9858\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0321 - acc: 0.9894 - val_loss: 0.0445 - val_acc: 0.9875\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0279 - acc: 0.9911 - val_loss: 0.0445 - val_acc: 0.9889\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0256 - acc: 0.9918 - val_loss: 0.0486 - val_acc: 0.9872\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 50us/step - loss: 0.0225 - acc: 0.9927 - val_loss: 0.0448 - val_acc: 0.9877\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 3s 50us/step - loss: 0.0186 - acc: 0.9938 - val_loss: 0.0575 - val_acc: 0.9847\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "8ieMqNaaqb6g",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "kCVoWKkYqb6h",
        "outputId": "51d1ccf9-dcc4-4d26-dd70-ccc0648031f9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.05752517269679956\n",
            "Test accuracy: 0.9847\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Q39KS57eqb6j"
      },
      "source": [
        "\n",
        "### WITHOUT DROPOUT "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Y3C_IPuzqb6j",
        "outputId": "2e3f9c97-80fc-406f-8f5c-44dd24341123",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "from keras.layers.normalization import BatchNormalization\n",
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d-->BN-->Pooling-->FC--->Softmax\n",
        "#We will use Relu as Activation function and Adam as Optimizer\n",
        "#We will also use BatchNormalization and Dropout at suitable places\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_9\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_25 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_25 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_26 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_26 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_27 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 3, 3, 64)          12        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_27 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_9 (Flatten)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_17 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_9 (Activation)    (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_18 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,166\n",
            "Trainable params: 94,160\n",
            "Non-trainable params: 6\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "F3-Rk2gXqb6l",
        "outputId": "12595aa4-2a34-42fd-ab9c-c7dd42407476",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 5s 80us/step - loss: 0.1998 - acc: 0.9387 - val_loss: 0.0811 - val_acc: 0.9746\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0646 - acc: 0.9798 - val_loss: 0.0665 - val_acc: 0.9787\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0457 - acc: 0.9857 - val_loss: 0.0521 - val_acc: 0.9850\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0348 - acc: 0.9890 - val_loss: 0.0514 - val_acc: 0.9851\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0272 - acc: 0.9909 - val_loss: 0.0525 - val_acc: 0.9845\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0234 - acc: 0.9923 - val_loss: 0.0541 - val_acc: 0.9857\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 57us/step - loss: 0.0189 - acc: 0.9940 - val_loss: 0.0494 - val_acc: 0.9871\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0143 - acc: 0.9955 - val_loss: 0.0582 - val_acc: 0.9848\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0156 - acc: 0.9949 - val_loss: 0.0567 - val_acc: 0.9855\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0125 - acc: 0.9959 - val_loss: 0.0535 - val_acc: 0.9874\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0126 - acc: 0.9958 - val_loss: 0.0551 - val_acc: 0.9863\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0079 - acc: 0.9973 - val_loss: 0.0614 - val_acc: 0.9848\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "076Soc3Hqb6m",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()\n",
        "    \n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "oGARQ8OQqb6o",
        "outputId": "2dc5690d-b4d8-48d4-903e-3c0d2c86d3b8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.06143429972460553\n",
            "Test accuracy: 0.9848\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nj9F4wqkuAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "8vfRX1Fyqb6q"
      },
      "source": [
        "# CNN WITH 3 LAYERS CONSIDERING SIGMOID AS A ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "yrkAMJGkqb6q"
      },
      "source": [
        "### USING BN + DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Yd_gJOllqb6r",
        "outputId": "781fe1ee-106c-4c00-9aaa-b3f0a8e8bf10",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 554
        }
      },
      "source": [
        "from keras.layers.normalization import BatchNormalization\n",
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d-->BN-->Pooling-->FC-->Dropout-->Softmax\n",
        "#We will use sigmoid as Activation function and Adam as Optimizer\n",
        "#We will also use BatchNormalization and Dropout at suitable places\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_10\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_28 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_28 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_29 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_29 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_30 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 3, 3, 64)          12        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_30 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_10 (Flatten)         (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_19 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_10 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_8 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_20 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,166\n",
            "Trainable params: 94,160\n",
            "Non-trainable params: 6\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2CBeVbulqb6s",
        "outputId": "2470e262-dc29-4096-b3f1-9bf9703f655b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 5s 79us/step - loss: 1.8599 - acc: 0.3115 - val_loss: 0.5820 - val_acc: 0.8273\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.2218 - acc: 0.9327 - val_loss: 0.1321 - val_acc: 0.9566\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.1352 - acc: 0.9585 - val_loss: 0.0884 - val_acc: 0.9733\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.1052 - acc: 0.9678 - val_loss: 0.0868 - val_acc: 0.9735\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0871 - acc: 0.9729 - val_loss: 0.0884 - val_acc: 0.9718\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0768 - acc: 0.9762 - val_loss: 0.0776 - val_acc: 0.9758\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0662 - acc: 0.9795 - val_loss: 0.0688 - val_acc: 0.9789\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0605 - acc: 0.9814 - val_loss: 0.0679 - val_acc: 0.9797\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0552 - acc: 0.9826 - val_loss: 0.0605 - val_acc: 0.9827\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0487 - acc: 0.9845 - val_loss: 0.0559 - val_acc: 0.9832\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 56us/step - loss: 0.0428 - acc: 0.9865 - val_loss: 0.0621 - val_acc: 0.9817\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0413 - acc: 0.9869 - val_loss: 0.0511 - val_acc: 0.9853\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "XurgzZZHqb6u",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "22Xn2e0_qb6w",
        "outputId": "4e36bc5d-4071-4342-c035-5972772c4d4e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.051126524720713495\n",
            "Test accuracy: 0.9853\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nF8n/B2N0hEJSYZDlAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "0LdqRsKMqb60"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "n5HUqTiUqb61",
        "outputId": "9b6635a3-b0c2-4705-abc6-f974a0b2a2d5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d---->Pooling-->FC-->Dropout-->Softmax\n",
        "#We will use Relu as Activation function and Adam as Optimizer\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_11\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_31 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_31 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_32 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_32 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_33 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_33 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_11 (Flatten)         (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_21 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_11 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_9 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_22 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,154\n",
            "Trainable params: 94,154\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "pda9F2eXqb62",
        "outputId": "008f21f7-015b-4f8f-c714-fcb902406169",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 4s 74us/step - loss: 2.3540 - acc: 0.1027 - val_loss: 2.3046 - val_acc: 0.0974\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 2.3052 - acc: 0.1054 - val_loss: 2.3026 - val_acc: 0.0974\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 2.3036 - acc: 0.1066 - val_loss: 2.3024 - val_acc: 0.1135\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 52us/step - loss: 2.3024 - acc: 0.1089 - val_loss: 2.3013 - val_acc: 0.1135\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 2.1929 - acc: 0.1583 - val_loss: 1.2194 - val_acc: 0.6413\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 52us/step - loss: 0.4917 - acc: 0.8502 - val_loss: 0.2106 - val_acc: 0.9333\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 52us/step - loss: 0.2133 - acc: 0.9356 - val_loss: 0.1463 - val_acc: 0.9534\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.1559 - acc: 0.9520 - val_loss: 0.1139 - val_acc: 0.9646\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.1280 - acc: 0.9612 - val_loss: 0.0952 - val_acc: 0.9716\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.1119 - acc: 0.9655 - val_loss: 0.0963 - val_acc: 0.9705\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0975 - acc: 0.9702 - val_loss: 0.0795 - val_acc: 0.9757\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 3s 51us/step - loss: 0.0873 - acc: 0.9731 - val_loss: 0.0753 - val_acc: 0.9779\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "HcSosV6Uqb63",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "0RyNUBbuqb65",
        "outputId": "da14ff20-e528-4ca8-8aec-c1573a475149",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.07533405585018918\n",
            "Test accuracy: 0.9779\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "YwIin4c-qb67"
      },
      "source": [
        "### WITHOUT DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "DGupo9Pbqb68",
        "outputId": "94bb458f-308e-4dae-831b-e3866873651b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 521
        }
      },
      "source": [
        "from keras.layers.normalization import BatchNormalization\n",
        "#Conv2d-->Pooling-->Conv2d-->Pooling-->Conv2d-->BN-->Pooling-->FC---->Softmax\n",
        "#We will use sigmoid as Activation function and Adam as Optimizer\n",
        "#We will also use BatchNormalization and Dropout at suitable places\n",
        "\n",
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_12\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_34 (Conv2D)           (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_34 (MaxPooling (None, 13, 13, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_35 (Conv2D)           (None, 11, 11, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_35 (MaxPooling (None, 5, 5, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_36 (Conv2D)           (None, 3, 3, 64)          36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_7 (Batch (None, 3, 3, 64)          12        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_36 (MaxPooling (None, 1, 1, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_12 (Flatten)         (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_23 (Dense)             (None, 512)               33280     \n",
            "_________________________________________________________________\n",
            "activation_12 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_24 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 94,166\n",
            "Trainable params: 94,160\n",
            "Non-trainable params: 6\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "nxtqDAwCqb6-",
        "outputId": "f34ce532-9e84-49d4-da35-3b8f566dddbb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 5s 83us/step - loss: 1.4658 - acc: 0.4630 - val_loss: 0.3256 - val_acc: 0.9233\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.1915 - acc: 0.9427 - val_loss: 0.1524 - val_acc: 0.9523\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.1274 - acc: 0.9606 - val_loss: 0.1009 - val_acc: 0.9699\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.1000 - acc: 0.9694 - val_loss: 0.0800 - val_acc: 0.9762\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0825 - acc: 0.9748 - val_loss: 0.0878 - val_acc: 0.9740\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0738 - acc: 0.9770 - val_loss: 0.0650 - val_acc: 0.9799\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0659 - acc: 0.9801 - val_loss: 0.0957 - val_acc: 0.9701\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0613 - acc: 0.9808 - val_loss: 0.0739 - val_acc: 0.9781\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.0536 - acc: 0.9829 - val_loss: 0.0715 - val_acc: 0.9782\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0478 - acc: 0.9848 - val_loss: 0.0624 - val_acc: 0.9834\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 3s 54us/step - loss: 0.0459 - acc: 0.9857 - val_loss: 0.0643 - val_acc: 0.9824\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 3s 55us/step - loss: 0.0395 - acc: 0.9874 - val_loss: 0.0772 - val_acc: 0.9788\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "QlObdzmMqb6_",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "aJAsesSmqb7A",
        "outputId": "a96891ee-9ffb-470e-e4a2-e7c0e4be0833",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.07715547566683963\n",
            "Test accuracy: 0.9788\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njiUFY4wxdSwpGGOMqWNJwRhjTJ3/D2eE7RYf35e+AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "7W_oSLDiC0rR"
      },
      "source": [
        "### CNN WITH 5 LAYERS AS RELU ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "S93Rqe3Fqb7C"
      },
      "source": [
        "### USING BN+DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "OnQdwZBZA3Gl",
        "outputId": "1127191b-2102-4fe6-87b9-522f787a8836",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(Dropout(0.20))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.20))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_13\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_37 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_37 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_38 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "dropout_10 (Dropout)         (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_38 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_39 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "batch_normalization_8 (Batch (None, 10, 10, 32)        40        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_39 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_40 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "batch_normalization_9 (Batch (None, 9, 9, 32)          36        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_40 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_41 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_41 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "dropout_11 (Dropout)         (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_13 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_25 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_13 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_12 (Dropout)         (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_26 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,134\n",
            "Trainable params: 76,096\n",
            "Non-trainable params: 38\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "4SeEsK4KCATs",
        "outputId": "dc116839-ed2c-4202-9e9a-ffbb7c7fab14",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 11s 175us/step - loss: 0.5994 - acc: 0.7974 - val_loss: 0.1416 - val_acc: 0.9581\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.2292 - acc: 0.9289 - val_loss: 0.1115 - val_acc: 0.9662\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.1768 - acc: 0.9450 - val_loss: 0.0797 - val_acc: 0.9771\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.1460 - acc: 0.9555 - val_loss: 0.0742 - val_acc: 0.9795\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.1314 - acc: 0.9597 - val_loss: 0.0776 - val_acc: 0.9786\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 8s 137us/step - loss: 0.1154 - acc: 0.9644 - val_loss: 0.0670 - val_acc: 0.9814\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 8s 136us/step - loss: 0.1089 - acc: 0.9665 - val_loss: 0.0698 - val_acc: 0.9801\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 8s 138us/step - loss: 0.1030 - acc: 0.9678 - val_loss: 0.0578 - val_acc: 0.9834\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 8s 138us/step - loss: 0.0928 - acc: 0.9721 - val_loss: 0.0524 - val_acc: 0.9860\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 8s 139us/step - loss: 0.0872 - acc: 0.9732 - val_loss: 0.0510 - val_acc: 0.9857\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 8s 138us/step - loss: 0.0853 - acc: 0.9738 - val_loss: 0.0537 - val_acc: 0.9851\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 8s 138us/step - loss: 0.0814 - acc: 0.9756 - val_loss: 0.0524 - val_acc: 0.9852\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "g5LuFNpxC8An",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "3aYmRHoUTIaS",
        "outputId": "db5bece3-6e5e-4998-d57d-19095fbe617c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.05238832147307694\n",
            "Test accuracy: 0.9852\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "dazm_Tmxqb7I"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ycJhkLbbqb7I",
        "outputId": "695e8f40-93a4-445d-f8b1-507a383f5e1f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 722
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(Dropout(0.20))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.20))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_14\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_42 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_42 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_43 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "dropout_13 (Dropout)         (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_43 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_44 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_44 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_45 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_45 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_46 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_46 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "dropout_14 (Dropout)         (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_14 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_27 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_14 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_15 (Dropout)         (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_28 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,058\n",
            "Trainable params: 76,058\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ZrCtP_Wdqb7K",
        "outputId": "d7c29c31-5be1-416a-ef1f-d1fec71d049b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 10s 159us/step - loss: 0.7369 - acc: 0.7520 - val_loss: 0.2068 - val_acc: 0.9434\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.3211 - acc: 0.9007 - val_loss: 0.1479 - val_acc: 0.9586\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.2396 - acc: 0.9259 - val_loss: 0.1113 - val_acc: 0.9684\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.2066 - acc: 0.9379 - val_loss: 0.1184 - val_acc: 0.9664\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.1791 - acc: 0.9450 - val_loss: 0.0966 - val_acc: 0.9745\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.1571 - acc: 0.9525 - val_loss: 0.0849 - val_acc: 0.9769\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.1442 - acc: 0.9563 - val_loss: 0.0780 - val_acc: 0.9806\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 123us/step - loss: 0.1355 - acc: 0.9587 - val_loss: 0.0794 - val_acc: 0.9780\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.1243 - acc: 0.9623 - val_loss: 0.0707 - val_acc: 0.9804\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.1155 - acc: 0.9656 - val_loss: 0.0650 - val_acc: 0.9815\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.1107 - acc: 0.9667 - val_loss: 0.0604 - val_acc: 0.9848\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 7s 120us/step - loss: 0.1019 - acc: 0.9694 - val_loss: 0.0604 - val_acc: 0.9832\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "cjwML2t-qb7L",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "qEjgP6Vfqb7M",
        "outputId": "9a12e4ed-1d0b-4bf0-94ff-98ff6ead597a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.06042629976607859\n",
            "Test accuracy: 0.9832\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npGCMMabA/wcywUCS+FO4WAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "58ik5_JYqb7O"
      },
      "source": [
        "### WITHOUT DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "JQ7k2eqbqb7O",
        "outputId": "43d346ef-d954-4453-8e80-10ab78284707",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 689
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('relu'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_15\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_47 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_47 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_48 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_48 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_49 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "batch_normalization_10 (Batc (None, 10, 10, 32)        40        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_49 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_50 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "batch_normalization_11 (Batc (None, 9, 9, 32)          36        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_50 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_51 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_51 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_15 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_29 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_15 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_30 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,134\n",
            "Trainable params: 76,096\n",
            "Non-trainable params: 38\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "f4P8-gilqb7P",
        "outputId": "1b7c6848-1b6f-4ef7-cf4c-ae70e80dc03b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 9s 155us/step - loss: 0.2806 - acc: 0.9127 - val_loss: 0.0900 - val_acc: 0.9721\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0881 - acc: 0.9727 - val_loss: 0.0809 - val_acc: 0.9754\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 7s 116us/step - loss: 0.0641 - acc: 0.9799 - val_loss: 0.0663 - val_acc: 0.9813\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0526 - acc: 0.9835 - val_loss: 0.0566 - val_acc: 0.9831\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0445 - acc: 0.9860 - val_loss: 0.0542 - val_acc: 0.9848\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0403 - acc: 0.9865 - val_loss: 0.0511 - val_acc: 0.9858\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 7s 116us/step - loss: 0.0336 - acc: 0.9887 - val_loss: 0.0632 - val_acc: 0.9801\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 117us/step - loss: 0.0295 - acc: 0.9902 - val_loss: 0.0523 - val_acc: 0.9860\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 7s 118us/step - loss: 0.0274 - acc: 0.9908 - val_loss: 0.0496 - val_acc: 0.9875\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0262 - acc: 0.9910 - val_loss: 0.0601 - val_acc: 0.9838\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 7s 115us/step - loss: 0.0229 - acc: 0.9918 - val_loss: 0.0519 - val_acc: 0.9860\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 7s 116us/step - loss: 0.0210 - acc: 0.9932 - val_loss: 0.0473 - val_acc: 0.9867\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "10ExQLmRqb7Q",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "8SRyVRHUqb7S",
        "outputId": "6e9f54ec-e9e3-4092-c371-1fef997d1b25",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.04732237683171697\n",
            "Test accuracy: 0.9867\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl4VOXZ+PHvTQgESAg7EkBWRQKE\nBFLAihAUFUFB+8MFl7rzVmutWn1rrS2+VOtSq2hdcWutCFUQVxBRiUtFVgVkE0TAAAoEhIQ94f79\n8ZwhQ0gyJ8tkMpP7c13nysyZc87cTwJzz7Oc5xFVxRhjjClLnUgHYIwxpuazZGGMMSYkSxbGGGNC\nsmRhjDEmJEsWxhhjQrJkYYwxJiRLFsZUgIioiHSNdBzGVBdLFibqich6EdknIvlB2+ORjqsqiUhH\nL0HVjXQspnayf3gmVpyrqh9EOghjYpXVLExME5ErReS/IvK4iOwSkVUicnrQ6yki8paI7BCRtSJy\nXdBrcSJyp4h8KyJ5IrJIRNoHXX6oiKwRkZ9E5AkRkRLeP8Wr9TQL2pchIttFJF5EuorIx15s20Xk\nPxUoY30RmSAim71tgojU915rISLveDHuEJFPRaSO99rvRWSTV7bVwb8XY4qzZGFqg/7At0ALYBzw\netCH9xQgB0gBRgN/FZHTvNduBcYAw4HGwNXA3qDrngP8DEgDLgTOKv7GqroZmAv8v6DdlwBTVfUQ\n8BfgfaAp0A74RwXK90dgAJAO9Ab6AXd5r/3OK19LoDVwJ6Ai0g24EfiZqiZ5sa+vwHubWsKShYkV\nb3jfngPbdUGvbQUmqOohVf0PsBoY4dUSTgF+r6r7VfUr4Dngl9551wJ3qepqdZaoam7Qde9X1Z9U\ndSMwB/dhXZJXcEkHr/ZxsbcP4BDQAUjxYvisAmW/FBivqltVdRvwf8DlQddvA3Twyv+pugnhCoH6\nQKqIxKvqelX9tgLvbWoJSxYmVpynqk2CtmeDXtukR8+YuQFXk0gBdqhqXrHX2nqP2+NqJKX5Iejx\nXiCxlOOmASeLSBtgEHAY+NR77X8BAeaLyHIRubqM9ytNihd3QKB8AH8D1gLvi8g6EbkDQFXXAjcD\ndwNbRWSKiKRgTCksWZjaoG2x/oTjgc3e1kxEkoq9tsl7/D3QpbJvrqo7cU1NF+GaoKYEkpeq/qCq\n16lqCvA/wJMVGJK7GVc7CQiUD1XNU9XfqWpnYCRwa6BvQlVfUdWB3rkKPFDhQpqYZ8nC1AatgJu8\nDuULgO7ADFX9HvgcuE9EEkQkDbgGeNk77zngLyJygjhpItK8gjG8gmveGk1RExQicoGItPOe7sR9\naB8u4zr1vVgDWx1gMnCXiLQUkRbAnwNlEJFzvE50AXbhmp8Oi0g3ETnN6wjfD+wL8b6mlrOhsyZW\nvC0ihUHPZ6vq+d7jecAJwHbgR2B0UN/DGOBp3DfxncC4oCG4D+Pa9d/HdY6vAgLXLK+3cMlno6ou\nCdr/M2CCiCR7sf1WVdeVcZ38Ys/PAO7BdcAv9fa95u0DV+7HcR3cO4EnVXWOlxjvxyXOQ7ikObaC\nZTO1gNjiRyaWiciVwLVec4sxpoKsGcoYY0xIliyMMcaEZM1QxhhjQrKahTHGmJBiZjRUixYttGPH\njpEOw5c9e/bQqFGjSIcRNrFcPitb9Irl8lWmbIsWLdquqi1DHRczyaJjx44sXLgw0mH4kp2dTVZW\nVqTDCJtYLp+VLXrFcvkqUzYR2RD6KGuGMsYY44MlC2OMMSFZsjDGGBNSzPRZGGOqx6FDh8jJyWH/\n/v2RDqVckpOTWblyZaTDCAs/ZUtISKBdu3bEx8dX6D0sWRhjyiUnJ4ekpCQ6duxICYsD1lh5eXkk\nJSWFPjAKhSqbqpKbm0tOTg6dOnWq0HtYM5Qxplz2799P8+bNoypR1HYiQvPmzStVG7RkYYwpN0sU\n0aeyfzNLFjt3wvjxsGhRpCMxxpgay5JFXByMGwezZkU6EmOMD0OGDGFWsf+vEyZM4Prrry/zvDZt\n2gCwefNmRo8eXeIxWVlZIW/unTBhAnv37j3yfPjw4fz0009+Qi/T3XffzUMPPVTp64SLJYvGjaFL\nF/jyy0hHYozxYcyYMUyZMuWofVOmTGHMmDG+zk9JSWHq1KkVfv/iyWLGjBk0adKkwteLFpYsANLT\n4auvIh2FMcaH0aNH8+6773Lw4EEA1q9fz+bNmzn11FPJz8/n9NNPp0+fPvTq1Ys333zzmPPXr19P\nz549Adi3bx8XX3wx3bt35/zzz2ffvn1Hjrv++uvJzMykR48ejBs3DoDHHnuMzZs3M2TIEIYMGQK4\nqYa2b98OwMMPP0zPnj3p2bMnEyZMOPJ+3bt357rrrqNHjx6ceeaZR71PKCVdc8+ePYwYMYLevXvT\ns2dPpk2bBsAdd9xBamoqaWlp3HbbbeX6vYZiQ2cBMjJg2jTIy4MYHVpnTDjcfHPVf89KTwfvM7FE\nzZo1o1+/fsycOZNRo0YxZcoULrzwQkSEhIQEpk+fTuPGjdm+fTsDBgxg5MiRpXbuPvXUUzRs2JCV\nK1eydOlS+vTpc+S1e++9l2bNmlFYWMjpp5/O0qVLuemmm3j44YeZM2cOLVq0OOpaixYt4sUXX2Te\nvHmoKv3792fw4ME0bdqUNWvWMHnyZJ599lkuvPBCpk2bxmWXXRbyd1HaNdetW0dKSgrvvvsu4IYz\n5+bmMn36dFatWoWIVEnTWDCrWYD71wmwZEnZxxljaoTgpqjgJihV5c477yQtLY2hQ4eyadMmfvzx\nx1Kv88knnxz50E5LSyMtLe3Ia6+++ip9+vQhIyOD5cuXs2LFijJj+uyzzzj//PNp1KgRiYmJ/OIX\nv+DTTz8FoFOnTqR7nzN9+/Zl/fr1vspZ2jV79erF7Nmz+f3vf8+nn35KcnIyycnJJCQkcM011/D6\n66/TsGFDX+/hl9UsoChZfPUVDLSlmo3xq6waQDiNGjWKW265hcWLF7N371769u0LwKRJk9i2bRuL\nFi0iPj6ejh07Vujegu+++46HHnqIBQsW0LRpU6688spK3aNQv379I4/j4uLK1QxVkhNPPJHFixcz\nY8YM7rrrLgYOHMi9997L/Pnz+fDDD5k6dSqPP/44H330UaXeJ5jVLABSUqBlS+u3MCZKJCYmMmTI\nEK6++uqjOrZ37dpFq1atiI+PZ86cOWzYUPbs24MGDeKVV14B4Ouvv2bp0qUA7N69m0aNGpGcnMyP\nP/7IzJkzj5yTlJREXl7eMdc69dRTeeONN9i7dy979uxh+vTpnHrqqZUqZ2nX3Lx5Mw0bNuSyyy7j\n9ttvZ8mSJeTn57Nr1y6GDx/OI488wpIqbimxmgWAiKtd2IgoY6LGmDFjOP/8848aGXXppZdy7rnn\n0qtXLzIzMznppJPKvMb111/PVVddRffu3enevfuRGkrv3r3JyMjgpJNOon379pxyyilHzhk7dizD\nhg0jJSWFOXPmHNnfp08frrzySvr16wfAtddeS0ZGhu8mJ4B77rnnSCc2uL6Ikq45a9Ysbr/9durU\nqUN8fDwPPfQQeXl5jBo1iv3796OqPPzww77f1xdVjYmtb9++Wim3365ar57qwYOVu44Pc+bMCft7\nRFIsl8/KprpixYrwBhImu3fvjnQIYeO3bCX97YCF6uMz1pqhAtLT4eBBWLUq0pEYY0yNY8kiICPD\n/bR+C2OMOYYli4ATT4QGDazfwhhjSmDJIiAuDnr1spqFMcaUwJJFsIwMlyxUIx2JMcbUKJYsgqWn\nuynLN26MdCTGGFOjWLIIFnwntzGmRsrNzSU9PZ309HSOO+442rZte+R5YHLBUK666ipWr17t+z2f\ne+45br755oqGHBPsprxgaWlQp45LFqNGRToaY0wJmjdvzlfeF7q7776bxMTEY2ZYPXJvQJ2Svw+/\n+OKLYY8z1ljNIljDhm5UlI2IMibqrF27ltTUVC699FJ69OjBli1bGDt27JFpxu+///4jxw4cOJCv\nvvqKgoICmjRpwh133EHv3r05+eST2bp1q+/3fPnll+nVqxc9e/bkzjvvBKCgoIDLL7/8yP7HHnsM\ngEceeeTI9OF+ZpytaaxmUVx6OsydG+kojIkOkZijvAyrVq3ipZdeIjMzE4D777+fZs2aUVBQwKBB\ng1ixYgWpqalHnbNr1y4GDx7M/fffz6233soLL7zAHXfcEfK9cnJyuOuuu1i4cCHJyckMHTqUd955\nh5YtW7J9+3aWLVsGcGSq8AcffJANGzZQr169Kp8+vDpYzaK4jAzYsMF1dBtjokqXLl2OJAqAyZMn\n06dPH/r06cPq1atLnGa8QYMGnH322UD5pg+fN28ep512Gi1atCA+Pp5LLrmETz75hK5du7J69Wpu\nuukmZs2aRXJyMgA9evTgsssuY9KkScTHx1e+sNUsrDULERkGPArEAc+p6v3FXr8VuBYoALYBV6vq\nBu+1QmCZd+hGVR0ZzliPCO7k9lbCMsaUIlJzlJeiUaNGRx6vWbOGRx99lPnz59OkSRMuuuiiEqcZ\nr1ev3pHHcXFxFBQUVCqG5s2bs3TpUmbOnMkTTzzBtGnTmDhxIrNmzeLjjz/mrbfe4q9//StLly4l\nLi6uUu9VncJWsxCROOAJ4GwgFRgjIqnFDvsSyFTVNGAq8GDQa/tUNd3bqidRAPTu7X7aiChjotru\n3btJSkqicePGbNmyhQ8//LBKr9+/f3/mzJlDbm4uBQUFTJkyhcGDB7Nt2zZUlQsuuIDx48ezePFi\nCgsLycnJ4bTTTuPBBx9k+/btR63jHQ3CWbPoB6xV1XUAIjIFGAUcqQeq6pyg478AIt/r07o1tGlj\nycKYKNenTx9SU1M56aST6NChAwMGDKjU9Z5//nmmTp165PnChQv5y1/+QlZWFqrKueeey4gRI1i8\neDHXXHMNqoqI8MADD1BQUMAll1xCXl4ehw8f5rbbbiMp2pZw9jM1bUU2YDSu6Snw/HLg8TKOfxy4\nK+h5AbAQl0TOC/V+lZ6iPNjw4appaVV3vWJieZpr1dgun5XNpiiviapjivIaMRpKRC4DMoHBQbs7\nqOomEekMfCQiy1T122LnjQXGArRu3Zrs7OwqiadT06a0nzWLT99/Hw1qz6wq+fn5VRZrTRTL5bOy\nQXJycokrxdV0hYWFURm3H37Ltn///gr/+w1nstgEtA963s7bdxQRGQr8ERisqgcC+1V1k/dznYhk\nAxnAUclCVScCEwEyMzM1KyuraiLftg0mTWJwixbQp0/VXDNIdnY2VRZrDRTL5bOywcqVK6OvCQXI\ny8uLyrj98Fu2hIQEMgLLMZRTyA5uEblARJK8x3eJyOsi4ucTdAFwgoh0EpF6wMXAW8WunQE8A4xU\n1a1B+5uKSH3vcQvgFIL6OsLO1rYwpkxqk21Gncr+zfyMhvqTquaJyEBgKPA88JSPwAqAG4FZwErg\nVVVdLiLjRSQwuulvQCLwmoh8JSKBZNIdWCgiS4A5wP2qWn3JonNnSEy0O7mNKUFCQgK5ubmWMKKI\nqpKbm0tCQkKFr+GnGarQ+zkCmKiq74rIPT4DnAHMKLbvz0GPh5Zy3udALz/vERZ16rghtFazMOYY\n7dq1Iycnh23btkU6lHLZv39/pT4sazI/ZUtISKBdu3YVfg8/yWKTiDwDnAE84DUPxf6d3xkZ8K9/\nweHDLnkYYwCIj4+nU6dOkQ6j3LKzsyvcXl/TVUfZ/HwKXohrSjpLVX8CmgG3hzWqmiA9HfLyYN26\nSEdijDER5ydZtAHeVdU1IpIFXADMD2tUNYGtbWGMMUf4SRbTgEIR6YobptoeeCWsUdUEPXpA3bqW\nLIwxBn/J4rA3sukXwD9U9XZcbSO2JSRA9+42IsoYY/CXLA6JyBjgl8A73r7om1+3ItLTrWZhjDH4\nSxZXAScD96rqdyLSCfh3eMOqIdLTYfNmKMfKWcYYE4tCJgvvZrjbgGUi0hPIUdUHwh5ZTRAYirZk\nSWTjMMaYCPMz3UcWsAa3NsWTwDciMijMcdUMgbUtrN/CGFPL+bkp7+/Amaq6GkBETgQmA33DGViN\n0KwZHH+89VsYY2o9P30W8YFEAaCq31BbOrjBNUVZsjDG1HJ+ksVCEXlORLK87VncokS1Q3o6rF4N\nUbYEojHGVCU/yeJ63PTgN3nbCuBX4QyqRklPd/NDLVsW6UiMMSZi/IyGOqCqD6vqL7ztEWrL0Fmw\ntS2MMYaKzx57cpVGUZMdfzw0aWIjoowxtZrNvR2KiN3JbYyp9UodOlvG0qlCbRoNBa4p6umnobAQ\n4uIiHY0xxlS7su6z+HsZr62q6kBqtPR02LcPvvnGTS5ojDG1TKnJQlWHVGcgNVrw2haWLIwxtZD1\nWfjRvTvUq2f9FsaYWsuShR/x8dCzpyULY0ytZcnCr/R0N3xWNdKRGGNMtfMz6+zrIjJCRGp3YklP\nh23bYMuWSEdijDHVzk8CeBK4BFgjIveLSLcwx1Qz2Z3cxphazM90Hx+o6qVAH2A98IGIfC4iV4lI\n7bnfIi3N/bQ7uY0xtZCvpiURaQ5cCVwLfAk8ikses8MWWU3TuDF06WI1C2NMrRRy8SMRmQ50w00e\neK6qBhrt/yMitWeqcrC1LYwxtZaflfIeU9U5Jb2gqplVHE/Nlp4OU6fC7t2upmGMMbWEn2aouSJy\nqzcqapqI3CIiCWGPrCYK3Mm9dGlk4zDGmGrmJ1m8BPQA/gE8DqTicz0LERkmIqtFZK2I3FHC67eK\nyAoRWSoiH4pIh6DXrhCRNd52hb/ihJmNiDLG1FJ+mqF6qmpq0PM5IrIi1EkiEgc8AZwB5AALROQt\nVQ0+90sgU1X3isj1wIPARSLSDBgHZAIKLPLO3emvWGHSpg20bGkjoowxtY6fmsViERkQeCIi/fG3\nBnc/YK2qrlPVg8AUYFTwAao6R1UDi1t/AbTzHp8FzFbVHV6CmA0M8/Ge4WVrWxhjaik/yaIv8LmI\nrBeR9cBc4GciskxEymq8bwt8H/Q8x9tXmmuAmRU8t/qkp8PXX8OhQ5GOxBhjqo2fZqiwf6MXkctw\nTU6Dy3neWGAsQOvWrcnOzq764IppVb8+qQcPsuCll9jTpUuFrpGfn18tsUZKLJfPyha9Yrl81VG2\nkMlCVTeISG/gVG/Xp6q6xMe1NwHtg5638/YdRUSGAn8EBqvqgaBzs4qdm11CbBOBiQCZmZmalZVV\n/JCq17o13HMPP4uPhwq+X3Z2NtUSa4TEcvmsbNErlstXHWXzM5Hgb4FJQCtve1lEfuPj2guAE0Sk\nk4jUAy4G3ip27QzgGWCkqm4NemkWcKaINBWRpsCZ3r7IO/FEaNDA+i2MMbWKn2aoa4D+qroHQEQe\nwPVb/KOsk1S1QERuxH3IxwEvqOpyERkPLFTVt4C/AYnAayICsFFVR6rqDhH5Cy7hAIxX1R0VKF/V\ni4tz80RZsjDG1CJ+koUAhUHPC719IanqDGBGsX1/Dno8tIxzXwBe8PM+1S49Hf7zH7e2hfj6VRhj\nTFTzMxrqRWCeiNwtInfjhrg+H9aoarr0dPjpJ9i4MdKRGGNMtfDTwf2wiGQDA71dV6lq7b4rLfhO\n7g4dyj7WGGNiQJnJwrsLe7mqngQsrp6QokCvXlCnjruTe9So0McbY0yUK7MZSlULgdUicnw1xRMd\nGjZ0o6Ksk9sYU0v46eBuCiwXkfnAnsBOVR0ZtqiiQUYGfP55pKMwxphq4SdZ/CnsUUSj9HSYPBl2\n7IBmzSIdjTHGhJWf0VDDVfXj4A0YHu7AarzA2hZL/NzMbowx0c1PsjijhH1nV3UgUSeQLKzfwhhT\nC5TaDOWtL3ED0LnY7LJJgDXWt2oFKSm2toUxplYoq8/iFdyU4fcBwavc5dWYqTcizda2MMbUEqU2\nQ6nqLlVdr6pjcOtJHMKtWpdoQ2k96emwciXs3x/pSIwxJqxCjobyJgO8G/gROOztViAtfGFFiYwM\nKCiAFSugT59IR2OMMWHjZ+jszUA3Vc0NdzBRJ9DJ/eWXliyMMTHNz2io74Fd4Q4kKnXuDElJ1m9h\njIl5fmoW64BsEXkXCKxkh6o+HLaookWdOtC7tyULY0zM81Oz2AjMBurhhs0GNgNFI6IOHw59rDHG\nRCk/U5T/H4CINFTVveEPKcqkp0N+PqxbB127RjoaY4wJCz9rcJ8sIiuAVd7z3iLyZNgjixbBa1sY\nY0yM8tMMNQE4C8gFUNUlwKBwBhVVUlOhbl27k9sYE9P8JAtU9ftiuwpLPLA2SkiA7t2tZmGMiWm+\nhs6KyM8BFZF4EbkNWBnmuKKLTfthjIlxfpLFr4BfA22BTUC699wEZGTA5s2wdWukIzHGmLDwMxpq\nO3BpNcQSvYKnKz/zzMjGYowxYeBnNNSDItLYa4L6UES2ichl1RFc1Ojd2/20pihjTIzy0wx1pqru\nBs4B1gNdgdvDGVTUadYMOnSwZGGMiVl+kkWgqWoE8Jqq2jxRJUlPt+GzxpiY5SdZvCMiq4C+wIci\n0hKwBRyKS0+H1athz55IR2KMMVUuZLJQ1TuAnwOZqnoI2AOMCndgUScjA1Th668jHYkxxlQ5Px3c\nFwCHVLVQRO4CXgZSwh5ZtAle28IYY2KMn2aoP6lqnogMBIYCzwNPhTesKHT88dC0qXVyG2Nikp9k\nEZjaYwQwUVXfxU1XHpKIDBOR1SKyVkTuKOH1QSKyWEQKRGR0sdcKReQrb3vLz/tFlIjdyW2MiVl+\nksUmEXkGuAiYISL1/ZwnInHAE8DZQCowRkRSix22EbgSeKWES+xT1XRvG+kjzshLT4elS6HQps4y\nxsQWP8niQmAWcJaq/gQ0w999Fv2Ataq6TlUPAlMo1jGuqutVdSkQGysHpafDvn3wzTeRjsQYY6qU\nn+k+9orIt8BZInIW8Kmqvu/j2m1x63cH5AD9yxFbgogsBAqA+1X1jeIHiMhYYCxA69atyc7OLsfl\nq16jQ4f4GbDilVfYevrppR6Xn58f8VjDKZbLZ2WLXrFcvuooW8hkISK/Ba4DXvd2vSwiE1X1H2GN\nDDqo6iYR6Qx8JCLLVPXb4ANUdSIwESAzM1OzsrLCHFIIp5wCN9xA6oEDpJYRS3Z2NhGPNYxiuXxW\ntugVy+WrjrKFTBbANUB/Vd0DICIPAHOBUMliE9A+6Hk7b58vqrrJ+7lORLKBDODbMk+KtPh46NnT\nOrmNMTHHT5+FcPRiR4XevlAWACeISCcRqQdcDPga1SQiTb2OdESkBXAKsMLPuREXGBGlGulIjDGm\nyvhJFi8C80TkbhG5G/gCd69FmVS1ALgR1zm+EnhVVZeLyHgRGQkgIj8TkRzgAuAZEVnund4dWCgi\nS4A5uD6L6EgWGRmwbRts2RLpSIwxpsr46eB+2GsGGujtukpVfd2mrKozgBnF9v056PECXPNU8fM+\nB3r5eY8aJ/hO7hS70d0YExvKTBbevRLLVfUkYHH1hBTl0tLcz6++ghEjIhuLMcZUkTKboVS1EFgt\nIsdXUzzRr3Fj6NrVOrmNMTHFz2iopsByEZmPm3EWgKi5qzoSbG0LY0yM8ZMs/hT2KGJNejpMnQq7\nd7uahjHGRLlSk4WIdAVaq+rHxfYPBGyoT1kyMtzPpUth4MCyjzXGmChQVp/FBGB3Cft3ea+Z0tja\nFsaYGFNWsmitqsuK7/T2dQxbRLGgTRto2dI6uY0xMaOsZNGkjNcaVHUgMcXWtjDGxJiyksVCEbmu\n+E4RuRZYFL6QYkRGhluP+9ChSEdijDGVVtZoqJuB6SJyKUXJIRO3St754Q4s6qWnw8GDsHJl0Y16\nxhgTpUpNFqr6I/BzERkC9PR2v6uqH1VLZNEu0Mn91VeWLIwxUc/P3FBzcJP5mfI48URo0MAli1/+\nMtLRGGNMpfiZddZURFycq1HY8FljTAywZBFOtraFMSZGWLIIp4wM+Okn2Lgx0pEYY0yllDXdRx5Q\n0ldiAVRVbdKjUILv5O7QIbKxGGNMJZRas1DVJFVtXMKWZInCp169oE4duznPGBP1/Mw6C4CItAIS\nAs9V1dpWQmnYELp1s2RhjIl6IfssRGSkiKwBvgM+BtYDM8McV+ywtS2MMTHATwf3X4ABwDeq2gk4\nHfgirFHFkvR018G9Y0ekIzHGmArzkywOqWouUEdE6ng36WWGOa7YEejkXrIksnEYY0wl+EkWP4lI\nIvAJMElEHiVoeVUTgq1tYYyJAX6SxShgL3AL8B7wLXBuOIOKKa1aQUqKdXIbY6Kan9FQrYAtqrof\n+JeINABaA7lhjSyW2NoWxpgo56dm8RpwOOh5obfP+JWR4aYq378/0pEYY0yF+EkWdVX1YOCJ97he\n+EKKQenpUFAAy5dHOhJjjKkQP8lim4iMDDwRkVHA9vCFFIOC17Ywxpgo5KfP4le4UVCP4+aF+h6w\nBRrKo3NnSEqyZGGMiVp+Fj/6FhjgDZ9FVfPDHlWsqVMHeve24bPGmKhVajOUiFzm/bxVRG4FxgJj\ng56HJCLDRGS1iKwVkTtKeH2QiCwWkQIRGV3stStEZI23XVG+YtVA6enuxrzDh0Mfa4wxNUxZfRaN\nvJ9JpWxlEpE44AngbCAVGCMiqcUO2whcCbxS7NxmwDigP9APGCciTUO9Z42WkQH5+bBuXaQjMcaY\nciu1GUpVn/E+8Her6iMVuHY/YK2qrgMQkSm4G/xWBL3Heu+14l+3zwJmq+oO7/XZwDBgcgXiqBmC\n7+Ru2TKysRhjTDmV2WehqoUiMgaoSLJoi+sMD8jB1RQqem7b4geJyFhc8xitW7cmOzu7AmFWjzoH\nDzIwLo7v33yT/IsvrtGxVlZ+fn7Mls/KFr1iuXzVUTY/o6H+642E+g9Bc0Kp6uKwReWTqk4EJgJk\nZmZqVlZWZAMKJTWVDjt38l1iIjU+1krIzs6O2fJZ2aJXLJevOsrmJ1l47SeMD9qnwGkhztsEtA96\n3s7b58cmIKvYudk+z625MjJg9uxIR2GMMeXmZ+jskApeewFwgoh0wn34Xwxc4vPcWcBfgzq1zwT+\nUME4ao70dHjpJeJtbQtjTJTxs1Jesog8LCILve3vIpIc6jxVLQBuxH3wrwReVdXlIjI+cEe4iPxM\nRHKAC4BnRGS5d+4O3KJLC7xwzq+BAAAZkklEQVRtfKCzO6p5ndyJ334b4UCMMaZ8/DRDvQB8DVzo\nPb8ceBH4RagTVXUGMKPYvj8HPV6Aa2Iq6dwXvPeOHYFksXZthAMxxpjy8TM3VBdVHaeq67zt/4DO\n4Q6suqjCU0+5Ea2FhWF+s6ZNoUMHSxbGmKjjp2axT0QGqupnACJyCrAvvGFVnw0b4IYb3OOmTWHQ\nIMjKgiFDoFcvN1NHlUpPJ9Gm/TDGRBk/yeJ63KJHybiJBHfg7rqOCR07wvffw8cfw5w5kJ0Nb77p\nXmvaFAYPdokjKwt69qyC5JGRQcO33oIPPoChQyt5MWOMqR5+RkN9BfQWkcbe891hj6qatWsHl17q\nNnDJIzu7aHvjDbe/eXOXPAI1j9TUCiSPK65g3/PP0/CMM+C66+Bvf4PkkOMFjDEmokImi+KTBooI\nwC5gkZdIYk779nD55W4D11QVXPN4/XW3v0WLo2seqangfj1l6NiRhc8+y6APP4S//x1mzoRnn4Vh\nw8JYImOMqRw/34szcWtatPW2/8HN0/SsiPxvGGOrMTp0gF/+El58Eb77zm0vvggjRsCCBXDjja6J\nqnVruPBCePJJt4qqasnXO1y/Pjz4IHz+OTRuDGefDVddBTt3Vm/BjDHGJz/Joh3QR1V/p6q/A/oC\nrYBBxFDfRXl07AhXXgn//CesX+8mkn3+eVc5mDsXfv1rV8s47ji46CJ4+mlYtaqE5NG/PyxeDHfe\nCf/+N/ToAW+/Xe3lMcaYUPwki1bAgaDnh4DWqrqv2P5aSQQ6dYKrr4aXXoKNG2HtWnjuOTjjDPjs\nM7j+eujeHVJSYMwYePfd49i1y7tA/fpw770wf76bjXbkSNd5kpsb0XIZY0wwP8liEjBPRMaJyDjg\nv8ArItKIoOnGjSMCXbrANdfAyy9DTg588w1MnAinneb6Ph566CRSUtwx8+d7NY4+fVyb1t13w6uv\nuqrJtGmRLo4xxgA+koWq/gU3DfhP3vYrVR2vqntU9dJwBxjtROCEE9zAp0mTYNMmePLJRYwZA1Om\nuJaoPn3cjYG79tWDceNg0SI3RGv0aNcJsnVrpIthjKnl/A78TMAtgvQosMGbHNBUgAh0757Hc8/B\nli2uM1zV3RiYkgLXXgvz96ehc79wzVNvvun6MqZMKb3H3BhjwszPRILjgN9TNOtrPPByOIOqLRo3\ndv0ZX34J8+a5/ozJk73aRv94nmp6J/mfLIbOnd2Lv/gF/PBDpMM2xtRCfmoW5wMj8RY+UtXN+FiD\n2/gnAv36uU7xzZuPrm20Pq0HY3v8l+9/8yA6c6bry/j3v62WYYypVn6SxUFVVdyCR3gd2yZMkpOP\nrW1M+k9djv/H7ZzfaQlbmnZ3N32ce67rADHGmGrgJ1m8KiLPAE1E5DrgA+C58IZlSqptrK/fjXbr\nPuF/603gwHsfUdAtFX3ueatlGGPCzs9oqIeAqcA0oBvwZ1V9LNyBmSLBtY258+LIvey39I1fxmd7\nMpDrruX7HsPIW74x0mEaY2KYnw7uB1R1tqrerqq3qepsEXmgOoIzRwvUNp5/Hv77QxdW/OMj7k15\ngqYr/ws9e/DywKdZMO+wVTSMMVXOTzPUGSXsO7uqAzHlk5wMN9xYhztzbmDN61+zMWUAl/33evIG\nDGVkz3U8/TTsjrn5gY0xkVJqshCR60VkGdBNRJYGbd8BS6svRFMWEcg4vyM9ct5n76MTGZiwkFdX\n9mL59f+gbZvDXHcdfPIJHDwY6UiNMdGsrJrFK8C5wFvez8DWV1Uvq4bYTHmI0PCm66j3zXISzhrE\nP7iJRUlZzHt5DYMHQ7NmcM45MGECLF9ufeLGmPIpNVmo6i5VXa+qY1R1A24pVQUSReT4aovQlE/7\n9siMGfDii5x4YBlLSGPVL+7kj8O/5JvVyi23uOnU27Z1I3BfesmNtjLGmLL46eA+V0TWAN8BHwPr\ngZlhjstUhoibQ335cmT4cLq98QB/eK0P3xR0ZtfVt/DO7R8zeGAhM2fCFVe4xNGjB9x8M7z7LuTl\nRboAtdeBA/DOO+7v0ro1/PzncN99Vhs0keeng/seYADwjap2Ak4HvghrVKZqpKS4mWt/+MENoerZ\nk8aTnmLE37KYPOc4tp5zNWsfeZuH791H27bwzDOuqapZMxg0CMaPd+tzFBREuiCx7eBBl6SvvNIl\niHPPdVOCDRniksedd7raYJcu8NvfuuXbrQ/KVLeQy6oCh1Q1V0TqiEgdVZ0jIhPCHpmpOi1bugU3\nrr7aVRtmzYLp05Hpr9Plny9yS6NG3DJsGAefPJ+5zUbw3hdNmD3bzZY+bpybw2rIEBg61K3RceKJ\nPpaPNWU6eNB96L/6qksMP/3kRridd56baHjoUKhXzx27aZOrbbz9tpvq/rHH3N/krLNcYhk+3K0P\nb4514IAb4JGdDfn5bWncGNLSoK6fTz5zFD+/sp9EJBH4BJgkIlvx5okyUSgpyU19Pnq0+8TKzobp\n0+HNN6k3bRqD69Zl8JAh3Hf1eez85yg+WNmWDz6A2bPdhxq4NcqHDnXb6ae7b8MmtECCeO01eOON\noxPEBRe432f9+see17Yt/M//uG3PHneNt992CeS116BOHTjlFJc4zj0XunWr3cl8wwa3tP2MGfDh\nh7B3r/sdHT58Ao89Bg0busk6Tz7ZNfMNGGDJ1hdVLXMDGuGaq+oCVwA3Ac1DnVfdW9++fTVazJkz\nJ9IhHKuwUHXuXNXf/171xBNVXRO5av/+qvfdp7pqlX77rerTT6uOHq3atGnRIWlpqr/7nerMmap7\n9tTQ8lWR8pbtwAHVGTNUr7pKtUkT9/tKTlb95S9V33lHdf/+isdSWKg6b57qXXep9u5d9Pfo2lX1\nlltUP/pI9eBB/9eL1r/bgQOurLfdptqjR9HvoWNH1RtucL/nPXtUJ0/+XF95RfU3v1Ht21c1Lq7o\n2G7d3N/o2WdVly93v9toUpm/HbBQfXzGipbSayYiXXHLp/632P6BwBZV/TaMOazcMjMzdeHChZEO\nw5fs7GyysrIiHUbpVN2i4dOnuy3wez3pJDj/fDjvPAozMvlySR1mz3bfdD/7zH1zjo+HJk32c9xx\nCTRtSrm2QLNLTebnb3fwoPtGG6hB7Nzpmo0CNYgzzii5BlFZGzYUNVfNmePiaNIEzj7b1TiGDXO/\n59LU+H+XQTZtcrWHmTNdrTcvz/3bGzTINcsNH35sDat4+fbscf+0P/+8aNuxw73WpImrcfz8527r\n189VymuqyvztRGSRqmaGOq6sZqgJFK1hEWyX99q5FYrM1HxuhSa33XknfP+9a4N64w148EG47z7i\n2rYl87zzyDzvPP5w22D2Horns89cq9aXX+6kfv027NwJ69a5D8udO91/zrI0bFi+5NK0qeuMb948\n8m3Qhw4VJYjp04sSxKhRrg8iXAkiWIcO8Otfuy0vz32Ivv226zyfPBni4uDUU4uaq044IbzxVKWC\nAjfYItC8tGSJ29+unZuZefhwt2xxeT7QGzWCwYPdBu470po1LmnMnet+jhvn9tepA716ucQRaL7q\n3Ll2NfeVVbNYoKo/K+W1ZaraK6yRlZPVLKrJjh3u02f6dHjvPdi3z30NO+ccV+s46yyyFywosXwH\nD7p2+kDyKM+Wn196SCIuabRqVfLWsuXRz5s0qfh/8uC/3aFD8NFHrpP6jTfcryaQIC64AM48M/wJ\nwo/CQrfW+9tvu+3rr93+bt1c0hg50n0AfvZZzfp3+eOP7p/YjBnw/vvu307duq5/JlB76NHD/9+y\nIv/vdu1ySwUEah5ffFE0tLxVq6OTR9++0KBB+cpYVaqjZlFWslijqiV+9xCRtara1UcQw4BHgTjg\nOVW9v9jr9YGXgL5ALnCRqq4XkY7ASmC1d+gXqvqrst7LkkUE7N3rvr5On+4+hXbsgIQEfurWjSZn\nn13Ue9iyZaXf6tChkhPNjh2wbZtbpjx427atqEmhuPj4oxNI8WRSPNE0ClrB5YMPPqawcPCRGkQg\nQYwc6WoQNSVBlOW774oSx8cfu99ts2bQq9dWevduRfPmHNkCNbfAlpgYvm/TgaQWqD0sWuT2t2nj\nmtKGD3eDAJKTK3b9qvh/V1gIK1YcXftYs8a9Fh8PGRlFzVZdu0LHjtCiRfhrIJFOFpOBj1T12WL7\nrwXOUNWLQgQQB3yDm4gwB1gAjFHVFUHH3ACkqeqvRORi4HxVvchLFu+oas9QBQiwZBFhBQXw6afw\n1lvsfu89Gq9dW3SDRteu7utX4CtYz56uTSTMDh2C7duPTSSBZFJ8X2nNZA0bFiWPVasOsXt3PElJ\nR9cgEhLCXpyw2L3bjaR++2344IN97NnToMwJKOPjS04ipSWXwL7S+qO2b3fvP2OG+5mb65p8Tj65\nqPbQu3fVfNiG6//dtm2uxhGofSxY4CrcAY0auaRRfOvUyf1s1qzy5Yt0n8XNwHQRuRTwcjyZQD3c\nUquh9APWquo6L6ApwChgRdAxo4C7vcdTgcdFalMrYAypW9fdjDFkCItHjSKrf3/Xezh3rtvef98t\nBwvu62m/fkUJJExjF+Pj3bfSNm38Hb9nj/uPX1IiCWz9++dy443HRXWCCNa4sUt4F1wA2dnzyMrK\n4tAhV2PKzXVb8OPi+9audc00ubll3yiYmHhsEvnuO1eTUHU1uBEjXHI44wz3ARotWrYs6gcC9yVl\n1SpYv96Vcf36ou3zz12NOFhi4rEJJHhr2rRm9I2UWrM4coDIECDwDX+5qn7k68Iio4Fhqnqt9/xy\noL+q3hh0zNfeMTne82+B/kAisBxXM9kN3KWqn5bwHmOBsQCtW7fuO2XKFD+hRVx+fj6JiYmRDiNs\nSiyfKglbttB4+XKSV6yg8fLlJH77LXL4MAB727dnV48e7E5NZXfPnuzp0MF9xaxhYvlvV5myqcL+\n/XXYvTv+yLZrV13y8gLP67Jrl3ucl1eX3bvjadz4EP367WDAgB2ccEJe2P/cNeVvl59flx9+qM8P\nPyR4W4Ogxwns2XP0d/hGjQpo3Xo/xx1XtLVpU/Q4MbGgUmUbMmRIpWsWAKjqHGBOhaKouC3A8eru\nHO8LvCEiPVT1qAqyqk4EJoJrhoqWpp2YbIYK4rt8e/a4OvvcuTT0tjbvvedea9y46M6pQO2jSZOw\nxu1HLP/tIlO2ZKBTtbxTtPztdu48ujby3Xd1Wb8+kfXrE1m69NjBHsnJkJa2jU8+qXzfYFnCOeBw\nE9A+6Hk7b19Jx+SISF3cv5xc70aRAwCqusircZwIREenhPGnUSPIynIbuK+na9cWNV3NnQv33ANe\n7YPU1KLkcfLJ7r6PGlj7MKYyAsPCMzKOfU3VNQEGJ5P16yEvbw8QvcliAXCCiHTCJYWLgUuKHfMW\n7q7wucBoXIe6ikhLYIeqFopIZ+AEYF0YYzU1gYgb/H/CCW7+dHDjFOfPL0oe06e7SRHB1TQyM11D\n7/HHH721axcdd/kZUw4iRX0+ffsW7c/OXg90DOt7hy1ZqGqBiNwIzMINnX1BVZeLyHjc7eVvAc8D\n/xaRtcAOXEIBGASMF5FDwGHgV6paykBIE9OSktwEVKef7p6rwjffFI1d/PJLd8Pg1q1HnycCxx13\nbBIJ3po3rxk9h8ZEgbDe96qqM4AZxfb9OejxfuCCEs6bBkwLZ2wmSom4u8m6dYOrrirav28f5OTA\nxo3HbkuWuLGh+/cffa0GDcpOJu3axcaQJ2OqgE3Ua2JDgwZFTVglUXWD+ktKJhs3urvSf/jh2PNa\ntz4qgaQcPuyGCaeluU54Y2oJSxamdhBxA+Jbtjy6sTfYgQOl105WrICZMzlx71549FF3fKdO7o6x\n3r1d8ujd2+2zTncTgyxZGBNQv75bjq5Ll5JfV2Xua69xcsOGsHSpa95assT1mQTuV0pMdIkjkDx6\n93Yz0NWA8f3GVIYlC2P8EuFAq1ZuqO855xTt37vXzc4XnEAmT4annz5yHl26HJ1Aevd208RaB7uJ\nEpYsjKmshg3d9CX9+hXtUy3qXA9sS5e6ob+BWkhgjc/gpqyePY+eudCYGsKShTHhIOJqDh06uClp\nA/LzXS0kkDyWLIGXXiqa9zpwr0kgeXTp4taxbd8eUlLchFfGRIAlC2OqU2Kim7pkwICifYcPu9tw\ng5uxFi1yKykFq1On6N6RQAIJbIF9rVpZB7sJC0sWxkRanTpu2bXOnd3aqwH5+a4p6/vv3Rb8eMkS\nt4Zq8FzY4Goe7dodm0SCt5oyjamJKpYsjKmpEhPdfFipqSW/rurmBg8kkOBt40a3MPqmTUXrigQ0\nanRsAmnfnma5ue4mxOTkoq1hQ0ssBrBkYUz0EnHLsLVoUfKsc+CWdvvxx2NrJoFt2TL3uippJZ0f\nF+c64oMTSHJyyftK25+UZE1jMcCShTGxLC7OdYynpLgp30ty8CBs2sTid96hT5cubuHpXbvcMnqB\nx8Hb998f/bywMHQcSUnHJpHERHfnfUJC1fysX99qQWFkycKY2q5ePejUid29ehVNF++XqrvPpKzk\nUtL+rVvdMnL79rk5uwI/y1puz4/SEkmDBvRUhR493BQuJW2NG1uyKYMlC2NMxYm4PpBGjVztpbIK\nC920K8WTyL59Je/z+3PvXhI2bHAzFm/fXnSvS7CEhNITSfGtSZNal1gsWRhjao64ONep3rBhlV96\nYWClvIIClzB+/LH0beNGt4rj1q1Fi28Fq1fPDVM+7riSk0nLlq6m0rixa4JLSnIJNYr7bixZGGNq\nl7p13Yf8cceFPraw0I04KyuxbN7s1lXZuvXYkWfBArWw4AQS2Cqyr5pv0LRkYYwxpYmLczWIVq3c\nhJBlOXzYLaD944+wbZvrq8nLO3oraZ9bF7VoO3DAX2z16x9JIN07dix/f1M5WbIwxpiqUKdO0Zqn\nlXHoUOmJpZR9+0vqg6liliyMMaYmiY+HZs3c5tN32dl0CGNIANHb22KMMabaWLIwxhgTkiULY4wx\nIVmyMMYYE5IlC2OMMSFZsjDGGBOSJQtjjDEhWbIwxhgTkmg13PlXHURkG7Ah0nH41ALYHukgwiiW\ny2dli16xXL7KlK2DqrYMdVDMJItoIiILVTUz0nGESyyXz8oWvWK5fNVRNmuGMsYYE5IlC2OMMSFZ\nsoiMiZEOIMxiuXxWtugVy+ULe9msz8IYY0xIVrMwxhgTkiULY4wxIVmyqEYi0l5E5ojIChFZLiK/\njXRMVU1E4kTkSxF5J9KxVDURaSIiU0VklYisFJGTIx1TVRGRW7x/k1+LyGQRSYh0TJUhIi+IyFYR\n+TpoXzMRmS0ia7yfTSMZY0WVUra/ef8ul4rIdBFpUtXva8miehUAv1PVVGAA8GsRSY1wTFXtt8DK\nSAcRJo8C76nqSUBvYqScItIWuAnIVNWeQBxwcWSjqrR/AsOK7bsD+FBVTwA+9J5Ho39ybNlmAz1V\nNQ34BvhDVb+pJYtqpKpbVHWx9zgP92HTNrJRVR0RaQeMAJ6LdCxVTUSSgUHA8wCqelBVf4psVFWq\nLtBAROoCDYHNEY6nUlT1E2BHsd2jgH95j/8FnFetQVWRksqmqu+raoH39AugXVW/ryWLCBGRjkAG\nMC+ykVSpCcD/AocjHUgYdAK2AS96zWzPiUijSAdVFVR1E/AQsBHYAuxS1fcjG1VYtFbVLd7jH4DW\nkQwmjK4GZlb1RS1ZRICIJALTgJtVdXek46kKInIOsFVVF0U6ljCpC/QBnlLVDGAP0duMcRSv7X4U\nLiGmAI1E5LLIRhVe6u4ZiLn7BkTkj7jm7klVfW1LFtVMROJxiWKSqr4e6Xiq0CnASBFZD0wBThOR\nlyMbUpXKAXJUNVATnIpLHrFgKPCdqm5T1UPA68DPIxxTOPwoIm0AvJ9bIxxPlRKRK4FzgEs1DDfQ\nWbKoRiIiuDbvlar6cKTjqUqq+gdVbaeqHXGdox+pasx8O1XVH4DvRaSbt+t0YEUEQ6pKG4EBItLQ\n+zd6OjHSeV/MW8AV3uMrgDcjGEuVEpFhuCbgkaq6NxzvYcmiep0CXI771v2Vtw2PdFDGt98Ak0Rk\nKZAO/DXC8VQJr7Y0FVgMLMN9LkT11BgiMhmYC3QTkRwRuQa4HzhDRNbgalP3RzLGiiqlbI8DScBs\n73Pl6Sp/X5vuwxhjTChWszDGGBOSJQtjjDEhWbIwxhgTkiULY4wxIVmyMMYYE5IlC2NqABHJisWZ\nek3ssGRhjDEmJEsWxpSDiFwmIvO9G5+e8dbvyBeRR7z1ID4UkZbeseki8kXQGgNNvf1dReQDEVki\nIotFpIt3+cSg9TImeXdTG1MjWLIwxicR6Q5cBJyiqulAIXAp0AhYqKo9gI+Bcd4pLwG/99YYWBa0\nfxLwhKr2xs3BFJgJNQO4GUgFOuPu+DemRqgb6QCMiSKnA32BBd6X/ga4yegOA//xjnkZeN1b/6KJ\nqn7s7f8X8JqIJAFtVXU6gKruB/CuN19Vc7znXwEdgc/CXyxjQrNkYYx/AvxLVY9ahUxE/lTsuIrO\noXMg6HEh9v/T1CDWDGWMfx8Co0WkFRxZ07kD7v/RaO+YS4DPVHUXsFNETvX2Xw587K2QmCMi53nX\nqC8iDau1FMZUgH1zMcYnVV0hIncB74tIHeAQ8GvcQkj9vNe24vo1wE2D/bSXDNYBV3n7LweeEZHx\n3jUuqMZiGFMhNuusMZUkIvmqmhjpOIwJJ2uGMsYYE5LVLIwxxoRkNQtjjDEhWbIwxhgTkiULY4wx\nIVmyMMYYE5IlC2OMMSH9f0W2AIrHjv90AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "UkQMnunpqb7T"
      },
      "source": [
        "### CNN WITH 5 LAYERS AS SIGMOID ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "p-Xoi65Eqb7U"
      },
      "source": [
        "### USING BN+DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "vei4fOnzqb7U",
        "outputId": "07395646-414d-4c69-9f3f-4e5bc005406f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(Dropout(0.20))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.20))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_16\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_52 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_52 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_53 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "dropout_16 (Dropout)         (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_53 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_54 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "batch_normalization_12 (Batc (None, 10, 10, 32)        40        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_54 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_55 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "batch_normalization_13 (Batc (None, 9, 9, 32)          36        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_55 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_56 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_56 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "dropout_17 (Dropout)         (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_16 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_31 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_16 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_18 (Dropout)         (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_32 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,134\n",
            "Trainable params: 76,096\n",
            "Non-trainable params: 38\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "s9h6NfhHqb7V",
        "outputId": "2b5dad72-b18d-40a2-d484-fa0808ec631a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 10s 173us/step - loss: 2.4012 - acc: 0.1036 - val_loss: 2.3044 - val_acc: 0.1028\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 8s 129us/step - loss: 2.3268 - acc: 0.0999 - val_loss: 2.3069 - val_acc: 0.0958\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 8s 129us/step - loss: 2.3136 - acc: 0.1036 - val_loss: 2.3033 - val_acc: 0.1028\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3084 - acc: 0.1033 - val_loss: 2.3023 - val_acc: 0.1135\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3069 - acc: 0.1063 - val_loss: 2.3018 - val_acc: 0.1135\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3056 - acc: 0.1060 - val_loss: 2.3018 - val_acc: 0.1135\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3051 - acc: 0.1051 - val_loss: 2.3026 - val_acc: 0.1135\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3042 - acc: 0.1095 - val_loss: 2.3018 - val_acc: 0.1135\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 8s 133us/step - loss: 2.3032 - acc: 0.1074 - val_loss: 2.3015 - val_acc: 0.1135\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 2.3030 - acc: 0.1096 - val_loss: 2.3015 - val_acc: 0.1135\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 8s 134us/step - loss: 2.3028 - acc: 0.1078 - val_loss: 2.3018 - val_acc: 0.1135\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 8s 129us/step - loss: 2.3028 - acc: 0.1093 - val_loss: 2.3016 - val_acc: 0.1135\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "nNYFUs4cqb7X",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "oEjD8Su3qb7Y",
        "outputId": "15809192-91b1-425f-8583-b0b9e37180c6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.03929063466354273\n",
            "Test accuracy: 0.9873\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLCkYY4ypY0nBGGNMnf8PZt0BHuZvpUYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Gl2gEC2Hqb7a"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "7Cd8G7Faqb7a",
        "outputId": "9c23340d-8b67-4b5a-8815-3b63a87d9c93",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 722
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(Dropout(0.20))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.20))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_17\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_57 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_57 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_58 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "dropout_19 (Dropout)         (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_58 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_59 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_59 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_60 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_60 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_61 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_61 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "dropout_20 (Dropout)         (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_17 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_33 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_17 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_21 (Dropout)         (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_34 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,058\n",
            "Trainable params: 76,058\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "X6rFjLAUqb7c",
        "outputId": "dae210d0-e01b-432f-ed66-423f19104835",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 10s 170us/step - loss: 2.4015 - acc: 0.1020 - val_loss: 2.3055 - val_acc: 0.1032\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 2.3254 - acc: 0.1023 - val_loss: 2.3061 - val_acc: 0.1028\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 2.3138 - acc: 0.1026 - val_loss: 2.3021 - val_acc: 0.1028\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 7s 125us/step - loss: 2.3090 - acc: 0.1046 - val_loss: 2.3033 - val_acc: 0.1009\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 7s 125us/step - loss: 2.3065 - acc: 0.1036 - val_loss: 2.3024 - val_acc: 0.1135\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 7s 124us/step - loss: 2.3053 - acc: 0.1056 - val_loss: 2.3023 - val_acc: 0.1135\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 7s 124us/step - loss: 2.3049 - acc: 0.1057 - val_loss: 2.3016 - val_acc: 0.1135\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 2.3040 - acc: 0.1073 - val_loss: 2.3021 - val_acc: 0.1135\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 2.3036 - acc: 0.1075 - val_loss: 2.3017 - val_acc: 0.1135\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 2.3034 - acc: 0.1092 - val_loss: 2.3016 - val_acc: 0.1135\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 2.3028 - acc: 0.1090 - val_loss: 2.3026 - val_acc: 0.1135\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 2.3028 - acc: 0.1087 - val_loss: 2.3017 - val_acc: 0.1135\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "sSaajoPUqb7e",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "W7Sx0uQKqb7g",
        "outputId": "d3570538-ae8c-4c44-b06b-37c501a5d334",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.03929063466354273\n",
            "Test accuracy: 0.9873\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLCkYY4ypY0nBGGNMnf8PZt0BHuZvpUYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "kj63DNBNqb7h"
      },
      "source": [
        "### WITHOUT DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "QXtSK7XSqb7h",
        "outputId": "efd4f2bf-beff-4665-ac22-1ab9762bf693",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 689
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Second layer\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#third layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth Layer\n",
        "model.add(Conv2D(32,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Fifth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512))\n",
        "model.add(Activation('sigmoid'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_18\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_62 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_62 (MaxPooling (None, 26, 26, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_63 (Conv2D)           (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_63 (MaxPooling (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_64 (Conv2D)           (None, 10, 10, 32)        18464     \n",
            "_________________________________________________________________\n",
            "batch_normalization_14 (Batc (None, 10, 10, 32)        40        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_64 (MaxPooling (None, 10, 10, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_65 (Conv2D)           (None, 9, 9, 32)          4128      \n",
            "_________________________________________________________________\n",
            "batch_normalization_15 (Batc (None, 9, 9, 32)          36        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_65 (MaxPooling (None, 4, 4, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_66 (Conv2D)           (None, 3, 3, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_66 (MaxPooling (None, 1, 1, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_18 (Flatten)         (None, 16)                0         \n",
            "_________________________________________________________________\n",
            "dense_35 (Dense)             (None, 512)               8704      \n",
            "_________________________________________________________________\n",
            "activation_18 (Activation)   (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_36 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 76,134\n",
            "Trainable params: 76,096\n",
            "Non-trainable params: 38\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "kz_NB0e_qb7i",
        "outputId": "77bb9b3d-c8dd-48c4-e5e8-651ef79a6752",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 10s 161us/step - loss: 2.1318 - acc: 0.2149 - val_loss: 1.2405 - val_acc: 0.6568\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 7s 111us/step - loss: 0.3614 - acc: 0.9150 - val_loss: 0.2074 - val_acc: 0.9415\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 7s 112us/step - loss: 0.1606 - acc: 0.9553 - val_loss: 0.1580 - val_acc: 0.9537\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 7s 112us/step - loss: 0.1218 - acc: 0.9644 - val_loss: 0.1252 - val_acc: 0.9613\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 7s 113us/step - loss: 0.1055 - acc: 0.9693 - val_loss: 0.1071 - val_acc: 0.9662\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 7s 114us/step - loss: 0.0901 - acc: 0.9735 - val_loss: 0.1035 - val_acc: 0.9685\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 7s 112us/step - loss: 0.0819 - acc: 0.9754 - val_loss: 0.0876 - val_acc: 0.9721\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 111us/step - loss: 0.0738 - acc: 0.9769 - val_loss: 0.0764 - val_acc: 0.9763\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 7s 112us/step - loss: 0.0682 - acc: 0.9799 - val_loss: 0.0794 - val_acc: 0.9772\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 7s 111us/step - loss: 0.0624 - acc: 0.9812 - val_loss: 0.0803 - val_acc: 0.9756\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 7s 111us/step - loss: 0.0593 - acc: 0.9822 - val_loss: 0.0753 - val_acc: 0.9780\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 7s 110us/step - loss: 0.0558 - acc: 0.9832 - val_loss: 0.0719 - val_acc: 0.9779\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "f4WEAN7Fqb7j",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "EYdvPNYXqb7l",
        "outputId": "a3f568bc-2d4c-45dc-ac7c-e540da2ef273",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.07189704030742869\n",
            "Test accuracy: 0.9779\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "CUuS1kP0qb7n",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "zswdtGECmWY5"
      },
      "source": [
        "### CNN WITH 6 LAYERS ASA RELU ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "4QVntOqmqb7r"
      },
      "source": [
        "### BN+ DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "5mKsKDwfTLL_",
        "outputId": "dbe60209-e6ad-47dc-e6a6-39f823c29f4c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 890
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(Dropout(0.25))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_19\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_67 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_67 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_68 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_16 (Batc (None, 22, 22, 64)        88        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_68 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_69 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "dropout_22 (Dropout)         (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_69 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_70 (Conv2D)           (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "dropout_23 (Dropout)         (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_70 (MaxPooling (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_71 (Conv2D)           (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "batch_normalization_17 (Batc (None, 6, 6, 16)          24        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_71 (MaxPooling (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_72 (Conv2D)           (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "batch_normalization_18 (Batc (None, 5, 5, 16)          20        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_72 (MaxPooling (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_19 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_37 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_19 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dropout_24 (Dropout)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_38 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,158\n",
            "Trainable params: 122,092\n",
            "Non-trainable params: 66\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "eo_7q34iTutw",
        "outputId": "f6211c1b-794b-44f2-9dfd-ae1c8585629a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 11s 187us/step - loss: 0.4371 - acc: 0.8586 - val_loss: 0.0679 - val_acc: 0.9780\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.1180 - acc: 0.9648 - val_loss: 0.0482 - val_acc: 0.9842\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0846 - acc: 0.9754 - val_loss: 0.0401 - val_acc: 0.9874\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0733 - acc: 0.9791 - val_loss: 0.0354 - val_acc: 0.9898\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0636 - acc: 0.9815 - val_loss: 0.0306 - val_acc: 0.9908\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0569 - acc: 0.9835 - val_loss: 0.0329 - val_acc: 0.9898\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0517 - acc: 0.9850 - val_loss: 0.0292 - val_acc: 0.9921\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 125us/step - loss: 0.0486 - acc: 0.9861 - val_loss: 0.0253 - val_acc: 0.9912\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0461 - acc: 0.9863 - val_loss: 0.0271 - val_acc: 0.9929\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0411 - acc: 0.9876 - val_loss: 0.0303 - val_acc: 0.9910\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0392 - acc: 0.9881 - val_loss: 0.0321 - val_acc: 0.9904\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0389 - acc: 0.9883 - val_loss: 0.0251 - val_acc: 0.9928\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "SihLwMX7XYkL",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "lkzyrqNpXbyt",
        "outputId": "74c89db6-b644-4d33-85e2-cb335cb8ca7b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.02512707673313598\n",
            "Test accuracy: 0.9928\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTMz/Bx16Uz2mYvOgAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "E_XnJVemqb7z"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "YAJqiOotqb70",
        "outputId": "decaa411-4356-476f-8713-bd3f9a189d40",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(Dropout(0.25))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_20\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_73 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_73 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_74 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_74 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_75 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "dropout_25 (Dropout)         (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_75 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_76 (Conv2D)           (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "dropout_26 (Dropout)         (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_76 (MaxPooling (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_77 (Conv2D)           (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_77 (MaxPooling (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_78 (Conv2D)           (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_78 (MaxPooling (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_20 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_39 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_20 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dropout_27 (Dropout)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_40 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,026\n",
            "Trainable params: 122,026\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ilCB6YeUqb71",
        "outputId": "1d81c68a-0dce-4665-8db1-79b607c7c78b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 10s 163us/step - loss: 0.4735 - acc: 0.8461 - val_loss: 0.0686 - val_acc: 0.9783\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 6s 101us/step - loss: 0.1209 - acc: 0.9649 - val_loss: 0.0562 - val_acc: 0.9824\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0902 - acc: 0.9741 - val_loss: 0.0354 - val_acc: 0.9882\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0776 - acc: 0.9775 - val_loss: 0.0365 - val_acc: 0.9887\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0682 - acc: 0.9804 - val_loss: 0.0372 - val_acc: 0.9887\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0606 - acc: 0.9826 - val_loss: 0.0269 - val_acc: 0.9913\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 0.0560 - acc: 0.9841 - val_loss: 0.0228 - val_acc: 0.9936\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 0.0509 - acc: 0.9853 - val_loss: 0.0297 - val_acc: 0.9912\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0495 - acc: 0.9857 - val_loss: 0.0228 - val_acc: 0.9927\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 0.0469 - acc: 0.9862 - val_loss: 0.0225 - val_acc: 0.9931\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0437 - acc: 0.9875 - val_loss: 0.0261 - val_acc: 0.9908\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 0.0416 - acc: 0.9886 - val_loss: 0.0234 - val_acc: 0.9930\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "L0jDzeomqb72",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "V24Eg0t3qb76",
        "outputId": "1dc4f963-5621-493c-9f37-464c134e2581",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.023400390856726152\n",
            "Test accuracy: 0.993\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl8FeW5wPHfkxAIBAg7GkChgkAg\nh1XEuoHY1rrWiguirSut91rrWq211atd1HoVW73WvVpRtCCtVby4EZdeFxYVZBMU1IACibKEsIU8\n9493TnKSnCST5Ewm55zn+/nMZ868M2fmeVnmOTPvO++IqmKMMcYAZIQdgDHGmNbDkoIxxphKlhSM\nMcZUsqRgjDGmkiUFY4wxlSwpGGOMqWRJwZg6iIiKyMCw4zCmJVlSMElBRNaJyE4RKY2Z7gk7rkQS\nkf5eImoTdiwmfdk/PpNMTlLVV8IOwphUZlcKJumJyHki8m8RuUdEtorIShGZFLM+T0SeE5GvRWSN\niFwcsy5TRK4XkU9EZLuILBKRfjG7P1ZEVovIFhG5V0QkzvHzvKuYbjFlo0SkWESyRGSgiLzuxVYs\nIk83oY7tRGS6iGzwpuki0s5b10NEnvdi/FpE3hSRDG/dtSKy3qvbqtg/F2PisaRgUsWhwCdAD+BG\n4NmYk/RMoAjIAyYDvxeRY7x1VwJTgOOBzsAFQFnMfk8EDgEiwBnA92oeWFU3AG8Dp8UUnw3MUtW9\nwC3AS0BXoC/w5ybU71fAeGAkMAIYB9zgrbvKq19PoDdwPaAiMhi4FDhEVTt5sa9rwrFNGrGkYJLJ\nP7xfw9Hp4ph1m4DpqrpXVZ8GVgEneL/6DweuVdVdqvoB8BDwI+97FwE3qOoqdT5U1ZKY/d6qqltU\n9XNgPu6kHM+TuOSCdzVxllcGsBc4EMjzYnirCXWfCtysqptUdTPwX8C5MfvfHzjQq/+b6gY12we0\nA/JFJEtV16nqJ004tkkjlhRMMvmBqnaJmR6MWbdeq4/u+BnuyiAP+FpVt9dY18f73A93hVGXr2I+\nlwEd69huNnCYiOwPHAVUAG96634BCPCeiCwTkQvqOV5d8ry4o6L1A/gjsAZ4SUQ+FZHrAFR1DXA5\ncBOwSURmikgextTDkoJJFX1q3O8/ANjgTd1EpFONdeu9z18ABzX34Kr6De4W0Zm4W0czo0lKVb9S\n1YtVNQ/4CfA/TejqugF3tREVrR+qul1Vr1LVbwEnA1dG2w5U9UlVPcL7rgK3NbmSJi1YUjCpohdw\nmdewezowFJirql8A/wf8QUSyRSQCXAg84X3vIeAWERkkTkREujcxhidxt6UmU3XrCBE5XUT6eovf\n4E7OFfXsp50Xa3TKAJ4CbhCRniLSA/hNtA4icqLXmC3AVtxtowoRGSwix3gN0ruAnQ0c1xjrkmqS\nyr9EZF/M8suqeqr3+V1gEFAMbAQmx7QNTAH+gvtl/Q1wY0zX1jtx991fwjVSrwSi+2ys53BJ5nNV\n/TCm/BBguojkerH9XFU/rWc/pTWWvwP8FtcQvsQr+7tXBq7e9+Aamr8B/kdV53sJ8FZcgtyLS47T\nmlg3kybEXrJjkp2InAdc5N0mMcY0g90+MsYYU8mSgjHGmEp2+8gYY0wlu1IwxhhTKel6H/Xo0UP7\n9+8fdhi+7Nixg5ycnLDDCEQq1w1Su35Wt+TVnPotWrSoWFV7NrRd0iWF/v37s3DhwrDD8KWwsJAJ\nEyaEHUYgUrlukNr1s7olr+bUT0Q+a3gru31kjDEmhiUFY4wxlSwpGGOMqZR0bQrGmODt3buXoqIi\ndu3aFXYojZKbm8uKFSvCDiMwfuqXnZ1N3759ycrKatIxLCkYY2opKiqiU6dO9O/fnzgvm2u1tm/f\nTqdOnRreMEk1VD9VpaSkhKKiIgYMGNCkY9jtI2NMLbt27aJ79+5JlRAMiAjdu3dv1hWeJQVjTFyW\nEJJTc//e0icp/Pvf8Mtfgg3rYYwxdUqfpLB4Mdx6K3z1VcPbGmNCNXHiRObNm1etbPr06VxyySX1\nfm///fcHYMOGDUyePDnuNhMmTGjwAdjp06dTVlZWuXz88cezZcsWP6HX66abbuKOO+5o9n6ClD5J\nIRJx8yVL6t/OGBO6KVOmMHPmzGplM2fOZMqUKb6+n5eXx6xZs5p8/JpJYe7cuXTp0qXJ+0sm6ZMU\nCgrc3JKCMa3e5MmTeeGFF9izZw8A69atY8OGDRx55JGUlpYyadIkRo8eTUFBAf/85z9rfX/dunUM\nHz4cgJ07d3LWWWcxdOhQTj31VHbu3Fm53SWXXMLYsWMZNmwYN954IwB/+tOf2LBhAxMnTmTixImA\nG16nuLgYgDvvvJPhw4czfPhwpk+fXnm8oUOHcvHFFzNs2DC++93vVjtOQ+Ltc8eOHZxwwgmMGDGC\n4cOH8/TTTwNw3XXXkZ+fTyQS4eqrr27Un6sf6dMltVs36NPHkoIxjXT55fDBB4nd58iR4J374urW\nrRvjxo3jxRdf5JRTTmHmzJmcccYZiAjZ2dnMmTOHzp07U1xczPjx4zn55JPrbGC977776NChAytW\nrGDJkiWMHj26ct3vfvc7unXrxr59+5g0aRJLlizhsssu484772T+/Pn06NGj2r4WLVrEo48+yrvv\nvouqcuihh3L00UfTtWtXVq9ezVNPPcWDDz7IGWecwezZsznnnHMa/LOoa5+ffvopeXl5vPDCCwBs\n3bqVkpIS5syZw8qVKxGRhNzSqil9rhTA3UJaujTsKIwxPsTeQoq9daSqXH/99UQiEY499ljWr1/P\nxo0b69zPG2+8UXlyjkQiRKK3koFnnnmG0aNHM2rUKJYtW8by5cvrjemtt97i1FNPJScnh44dO/LD\nH/6QN998E4ABAwYwcuRIAMaMGcO6det81bOufRYUFPDyyy9z7bXX8uabb5Kbm0tubi7Z2dlceOGF\nPPvss3To0MHXMRojfa4UwCWFV16BvXuhiU/7GZNu6vtFH6RTTjmFK664gsWLF1NWVsaYMWMAmDFj\nBps3b2bRokVkZWXRv3//JvXLX7t2LXfccQcLFiyga9eunHfeec3q39+uXbvKz5mZmY26fRTPwQcf\nzOLFi5k7dy433HADkyZN4oorruC9997j1VdfZdasWdxzzz289tprzTpOTel3pbB3L6xaFXYkxpgG\ndOzYkYkTJ3LBBRdUa2DeunUrvXr1Iisri/nz5/PZZ/WPCH3UUUfx5JNPAvDRRx+xxLuFvG3bNnJy\ncsjNzWXjxo28+OKLld/p1KkT27dvr7WvI488kn/84x+UlZWxY8cO5syZw5FHHtmseta1zw0bNtCh\nQwfOOeccrrnmGhYvXkxpaSlbt27l+OOP56677uLDDz9s1rHjSb8rBXDtCl4jlDGm9ZoyZQqnnnpq\ntZ5IU6dO5aSTTqKgoICxY8cyZMiQevdxySWXcP755zN06FCGDh1aecUxYsQIRo0axZAhQ+jXrx+H\nH3545XemTZvGcccdR15eHvPnz68sHz16NOeddx7jxo0D4KKLLmLUqFG+bxUB/Pa3v61sTAY3pEi8\nfc6bN49rrrmGjIwMsrKyuO+++ygtLWXq1Kns2rULVeXOO+/0fVzfVDWppjFjxmiT7dmjmpWleu21\nTd9HI8yfP79FjhOGVK6bamrXz0/dli9fHnwgAdi2bVvYIQTKb/3i/f0BC9XHOTa9bh9lZcHQodYD\nyRhj6pBeSQGsB5IxxtQjPZNCURF8/XXYkRhjTKuTnkkB7GrBGGPiSN+kYO0KxhhTS/olhf32gx49\nLCkYY0wc6ZcURNzgeJYUjGm1SkpKGDlyJCNHjmS//fajT58+lcvRQfIacv7557OqEQ+qPvTQQ1x+\n+eVNDTllpNfDa1GRCDz4IFRUQEb65UVjWrvu3bvzgTcK30033UTHjh1rjQha2a++jv/Djz76aOBx\npqL0PCNGIlBWBp9+GnYkxphGWLNmDfn5+UydOpVhw4bx5ZdfMm3atMrhr2+99dbKbY844gg++OAD\nysvL6dKlC9dddx0jRozgsMMOY9OmTb6P+cQTT1BQUMDw4cO5/vrrASgvL+fcc8+tLP/Tn/4EwF13\n3VU5rLWfEVJbo/S9UgB3C2ngwHBjMaa1C2Ps7HqsXLmSxx9/nLFjxwJw66230q1bN8rLyznqqKNY\nvnw5+fn51b6zdetWjj76aG699VauvPJKHnnkEa677roGj1VUVMQNN9zAwoULyc3N5dhjj+X555+n\nZ8+eFBcXs9TrxRgdwvr222/ns88+o23btoEMa90S0vNKIT/f3TaydgVjks5BBx1UmRAAnnrqKUaP\nHs3o0aNZtWpV3OGv27dvz/e//32gccNav/vuuxxzzDH06NGDrKwszj77bN544w0GDhzIqlWruOyy\ny5g3bx65ubkADBs2jHPOOYcZM2aQlaQjMafnlUKHDjBokCUFY/wIa+zsOuTk5FR+Xr16NXfffTfv\nvfceXbp04cwzz4w7/HXbtm0rP2dmZlJeXt6sGLp3786SJUt48cUXuffee5k9ezYPPPAA8+bN4/XX\nX+e5557j97//PUuWLCEzM7NZx2pp6XmlANYDyZgUsG3bNjp16kTnzp358ssvefXVVxO6/0MPPZT5\n8+dTUlJCeXk5M2fO5Oijj2bz5s2oKqeffjo333wzixcvZt++fRQVFXHMMcdw++23U1xcXO09z8ki\nPa8UwLUrzJ4NpaXQsWPY0RhjmmD06NHk5+czZMgQDjzwQMaPH9+s/T388MPMmjWrcnnhwoXccsst\nTJgwAVXlpJNO4oQTTmDx4sVceOGFqCoiwm233UZ5eTlnn30227dvp6KigquvvppOnTo1t4otz89Q\nqq1patbQ2bH+8Q9VUH3nncTsL450H345maVy/Wzo7ORlQ2cHyYa7MMaYWtI3KRx4IHTqZEnBGGNi\nNJgUROR0Eenkfb5BRJ4VkdHBhxawjAxrbDamHu6Og0k2zf1783Ol8GtV3S4iRwDHAg8D9zXrqK1F\nNCnYP35jqsnOzqakpMQSQ5JRVUpKSsjOzm7yPvz0PtrnzU8AHlDVF0Tkt00+YmsSicD998P69dC3\nb9jRGNNq9O3bl6KiIjZv3hx2KI2ya9euZp0QWzs/9cvOzqZvM85nfpLCehG5H/gOcJuItMNnW4SI\nHAfcDWQCD6nqrXVsdxowCzhEVRf6ijwRYhubLSkYUykrK4sBAwaEHUajFRYWMmrUqLDDCExL1M/P\nyf0MYB7wPVXdAnQDrmnoSyKSCdwLfB/IB6aISH6c7ToBPwfebUTciVFQ4ObWrmCMMYC/pLA/8IKq\nrhaRCcDpwHs+vjcOWKOqn6rqHmAmcEqc7W4BbgNqP5setNxc1wvJkoIxxgD+bh/NBsaKyEDgAeCf\nwJPA8Q18rw/wRcxyEXBo7AZeL6Z+XjtFnVcfIjINmAbQu3dvCgsLfYTtz/C8PNq//TYLErjPqNLS\n0oTG2pqkct0gtetndUteLVK/hp5uAxZ7818AP/M+v+/je5Nx7QjR5XOBe2KWM4BCoL+3XAiMbWi/\nCXuiOer661UzM1V37UrsftWeik1mqVw/q1vyak79SOATzXtFZArwI+B5r8zPmLDrgX4xy329sqhO\nwHCgUETWAeOB50RkLC0pEoF9+2DlyhY9rDHGtEZ+ksL5wGHA71R1rYgMAP7m43sLgEEiMkBE2gJn\nAc9FV6rqVlXtoar9VbU/8A5wsrZk7yOw4S6MMSZGg0lBVZcDVwNLRWQ4UKSqt/n4XjlwKa7n0grg\nGVVdJiI3i8jJzYw7cQYNgnbtLCkYYww+Gpq9HkePAesAAfqJyI9V9Y2Gvquqc4G5Ncp+U8e2ExoO\nNwBt2sCwYZYUjDEGf72P/hv4rqquAhCRg4GngDFBBtaiCgpg3rywozDGmND5aVPIiiYEAFX9GH8N\nzckjEoGvvoIke6TfGGMSzU9SWCgiD4nIBG96EGjZxuCgRRubly4NNw5jjAmZn6RwCbAcuMyblgM/\nDTKoFmc9kIwxBvDRpqCqu4E7vQkAEXkaODPAuFpWr17Qu7clBWNM2mvqm9cOS2gUrUEkYknBGJP2\n0vd1nDUVFMCyZVBeHnYkxhgTmjpvH9Xzyk0h1XofgbtS2LUL1qyBIUPCjsYYY0JRX5vCf9ezLvUG\nCortgWRJwRiTpupMCqo6sSUDCd3QoZCZ6doVTj897GiMMSYU1qYQlZ0NgwdbY7MxJq1ZUohlPZCM\nMWnOkkKsggJYtw62bg07EmOMCUWDSUFEnhWRE0Qk9RNItLH5o4/CjcMYY0Li50T/P8DZwGoRuVVE\nBgccU3hsDCRjTJrz85KdV1R1KjAa906FV0Tk/0TkfBFJrecV+vWD3FxrVzDGpC1ft4REpDtwHnAR\n8D5wNy5JvBxYZGEQscZmY0xa89OmMAd4E+gAnKSqJ6vq06r6M6Bj0AG2uGhSUA07EmOMaXF+3rz2\nJ1WdH2+Fqo5NcDzhKyiA7dvhs8+gf/+wozHGmBbl5/bR2yJypdcLabaIXCEi2YFHFhZrbDbGpDE/\nSeFxYBjwZ+AeIB/4W5BBhWr4cDe3dgVjTBryc/touKrmxyzPF5HlQQUUuk6d4FvfsqRgjElLfq4U\nFovI+OiCiBxKqr2juSbrgWSMSVN+ksIY4P9EZJ2IrAPeBg4RkaUikppnzkgEPv4Ydu4MOxJjjGlR\nfm4fHRd4FK1NQQFUVMDy5TBmTNjRGGNMi/HzRPNnQBfgJG/qoqqfRaegAwyF9UAyxqQpPw+v/RyY\nAfTypidE5GdBBxaqgw6C9u2tXcEYk3b83D66EDhUVXcAiMhtuHaFPwcZWKgyM13XVEsKxpg046eh\nWYB9Mcv7vLLUFonAhx/acBfGmLTiJyk8CrwrIjeJyE3AO8DDgUbVGhQUQHExbNwYdiTGGNNiGrx9\npKp3ikghcIRXdL6qvh9oVK1BtLF5yRLYb79wYzHGmBZSb1IQkUxgmaoOARa3TEitREGBmy9dCt/9\nbrixGGNMC6n39pGq7gNWicgBLRRP69GjB+TlWWOzMSat+Ol91BVYJiLvATuihap6cmBRtRY23IUx\nJs34SQq/DjyK1ioSgddeg717ISu13jxqjDHx+Ol9dLyqvh47Acf72bmIHCciq0RkjYhcF2f9T70x\nlD4QkbdEJD/efkJTUAB79rhxkIwxJg34SQrfiVP2/Ya+5DVS3+ttmw9MiXPSf1JVC1R1JHA7cKeP\neFpObA8kY4xJA3UmBRG5RESWAoNFZEnMtBbwMyjQOGCNqn6qqnuAmcApsRuo6raYxRygdT0pNmQI\ntGljYyAZY9KGaB1P7IpILq6R+Q9A7K2f7ar6dYM7FpkMHKeqF3nL5+KGy7i0xnb/CVwJtAWOUdXV\ncfY1DZgG0Lt37zEzZ870UbXEGHvhhezu1Yulf/hDo79bWlpKx44dA4gqfKlcN0jt+lndkldz6jdx\n4sRFqjq2wQ1VtcEJyATygAOik4/vTAYeilk+F7innu3PBh5raL9jxozRFjV1qmq/fk366vz58xMb\nSyuSynVTTe36Wd2SV3PqByxUH+d7P6OkXgpsBF4GXvCm530kpvVAv5jlvl5ZXWYCP/Cx35YVicAX\nX8A334QdiTHGBM5PQ/PlwGBVHaauUbhAVSM+vrcAGCQiA0SkLXAW8FzsBiIyKGbxBKDWraPQxT7Z\nbIwxKc5PUvgC2NrYHatqOXApMA9YATyjqstE5GYRiT74dqmILBORD3DtCj9u7HECZz2QjDFpxM/D\na58ChSLyArA7WqiqDXYfVdW5wNwaZb+J+fxz/6GGJC8PunWzKwVjTFrwkxQ+96a23pReRGy4C2NM\n2vAzdPZ/AYhIB1UtCz6kVigSgYcfhooKyPBzx80YY5KTn95Hh4nIcmCltzxCRP4n8Mhak0gEduyA\ntWvDjsQYYwLl52fvdOB7QAmAqn4IHBVkUK1OtAeS3UIyxqQ4X/dCVPWLGkX74m6YqoYNc20LlhSM\nMSnOT0PzFyLybUBFJAv4Oa6LafrIyYGBA60HkjEm5fm5Uvgp8J9AH9wTySO95fRiPZCMMWnAT++j\nYmBqC8TSukUi8OyzrsE5JyfsaIwxJhB+eh/dLiKdRSRLRF4Vkc0ick5LBNeqRCKgCsuWhR2JMcYE\nxs/to++qe+/BicA6YCBwTZBBtUrWA8kYkwb8JIXoLaYTgL+raqPHQUoJAwa420aWFIwxKcxP76Pn\nRWQlsBO4RER6AruCDasVyshwVwvWA8kYk8IavFJQ1euAbwNjVXUvsIMar9VMG9EeSHW8rc4YY5Kd\nn4bm04G9qrpPRG4AnsC9hS39RCLw9dewYUPYkRhjTCD8tCn8WlW3i8gRwLHAw8B9wYbVSlljszEm\nxflJCtEhLU4AHlDVF0jHIbTBkoIxJuX5SQrrReR+4Exgroi08/m91NO1K/TrZ0nBGJOy/Jzcz8C9\nUvN7qroF6EY6PqcQFYlYDyRjTMry0/uoDPgE+J6IXAr0UtWXAo+stYpEYMUK2LMn7EiMMSbh/PQ+\n+jkwA+jlTU+IyM+CDqzVikSgvBxWrgw7EmOMSTg/D69dCByqqjsAROQ24G3gz0EG1mrFNjZHIuHG\nYowxCeanTUGo/lKdfV5Zejr4YGjb1hqbjTEpyc+VwqPAuyIyx1v+Ae5ZhfSUlQX5+ZYUjDEpyc/7\nFO4UkULgCK/ofFV9P9CoWrtIBF55JewojDEm4epNCiKSCSxT1SHA4pYJKQlEIvD441BcDD16hB2N\nMcYkTL1tCqq6D1glIge0UDzJIdrAbM8rGGNSjJ82ha7AMhF5DzdCKgCqenJgUbV2sT2QJk4MNxZj\njEkgP0nh14FHkWx694aePa2x2RiTcupMCiIyEOitqq/XKD8C+DLowFo1kap3KxhjTAqpr01hOrAt\nTvlWb116i0Rg2TLYt6/hbY0xJknUlxR6q2qtllSvrH9gESWLSAR27oRPPgk7EmOMSZj6kkKXeta1\nT3QgSSfaA8luIRljUkh9SWGhiFxcs1BELgIWBRdSkhg6FDIyLCkYY1JKfb2PLgfmiMhUqpLAWNxb\n104NOrBWr317Nw6SJQVjTAqpMymo6kbg2yIyERjuFb+gqq+1SGTJIBKBBQvCjsIYYxLGz0t25qvq\nn72pUQlBRI4TkVUiskZErouz/koRWS4iS0TkVRE5sDH7D10kAmvXwvbtYUdijDEJEdi7lr1xk+4F\nvg/kA1NEJL/GZu8DY1U1AswCbg8qnkBEG5s/+ijcOIwxJkECSwrAOGCNqn6qqnuAmcApsRt4VyFl\n3uI7QN8A40k864FkjEkxfoa5aKo+wBcxy0XAofVsfyHwYrwVIjINmAbQu3dvCgsLExRiM6lyRE4O\nG198kdWDB9daXVpa2npiTbBUrhukdv2sbsmrJepX3zAX2wGNtwpQVe2cqCBE5Bxcz6aj461X1QeA\nBwDGjh2rEyZMSNShm2/kSPqUlNAnTkyFhYW0qlgTKJXrBqldP6tb8mqJ+tXX+6hTM/e9HugXs9zX\nK6tGRI4FfgUcraq7m3nMlheJwJNPgqobE8kYY5KY7zYFEeklIgdEJx9fWQAMEpEBItIWOAt4rsY+\nRwH3Ayer6qbGBN5qRCKwdSt88UXD2xpjTCvXYFIQkZNFZDWwFngdWEcd9/5jqWo5cCkwD1gBPKOq\ny0TkZhGJvovhj0BH4O8i8oGIPFfH7lova2w2xqQQPw3NtwDjgVdUdZT3MNs5fnauqnOBuTXKfhPz\n+dhGxNo6Dfee61uyBE48MdxYjDGmmfzcPtqrqiVAhohkqOp8XKOwAejcGfr3tysFY0xK8HOlsEVE\nOgJvADNEZBMxr+U02At3jDEpw8+VwilAGXAF8L/AJ8BJQQaVdCIR+Phj2LUr7EiMMaZZ/CSFXkBb\nVS1X1ceAB4HmdldNLZGIewPbihVhR2KMMc3iJyn8HaiIWd7nlZko64FkjEkRfpJCG2/sIgC8z22D\nCykJDRwI2dmWFIwxSc9PUtgc81wBInIKUBxcSEkoMxOGDbOkYIxJen56H/0U1+voHty4R18APwo0\nqmQUicALL4QdhTHGNIufl+x8oqrjce9EGKqq31bVNcGHlmQiEdi0CTZuDDsSY4xpsvpGST1HVZ8Q\nkStrlAOgqncGHFtyiTY2L10KvXuHG4sxxjRRfVcKOd68Ux2TiVVQ4ObWrmCMSWL1DZ19v/dKzW2q\nelcLxpScevaE/fazpGCMSWr1timo6j5gSgvFkvxsuAtjTJLz0yX13yJyj4gcKSKjo1PgkSWjSASW\nL4fy8rAjMcaYJvHTJXWkN785pkyBYxIfTpKLRGD3bli9GoYODTsaY4xptAaTgqpObIlAUkLscBeW\nFIwxScjPm9dyReROEVnoTf8tIrktEVzSGTIE2rSxdgVjTNLy06bwCLAdOMObtgGPBhlU0mrXDgYP\ntqRgjElaftoUDlLV02KW/0tEPggqoKQXicC//x12FMYY0yR+rhR2isgR0QURORzYGVxISS4Sgc8/\nhy1bwo7EGGMazc+VwiXAY147ggBfA+cFGVRSizY2f/RRuHEYY0wT+Ol99AEwQkQ6e8vbAo8qmcX2\nQMrPDzcWY4xppAaTQh0D4m0FFnkJw8Tq0we6drWkYIxJSn7aFMbi3qnQx5t+AhwHPCgivwgwtuQk\n4gbHsx5Ixpgk5Ccp9AVGq+pVqnoVMAboBRyFtS3EF4m4IbQrKhre1hhjWhE/SaEXsDtmeS/QW1V3\n1ig3UZEIlJaS/dVXYUdijDGN4qf30QzgXRH5p7d8EvCkiOQAywOLLJl5jc05n34aciDGGNM4fnof\n3SIiLwKHe0U/VdWF3uepgUWWzIYNAxE6WlIwxiQZP7ePALJxL9u5G/hMRAYEGFPy69gRDjqI3i+9\nBHPngmrYERljjC9+BsS7EbgW+KVXlAU8EWRQKWH6dDLKy+GEE2DsWJgzxxqejTGtnp8rhVOBk4Ed\nAKq6AXtHc8NOOIF3n3gCHnkEtm2DH/4QRoyAmTNh376wozPGmLj8JIU9qqq4F+vgNTAbH7RNGzj/\nfFixAmbMcFcKU6a4h9oeewz27g07RGOMqcZPUnhGRO4HuojIxcArwEPBhpVi2rSBs892zy78/e/Q\nvj2cd54bZvuBB9zb2owxphX5Eb3wAAATDklEQVRoMCmo6h3ALGA2MBj4jar+KejAUlJGBkyeDO+/\nD//6F/TsCT/5CQwcCH/+M+y0wWeNMeHy09B8m6q+rKrXqOrVqvqyiNzWEsGlLBE48UR45x146SUY\nMAAuu8zN77gDSkvDjtAYk6b83D76Tpyy7/vZuYgcJyKrRGSNiFwXZ/1RIrJYRMpFZLKffaYUEfjO\nd+CNN+D1192YSddcA/37w+9+B1u3hh2hMSbN1JkUROQSEVkKDBaRJTHTWqDB0d5EJBO4F5dA8oEp\nIlJz2NDPceMnPdnUCqSMo46Cl1+Gt9+G8ePhhhvgwAPhN7+BkpKwozPGpIn6rhSexA1p8Zw3j05j\nVPUcH/seB6xR1U9VdQ8wEzgldgNVXaeqSwDrwB81fjw8/zwsXgyTJsEtt7grh2uvhU2bwo7OGJPi\nRH0+bSsivXBPNgOgqp83sP1k4DhVvchbPhc4VFUvjbPtX4HnVXVWHfuaBkwD6N2795iZM2f6ijls\npaWldOzYsVn7yFm7lgOeeIJehYVUZGXx5Ykn8vmZZ7KnZ88ERdk0iahba5bK9bO6Ja/m1G/ixImL\nVHVsgxuqar0T7upgNe7htbW4X/XLfHxvMvBQzPK5wD11bPtXYHJD+1RVxowZo8li/vz5idvZqlWq\n552nmpmp2rat6k9/qrp2beL230gJrVsrlMr1s7olr+bUD1ioPs6xfhqafwuMBz5W1QHAJOAdH99b\nD/SLWe7rlZmmOPhgePRRWL3aPRD3yCMwaBBccIErM8aYBPCTFPaqagmQISIZqjof9za2hiwABonI\nABFpC5yFa58wzTFgAPzlL/DJJ/Af/wFPPQVDhriH455+GtassQH4jDFN5icpbBGRjsAbwAwRuRtv\nHKT6qGo5cCkwD1gBPKOqy0TkZhE5GUBEDhGRIuB04H4RWdbUiqSdvn3h7rth3Tq46ir3MNxZZ7mr\nh27dXCP1L37hEsUnn1iiMMb44uclO6cAO4ErcO9PyAVu9rNzVZ0LzK1R9puYzwtwt5VMU/XuDbff\nDr/9LXz0ESxaBAsXuvndd8OePW67Ll1g9GgYM8ZNY8fCt77lnpUwxhhPnUlBRAbiXrv5b6+oAnhM\nRI4AugDWeb41advWnfRHj4aLL3Zle/ZYojDGNEp9VwrTqXqHQqyt3rqTAonIJI7fRDF9etWIrZYo\njElr9SWF3qq6tGahqi4Vkf6BRWSCVV+iiCYJP4nC2iiMSUn1JYUu9axrn+hATIhiE0VUA4niiJwc\nGDrUDf998MHVpxR+eMiYVFdfUlgoIher6oOxhSJyEbAo2LBM6BpIFBvnzqVPWRm8+aZ7gVCsvLyq\nBBGbNAYMgKyslq2HMaZR6ksKlwNzRGQqVUlgLNAW94pOk25iEsXqgw+mz4QJrnznTvd8xMcfw6pV\nbv7xxzB7dvXB/Nq0ce0TNa8sBg+G/fe3dgtjWoE6k4KqbgS+LSITgeFe8Quq+lqLRJZgO3a4c1K7\ndmFHkoLat3fDfhcU1F5XUlKVJGKTxiuvwK5dVdvl5MS/ujj4YMjNbbm6GJPmGnxOwXuCeX4LxBKo\nhx+Gq6+G4cNdW2m03bSgwJ3TTEC6d4fDDnNTrIoKKCqqfXWxYIF7ZWlFzMC5vXu7K4wDD4QDDnDz\n2M+dO7dsnYxJYX4eXksJ48e7B38XL4Y5c+Ah7y3TmZkwbFhVkhg9GkaMcD9cTYAyMtxJ/YAD4Nhj\nq6/bvds9hR17dbF2rUsYs2dX9YqKys2tniRqzvfbzx3PGNOgtEkK48a5CVxvyi++cB1qFi9287lz\n4a9/deszMtxwQrFXFCNHQqdOoYWfXtq1g/x8N9VUUQEbN8Jnn7np88+rz996C7Zsqf6drCzo16/u\nxHHAAZCdXftYxqShtEkKsUSqzgWnek3mqrBhQ/VE8eqr8Le/VX3n4IOrX1GMHm23u1tcRoZrlN5/\nf3f5F8+2bbWTRXT+6qvuL7qixnudevWqliQOKC11o8927157ats2+HoaE5K0TArxiECfPm46+eSq\n8q++qkoSixe7H6JPPVW1fuDA2omiW7eWj9/E6NzZNR4NHx5//d69sH59/KSxbBm8+CLfKiuDBx+M\n//2OHasSRLdu8RNHzSk313pXmaRgSaEB++0Hxx/vpqjNm12CiCaLBQvgmWeq1vfv75JDXl4fBg50\nA5qaViQry/0l9e9f5yZv/O//ctSwYa73VOz09de1yz77zM2/+abuJ70zM+tPID16QM+e1eddulhb\niGlxlhSaoGdP+N733BT19dfw/vtVVxQLF8Kzzw7inntcx5vJk+G009wdCtP6VWRnu3aIfv0a3jhq\n3z7XnlEzacSb1q1z/1hKSqp3zY2VmVl3wojOa5ZZ24hpJksKCRJ9hcGkSVVljz/+LkVFhzJrluv5\ndNVVcMghVQnioIPCi9cEIHoS7969cd8rK4PiYncJWlxc/XNs2fLlbl5SUrtNJConp8HE0eWTT1ys\nbdtWn7Kyape1bWtXK2nGkkKADjhgJz/6EVx/vethOXs2zJoF117rplGjXIKYPNk1Yps01aFDVc8H\nP6JXJHUlj9iyFSvcfEfVe7FGNja+zMy6E0Z9ySQ6tW/v6piT4+Z1fY5X1r69tcW0MEsKLeSgg9yL\n0H7xC3cLOpogfvUrNxUUVCWIeD0xjakUe0UyeLC/7+zc6a4wNm/m/TffZNSwYW4sqz17XMN79HNd\nU0PbRNfv2uV6f0XLd+92xy4rc4mpvLzx9a0vkdT43L+4GN57z/Uf79zZzeNNOTl2BVQHSwohOPBA\nuPJKNxUVwbPPugRx001w441u8NFogigosB9KJgHat3c9Hvr2ZevWrRAdt6ql7d1blSDKyqp/jldW\n3/qSEvfAUcz6A8vK/A3rLuJ6kdVMFvUlknjrcnLcn2379i5ZpwBLCiHr2xcuu8xNX37pnraeNQt+\n9zu45Rb3yuVoghg1yhKESXJZWa57bkAP+Lz+2mtMGDcOtm+vPW3bFr88dv3atdXLdu/2f/A2bVxy\nyM6uPm9OWY11bbZvD+TPrVo1Aj+C8W3//eE//sNNmzbBP/7hEsTtt8Mf/uBGno4miEMOsQRhTC0Z\nGe4KoGNH9x+qufburT+plJW522M7d7pbZ7HzmmXFxfHX7dzp+6VVva64Ak4K9qWXlhRaqV69YNo0\nN5WUwD//6RLE9Onwxz+6NsnTTnMJYvz4cG+PVlS4ts/oPHaqWdbUbbp3d0OPdOgQXj1NGsrKcl0L\ng3wiVdUln3iJpUYS+bqu7ssJZEkhCXTvDhdc4KZvvoF//csliHvvhbvucu+0+cEPXI/DvXtdW155\nedXnxs79bXtU5cm7pYi4583y8127S3R4pKFDbaBUk8REqnpqNXBbbVdhYeDhWFJIMl27wo9+5KZt\n2+D5512CeOQR96OiTRs3ZWVVn8cri523a+euuBvaLjpfv76IAQMOIDOTalNGBg2W+dkmtiwjw91O\nW77cTStWuNcxxN7u7dMnfrLo0SO8vytjkpElhSTWuTOcfbabKircD46WamcoLPyUCRN89qtPkNNO\nq/q8b59rE4wmiWjCePjhal3y6dkzfrKwF70ZE58lhRSRbl2uMzPdYIQDB1YfwDD67p7Yq4rly2Hm\nzOojaufm1k4W+fluVItk/rMsLXX1LypyY/7F+1xe/m0GDKjsoVo59elTNbf3iaQvSwompcS+u+e4\n46rKVd2It7FXFcuXu9tvjzxStV2HDq5BOyurgIMOcmPSde3qpujnmmWdOwefSFTd+Fo1T/A1l7dt\nq/3d7t2rTvZjx8KGDcVUVOSxdq0b9ffrr2t/p2vX6okiXvJIl4Ffd+92f65bt1afb9vmfoTk5NQ/\ntW2bXH9OlhRMWhCpeg3DMcdUX1dc7JJFNGGsXAlr17blnXfc1cWWLXUPNRTdd25u7eRRXyKJnWdm\nuvcG1ffrfv362uPmRevUt697sHnSpNon8ry82q+bLSz8mAkT8iqXy8rqPm5RkRvocePG2vXOyamd\nKGomjx49wrvyKi+vfRKPd2JvqGzPnubFkZlZ9eB1Q1ND223ZkpWYP5x6WFIwaa9HDzjySDdFFRYu\nYoL31G9Fhbst8803btqypfbnmmXLl1d9bqgXoUjtbupZWVUn1kMOcS+Dij3x9unjhnXPSsA5okMH\n95DkoEF1b7Nnj3u4MjZZxCaQ+fPdu4tq9kaLDosUbe+KThkZtcsSsb60dFxlMigra7jumZkuoXfu\nXDXff393tRhbVtfnjAzXhlXfFH0AO960aVPtberr0Xf55T35wQ/8/b02lSUFYxqQkVF1MmjK0Oe7\ndsVPJNHPu3e7X/Sxv7jD/IUdT9u2ru711X/fPndFEXulsWGDSyiqtaeKivjlzVlfXFzKwIEdqp24\n453Mo/PWNt6eqvvzqiu5bNlSEngMlhSMCVh2tvtVv99+YUcSrMxMl9zy8tzVTRgKC5czYUKvcA6e\nACKue3i7dvGflyssbMSwG03Uin6LGGOMCZslBWOMMZUsKRhjjKlkScEYY0wlSwrGGGMqBZoUROQ4\nEVklImtE5Lo469uJyNPe+ndFpH+Q8RhjjKlfYElBRDKBe4HvA/nAFBGp+fbhC4FvVHUgcBdwW1Dx\nGGOMaViQVwrjgDWq+qmq7gFmAqfU2OYU4DHv8yxgkkhrepTEGGPSS5APr/UBvohZLgIOrWsbVS0X\nka1Ad6A4diMRmQZM8xZLRWRVIBEnXg9q1CWFpHLdILXrZ3VLXs2pn6/n8ZPiiWZVfQB4IOw4GktE\nFqrq2LDjCEIq1w1Su35Wt+TVEvUL8vbReqBfzHJfryzuNiLSBsgFgh/cwxhjTFxBJoUFwCARGSAi\nbYGzgOdqbPMc8GPv82TgNdWa40UaY4xpKYHdPvLaCC4F5gGZwCOqukxEbgYWqupzwMPA30RkDfA1\nLnGkkqS75dUIqVw3SO36Wd2SV+D1E/thbowxJsqeaDbGGFPJkoIxxphKlhQCICL9RGS+iCwXkWUi\n8vOwY0o0EckUkfdF5PmwY0kkEekiIrNEZKWIrBCRw8KOKVFE5Arv3+NHIvKUiGSHHVNziMgjIrJJ\nRD6KKesmIi+LyGpv3jXMGJuqjrr90ft3uURE5ohIlyCObUkhGOXAVaqaD4wH/jPOEB/J7ufAirCD\nCMDdwP+q6hBgBClSRxHpA1wGjFXV4bjOH8neseOvwHE1yq4DXlXVQcCr3nIy+iu16/YyMFxVI8DH\nwC+DOLAlhQCo6pequtj7vB13YukTblSJIyJ9gROAh8KOJZFEJBc4CtcrDlXdo6pbwo0qodoA7b1n\ngjoAG0KOp1lU9Q1cr8VYsUPnPAYE/Jr7YMSrm6q+pKrl3uI7uGe/Es6SQsC8kV9HAe+GG0lCTQd+\nAVSEHUiCDQA2A496t8YeEpGcsINKBFVdD9wBfA58CWxV1ZfCjSoQvVX1S+/zV0DvMIMJ0AXAi0Hs\n2JJCgESkIzAbuFxVt4UdTyKIyInAJlVdFHYsAWgDjAbuU9VRwA6S9/ZDNd699VNwiS8PyBGRc8KN\nKljeg7Ap1+deRH6Fu0U9I4j9W1IIiIhk4RLCDFV9Nux4Euhw4GQRWYcb+fYYEXki3JASpggoUtXo\nVd0sXJJIBccCa1V1s6ruBZ4Fvh1yTEHYKCL7A3jzTSHHk1Aich5wIjA1qNEfLCkEwBv++2Fghare\nGXY8iaSqv1TVvqraH9dQ+ZqqpsQvTlX9CvhCRAZ7RZOA5SGGlEifA+NFpIP373MSKdKIXkPs0Dk/\nBv4ZYiwJJSLH4W7bnqyqZUEdx5JCMA4HzsX9iv7Am44POyjjy8+AGSKyBBgJ/D7keBLCu/qZBSwG\nluL+7yf1kBAi8hTwNjBYRIpE5ELgVuA7IrIad3V0a5gxNlUddbsH6AS87J1T/hLIsW2YC2OMMVF2\npWCMMaaSJQVjjDGVLCkYY4ypZEnBGGNMJUsKxhhjKllSMKYFiciEVBtZ1qQWSwrGGGMqWVIwJg4R\nOUdE3vMeErrfe39EqYjc5b2T4FUR6eltO1JE3okZ576rVz5QRF4RkQ9FZLGIHOTtvmPMOxtmeE8Y\nG9MqWFIwpgYRGQqcCRyuqiOBfcBUIAdYqKrDgNeBG72vPA5c641zvzSmfAZwr6qOwI0zFB29cxRw\nOZAPfAv3BLwxrUKbsAMwphWaBIwBFng/4tvjBlarAJ72tnkCeNZ7B0MXVX3dK38M+LuIdAL6qOoc\nAFXdBeDt7z1VLfKWPwD6A28FXy1jGmZJwZjaBHhMVau92UpEfl1ju6aOEbM75vM+7P+haUXs9pEx\ntb0KTBaRXlD53t8Dcf9fJnvbnA28papbgW9E5Eiv/Fzgde+Ne0Ui8gNvH+1EpEOL1sKYJrBfKMbU\noKrLReQG4CURyQD2Av+Je+nOOG/dJly7A7ghmv/infQ/Bc73ys8F7heRm719nN6C1TCmSWyUVGN8\nEpFSVe0YdhzGBMluHxljjKlkVwrGGGMq2ZWCMcaYSpYUjDHGVLKkYIwxppIlBWOMMZUsKRhjjKn0\n/4bB17k7TOCHAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Mu6apn0jqb77"
      },
      "source": [
        "### WITHOUT DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "lEoYO0Wqqb77",
        "outputId": "959d842c-1d00-4220-dc92-801aa0aaeca7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='relu',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='relu'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('relu'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_21\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_79 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_79 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_80 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_19 (Batc (None, 22, 22, 64)        88        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_80 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_81 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_81 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_82 (Conv2D)           (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_82 (MaxPooling (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_83 (Conv2D)           (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "batch_normalization_20 (Batc (None, 6, 6, 16)          24        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_83 (MaxPooling (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_84 (Conv2D)           (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "batch_normalization_21 (Batc (None, 5, 5, 16)          20        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_84 (MaxPooling (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_21 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_41 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_21 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_42 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,158\n",
            "Trainable params: 122,092\n",
            "Non-trainable params: 66\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Om2fIqRgqb78",
        "outputId": "635e0f98-8d7e-4fb7-eefe-037be2e995a3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 11s 190us/step - loss: 0.1640 - acc: 0.9496 - val_loss: 0.0522 - val_acc: 0.9842\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0480 - acc: 0.9852 - val_loss: 0.0370 - val_acc: 0.9874\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0364 - acc: 0.9884 - val_loss: 0.0359 - val_acc: 0.9885\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0263 - acc: 0.9915 - val_loss: 0.0320 - val_acc: 0.9901\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 7s 121us/step - loss: 0.0229 - acc: 0.9923 - val_loss: 0.0337 - val_acc: 0.9897\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 7s 123us/step - loss: 0.0186 - acc: 0.9938 - val_loss: 0.0362 - val_acc: 0.9897\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0176 - acc: 0.9941 - val_loss: 0.0360 - val_acc: 0.9892\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0162 - acc: 0.9948 - val_loss: 0.0356 - val_acc: 0.9894\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0143 - acc: 0.9954 - val_loss: 0.0537 - val_acc: 0.9850\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0124 - acc: 0.9958 - val_loss: 0.0410 - val_acc: 0.9887\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 7s 123us/step - loss: 0.0085 - acc: 0.9973 - val_loss: 0.0272 - val_acc: 0.9923\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 7s 122us/step - loss: 0.0099 - acc: 0.9968 - val_loss: 0.0380 - val_acc: 0.9888\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "c-jnZdXPqb7-",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "EWgnnTSaqb8A",
        "outputId": "2039f5f8-40bb-4004-8b7e-3baa425035e1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.038047944817450934\n",
            "Test accuracy: 0.9888\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl4VNX5wPHvGwgJEAiyGNlBAWUP\nkILWDVQsblBbXBBtVYTqr2ipS4tLxaJt0VrFvYJoXVC0KEoVRUSiWHFBVFYRRIQAFQhbAgQIeX9/\nnDswDEnmksxklryf57lPZu76HkjmnXvOueeIqmKMMcaUJyXWARhjjIl/liyMMcaEZcnCGGNMWJYs\njDHGhGXJwhhjTFiWLIwxxoRlycKYChARFZF2sY7DmKpiycIkPBFZLSK7RaQwaHk01nFFkoi08RJU\nzVjHYqon+8UzyeICVX0v1kEYk6zszsIkNRG5UkT+KyKPish2EflGRM4M2t5MRKaLyBYRWSkiw4O2\n1RCR20TkOxEpEJEvRKRl0OnPEpEVIrJNRB4TESnl+s28u56GQet6iMhmEUkVkXYi8oEX22YRebkC\nZUwTkfEist5bxotImretsYi86cW4RUTmikiKt+2PIrLOK9vy4H8XY0JZsjDVQR/gO6AxMAZ4LejD\newqQBzQDBgN/FZEzvG03AkOAc4H6wNXArqDzng/8BOgGXAz8LPTCqroemAf8Mmj1ZcBUVd0H3A28\nCxwFtAAeqUD5bgdOBLKB7kBv4A5v201e+ZoAWcBtgIrI8cBI4CeqWs+LfXUFrm2qCUsWJlm87n17\nDizDg7ZtBMar6j5VfRlYDpzn3SWcDPxRVYtU9SvgKeBX3nHXAHeo6nJ1vlbV/KDzjlPVbaq6BpiD\n+7AuzYu4pIN393Gptw5gH9AaaObF8FEFyj4UGKuqG1V1E/Bn4Iqg8zcFWnvln6tuQLj9QBrQSURS\nVXW1qn5XgWubasKShUkWP1fVBkHLxKBt6/TQETN/wN1JNAO2qGpByLbm3uuWuDuSsvwv6PUuIKOM\n/V4FThKRpsBpQAkw19v2B0CAz0RkiYhcXc71ytLMizsgUD6AvwMrgXdFZJWIjAZQ1ZXAKOAuYKOI\nTBGRZhhTBksWpjpoHtKe0ApY7y0NRaReyLZ13uu1wHGVvbiqbsVVNV2Cq4KaEkheqvo/VR2uqs2A\n3wCPV6BL7nrc3UlAoHyoaoGq3qSqxwIDgRsDbROq+qKqnuIdq8C9FS6kSXqWLEx1cDRwg9egfBHQ\nEZihqmuBj4G/iUi6iHQDhgEveMc9BdwtIu3F6SYijSoYw4u46q3BHKyCQkQuEpEW3tutuA/tknLO\nk+bFGlhSgJeAO0SkiYg0Bu4MlEFEzvca0QXYjqt+KhGR40XkDK8hvAjYHea6ppqzrrMmWfxHRPYH\nvZ+lqhd6rz8F2gObgR+BwUFtD0OAf+K+iW8FxgR1wX0AV6//Lq5x/BsgcM4jNR2XfNao6tdB638C\njBeRTC+236nqqnLOUxjyvj9wD64BfqG37t/eOnDlfhTXwL0VeFxV53iJcRwuce7DJc0RFSybqQbE\nJj8yyUxErgSu8apbjDEVZNVQxhhjwrJkYYwxJiyrhjLGGBOW3VkYY4wJK2l6QzVu3FjbtGkT6zB8\n2blzJ3Xr1o11GFGTzOWzsiWuZC5fZcr2xRdfbFbVJuH2S5pk0aZNG+bPnx/rMHzJzc2lb9++sQ4j\napK5fFa2xJXM5atM2UTkh/B7WTWUMcYYHyxZGGOMCcuShTHGmLCSps3CGFM19u3bR15eHkVFRbEO\n5YhkZmaybNmyWIcRFX7Klp6eTosWLUhNTa3QNSxZGGOOSF5eHvXq1aNNmzaUMjlg3CooKKBevXrh\nd0xA4cqmquTn55OXl0fbtm0rdA2rhjLGHJGioiIaNWqUUImiuhMRGjVqVKm7QUsWxpgjZoki8VT2\n/8ySxdatMHYsfPFFrCMxxpi4ZcmiRg0YMwZmzox1JMYYH/r168fMkL/X8ePHc91115V7XNOmTQFY\nv349gwcPLnWfvn37hn24d/z48ezatevA+3PPPZdt27b5Cb1cd911F/fff3+lzxMtlizq14fjjoOv\nvop1JMYYH4YMGcKUKVMOWTdlyhSGDBni6/hmzZoxderUCl8/NFnMmDGDBg0aVPh8iSKqyUJEBojI\nchFZGZgoPmT7aSKyQESKRWRwyLZWIvKuiCwTkaUi0iZqgWZnW7IwJkEMHjyYt956i7179wKwevVq\n1q9fz6mnnkphYSFnnnkmPXv2pGvXrrzxxhuHHb969Wq6dOkCwO7du7n00kvp2LEjF154Ibt37z6w\n33XXXUdOTg6dO3dmzJgxADz88MOsX7+efv360a9fP8ANNbR582YAHnjgAbp06UKXLl0YP378get1\n7NiR4cOH07lzZ84+++xDrhNOaefcuXMn5513Ht27d6dLly68+uqrAIwePZpOnTrRrVs3br755iP6\ndw0nal1nRaQG8Bhu2sc84HMRma6qS4N2WwNcCZRWqueAv6jqLBHJIJrzA2dnw2uvQUEBJGnXOmOi\nYdSoyH/Pys4G7zOxVA0bNqR37968/fbbDBo0iClTpnDxxRcjIqSnpzNt2jTq16/P5s2bOfHEExk4\ncGCZjbtPPPEEderUYdmyZSxcuJCePXse2PaXv/yFhg0bsn//fs4880wWLlzIDTfcwAMPPMCcOXNo\n3LjxIef64osveOaZZ/j0009RVfr06cPpp5/OUUcdxYoVK3jppZeYOHEiF198Ma+++iqXX3552H+L\nss65atUqmjVrxltvvQW47sz5+flMmzaNb775BhGJSNVYsGjeWfQGVqrqKlXdC0wBBgXvoKqrVXUh\nIYlARDoBNVV1lrdfoaruIlqys0EVFi2K2iWMMZETXBUVXAWlqtx2221069aNs846i3Xr1vHjjz+W\neZ4PP/zwwId2t27d6Nat24Ftr7zyCj179qRHjx4sWbKEpUuXlnUaAD766CMuvPBC6tatS0ZGBr/4\nxS+YO3cuAG3btiU7OxuAXr16sXr1al/lLOucXbt2ZdasWfzxj39k7ty5ZGZmkpmZSXp6OsOGDeO1\n116jTp06vq7hVzQfymsOrA16nwf08XlsB2CbiLwGtAXeA0ar6v7gnURkBN4k81lZWeTm5lYo0LTd\nuzkJ+PaVV1jv3dpGU2FhYYVjTQTJXD4rm3tauKCgAIC7745OLN7py3TGGWcwatQo5s6dS2FhIR06\ndKCgoIDJkyezYcMGcnNzSU1NpUuXLmzevPnA8N0FBQUUFhZSUlJCQUEBxcXF7Nq160B5SkpK2Llz\nJ4sWLeK+++4jNzeXo446imuvvZZt27ZRUFCAqlJYWEhaWhrAgfdFRUXs2bPnwLn27NlDUVERhYWF\npKamHlhfXFzMzp07D7wP2LNnzyH7AWWes2nTpnzwwQe8++673HrrrZx22mnceuutzJ49m9zcXKZN\nm8ZDDz3Em2++ecg1ioqKKvz7G69PcNcETgV64KqqXsZVV00K3klVJwATAHJycrTCww+rQqNGdNi1\niw5VMIRxMg+VDMldPisbLFu2LOZPQterV48zzjiD66+/nqFDhx6IZ8+ePTRr1oyGDRsyZ84c1qxZ\nQ0ZGxoHt9erVIyMjg5SUlAPneP311zn//PNZvHgxixcvpm7dupSUlFCvXj1atGjBpk2beO+99+jf\nvz/16tWjfv36qOqBc4oIGRkZ9O/fnyuvvJIxY8agqsyYMYPnn3/+kOsBpKWlsW/fvsP+DdPS0khL\nSztkfVnnLCgoICsri+HDh9O0aVP++c9/IiKUlJQwePBg+vfvz7HHHnvYNdLT0+nRo0eF/s2jmSzW\nAS2D3rfw1vmRB3ylqqsAROR14ERCkkXEiFgjtzEJZsiQIVx44YWH9IwaOnQoF1xwAV27diUnJ4cT\nTjih3HNcd911XHXVVXTs2JGOHTvSq1cvALp3706PHj044YQTaNmyJSeffPKBY0aMGMGAAQNo1qwZ\nc+bMObC+Z8+eXHnllfTu3RuAa665hh49eviucgK45557DjRig2uLKO2cM2fO5JZbbiElJYXU1FTu\nv/9+CgoKGDRoEEVFRagqDzzwgO/r+qKqUVlwiWgVrhqpFvA10LmMff8FDA56X8Pbv4n3/hngt+Vd\nr1evXlopN92kmp6uum9f5c7jw5w5c6J+jVhK5vJZ2VSXLl0a3UCiZMeOHbEOIWr8lq20/ztgvvr4\nTI9aA7eqFgMjgZnAMuAVVV0iImNFZCCAiPxERPKAi4AnRWSJd+x+XA+p2SKyCBBgYrRiBdydRVER\nfPttVC9jjDGJKKptFqo6A5gRsu7OoNef46qnSjt2FtCttG1R4fVU4KuvoFOnKrusMcYkAnuCO+D4\n4yEtzdotjDGmFJYsAlJToUsXSxbGGFMKSxbBAj2iXKO6McYYjyWLYNnZsGkTbNgQ60iMMSauWLII\nFtzIbYyJS/n5+WRnZ5Odnc0xxxxD8+bND7zf63MEhquuuorly5f7vuZTTz3FqFGjKhpyUojXJ7hj\nIzAuzFdfwbnnxjYWY0ypGjVqxFfeF7q77rqLjIyMw0ZYPfBsQErp34efeeaZqMeZbOzOIpjNbWFM\nwlq5ciWdOnVi6NChdO7cmQ0bNjBixIgDw4yPGzfuwL6nnHIKX331FcXFxTRo0IDRo0fTvXt3Tjrp\nJDZu3Oj7mi+88AJdu3alS5cu3HbbbYAb++mKK644sP7hhx8G4MEHHzwwfLifEWfjjd1ZhLJhP4zx\nLxZjlJfjm2++4bnnniMnJweAcePG0bBhQ4qLiznttNNYunQpnUKeo9q+fTunn34648aN48Ybb+Tp\np59m9OjDpt85TF5eHnfccQfz588nMzOTs846izfffJMmTZqwefNmFnmjWAeGCr/vvvv44YcfqFWr\nVsSHD68KdmcRKjsbVq4MP+ylMSbuHHfccQcSBcBLL71Ez5496dmzJ8uXLy91mPHatWtzzjnnAEc2\nfPinn37KGWecQePGjUlNTeWyyy7jww8/pF27dixfvpwbbriBmTNnkpmZCUDnzp25/PLLmTx5Mqmp\nqZUvbBWzO4tQwXNb/PSnsY7GmPhWwTuAaAkMRQ6wYsUKHnroIT777DMaNGjAJZdcQlFR0WHH1KpV\n68DrGjVqUFxcXKkYGjVqxMKFC3n77bd57LHHePXVV5kwYQIzZ87kgw8+YPr06fz1r39l4cKF1KhR\no1LXqkp2ZxHKekQZkxR27NhxYEjxDRs2MHv27Iiev0+fPsyZM4f8/HyKi4uZMmUKp59+Ops2bUJV\nueiiixg7diwLFixg//795OXlccYZZ3DfffexefPmQ+bxTgR2ZxGqeXNo1MiShTEJrmfPnnTq1IkT\nTjiB1q1bc+KJJ1bqfJMmTWLq1KkH3s+fP5+7776bvn37oqpccMEFnHfeeSxYsIBhw4ahqogI9957\nL8XFxVx22WUUFBRQUlLCzTffHPM5QY6UaJI8rZyTk6Pz58+PzMnOOgt27IDPPovM+UIk8wQ6kNzl\ns7K5yY86duwY/YAirKCgIOE+oP3yW7bS/u9E5AtVzSnjkAOsGqo02dmuzaKSdZfGGJMsLFmUxua2\nMMaYQ4RNFiJykYjU817fISKviUjP6IcWQ9bIbUy5kqX6ujqp7P+ZnzuLP6lqgYicApyFmwf7iUpd\nNd7Z3BbGlCk9PZ38/HxLGAlEVcnPzyc9Pb3C5/DTG2q/9/M8YIKqviUi9/g5uYgMAB7Czan9lKqO\nC9l+GjAeNyPepao6NWR7fWAp8LqqjvRzzYiwuS2MKVOLFi3Iy8tj06ZNsQ7liBQVFVXqwzKe+Slb\neno6LVqUOjGpL36SxToReRLoD9wrImn4q76qATzmHZcHfC4i01U1+BHKNcCVuPm2S3M38KGPGCMv\nOxumT3cP6InEJARj4lFqaipt27aNdRhHLDc3lx49esQ6jKioirL5qYa6GJgJ/ExVtwENgVt8HNcb\nWKmqq1R1LzAFGBS8g6quVtWFQEnowSLSC8gC3vVxrcizuS2MMeYAP3cWTYG3VHWPiPTFVRk95+O4\n5sDaoPd5QB8/QYlICvAP4HJcO0lZ+40ARgBkZWWRm5vr5/S+ZKrSA1j43HNsqeTDPKEKCwsjGmu8\nSebyWdkSVzKXryrK5idZvArkiEg7YALwBvAiEM0JH/4PmKGqeVJOFZCqTvBiIicnRyP6sFTPnnDD\nDXQrKYEIP4SVzA92QXKXz8qWuJK5fFVRNj/JokRVi0XkF8AjqvqIiHzp47h1QMug9y28dX6cBJwq\nIv8HZAC1RKRQVcOPGxwpNreFMcYc4CdZ7BORIcCvgAu8dX7G1/0caC8ibXFJ4lLgMj9BqerQwGsR\nuRLIqdJEEWBzWxhjDOCvgfsq3Df9v6jq996H//PhDlLVYmAkrnF8GfCKqi4RkbEiMhBARH4iInnA\nRcCTIrKkogWJCpvbwhhjAB93Fqq6VERuBjqISBdguare6+fkqjoDmBGy7s6g15/jqqfKO8e/gH/5\nuV7E2dwWxhgD+Hteoi+wAvfMxOPAt97DdMnPhv0wxhjAX5vFP4CzVXU5gIh0AF4CekUzsLhgc1sY\nYwzgr80iNZAoAFT1W/w1cCc+EWvkNsYY/CWL+SLylIj09ZaJQIRmGUoANreFMcb4ShbX4Qbzu8Fb\nlgLXRjOouGJzWxhjjK/eUHuAB7wFABF5GbgkinHFj0Aj95dfQqdOsY3FGGNipKIz5Z0U0Sjimc1t\nYYwxNq1qWDa3hTHGlF0NVc7UqUJ16Q0VkJ0Nb7xhc1sYY6qt8tos/lHOtm8iHUhcy86GSZNg/Xr3\n7IUxxlQzZSYLVe1XlYHEteAnuS1ZGGOqIWuz8KNbN/fT2i2MMdWUJQs/bG4LY0w1Z8nCLxv2wxhT\njfkZdfY1ETnPmxe7+rK5LYwx1ZifBPA4boa7FSIyTkSOj3JM8SnQyL1wYWzjMMaYGAibLFT1PW+a\n057AauA9EflYRK4SkXKftxCRASKyXERWishh06KKyGkiskBEikVkcND6bBGZJyJLRGShiMR+aBGb\n28IYU435qloSkUbAlcA1wJfAQ7jkMaucY2rgJkw6B+gEDBGR0MGV1njnfTFk/S7gV6raGRgAjBeR\nBn5ijRqb28IYU42FHUhQRKYBx+Pm3b5AVTd4m14WkfKGKu8NrFTVVd55pgCDcKPWAqCqq71tJcEH\nenNmBF6vF5GNQBNgm48yRYfNbWGMqcb8zJT3sKrOKW2DquaUc1xzYG3Q+zygzxHEBoCI9AZqAd+V\nsm0EMAIgKyuL3NzcIz39ETmuUSOaf/ghc2fPRmvUqPB5CgsLox5rLCVz+axsiSuZy1cVZfOTLOaJ\nyI3AKYACHwFPqGpRVCMDRKQp7o7m16paErpdVScAEwBycnK0b9++0Q0oLw9eeYXTjzkGOneu8Gly\nc3OJeqwxlMzls7IlrmQuX1WUzU+bxXNAZ+AR4FFc+8PzPo5bB7QMet/CW+eLiNQH3gJuV9VP/B4X\nVdbIbYyppvzcWXRR1eCG6TkisrTMvQ/6HGgvIm1xSeJSXBfcsESkFjANeE5Vp/o5pkoEz20xdGis\nozHGmCrj585igYicGHgjIn3wMQe3qhYDI4GZwDLgFVVdIiJjRWSgd66fiEgecBHwpIgs8Q6/GDgN\nuFJEvvKW7CMqWTTY3BbGmGrKz51FL+BjEVnjvW8FLBeRRYCqareyDlTVGcCMkHV3Br3+HFc9FXrc\nC8ALPmKreja3hTGmGvKTLAZEPYpE0qOHzW1hjKl2wiYLVf1BRLoDp3qr5qrq19ENK47Z3BbGmGrI\nz0CCvwMmA0d7ywsicn20A4tbNreFMaYa8lMNNQzoo6o7AUTkXmAeritt9VOvHrRrZ8nCGFOt+OkN\nJcD+oPf7vXXVlw37YYypZvzcWTwDfOqNEQXwc2BS9EJKANnZMHWqm9uiXr1YR2OMMVHnZ4jyB4Cr\ngC3ecpWqjo92YHHN5rYwxlQz5d5ZeMOML1HVE4AFVRNSAgjuEXXyybGNxRhjqkC5dxaquh/3AF6r\nKoonMTRrBo0bW7uFMaba8NNmcRSwREQ+A3YGVqrqwKhFFe9sbgtjTDXjJ1n8KepRJKLsbHjkESgu\nhpp+/hmNMSZx+ek6e66qfhC8AOdGO7C4l50Ne/bA8uWxjsQYY6LOT7LoX8q6cyIdSMKxuS2MMdVI\nmclCRK7zRpY9XkQWBi3fA4uqLsQ4FTy3hTHGJLnyKttfBN4G/gaMDlpfoKpbohpVIqhZE7p2tWRh\njKkWyryzUNXtqrpaVYcAecA+3BzcGdaV1hPoEaUa60iMMSaq/Iw6OxL4EZiFmxP7LeBNPycXkQEi\nslxEVorI6FK2nyYiC0SkWEQGh2z7tYis8JZf+ypNVcvOhs2b3dwWxhiTxPz0+RwFHK+q+UdyYu/p\n78dwDeR5wOciMl1Vg+fvXgNcCdwccmxDYAyQg7ub+cI7duuRxBB1NreFMaaa8NMbai2wvQLn7g2s\nVNVVqroXmAIMCt7Bq+ZaCJSEHPszYJaqbvESxCziccY+m9vCGFNN+LmzWAXkishbwJ7ASm+AwfI0\nxyWagDygj8+4Sjs2/r6629wWxphqwk+yWOMttbwlbojICGAEQFZWFrm5uVUeQ6fmzcmYN4/PjuDa\nhYWFMYm1qiRz+axsiSuZy1cVZfMzB/efAUSkjqruOoJzrwNaBr1v4a3ze2zfkGNzS4ltAjABICcn\nR/v27Ru6S/T17w8ffEDfXr18z22Rm5tLTGKtIslcPitb4krm8lVF2fz0hjpJRJYC33jvu4vI4z7O\n/TnQXkTaikgt4FJgus+4ZgJni8hRInIUcLa3Lv7Y3BbGmGrATwP3eFyDcz6Aqn4NnBbuIFUtBkbi\nPuSXAa+o6hIRGSsiAwFE5CcikgdcBDwpIku8Y7cAd+MSzufA2Lh9ENCG/TDGVAO+hktV1bUih0y7\nvb+sfUOOmwHMCFl3Z9Drz3FVTKUd+zTwtJ/rxJTNbWGMqQb8JIu1IvJTQEUkFfgd7k7BgM1tYYyp\nFvxUQ10L/BbXdXUdkO29NwHZ2bBokZvbwhhjkpCf3lCbgaFVEEviCp7bonPnWEdjjDER56c31H0i\nUl9EUkVktohsEpHLqyK4hGGN3MaYJOenGupsVd0BnA+sBtoBt0QzqIRjc1sYY5Kcn2QRqKo6D/i3\nqlZknKjkZnNbGGOSnJ9k8aaIfAP0AmaLSBOgKLphJSCb28IYk8TCJgtVHQ38FMhR1X3ATkJGjzXY\n3BbGmKTmp4H7ImCfqu4XkTuAF4BmUY8s0VgjtzEmifmphvqTqhaIyCnAWcAk4InohpWAbG4LY0wS\n85MsAkN7nAdMUNW3iLOhyuOCzW1hjElifpLFOhF5ErgEmCEiaT6Pq35s2A9jTJLy86F/MW7k2J+p\n6jagIfacRemys2HlSigoiHUkxhgTUX56Q+0CvgN+JiIjgaNV9d2oR5aIbG4LY0yS8tMb6nfAZOBo\nb3lBRK6PdmAJyXpEGWOSlJ8hyocBfVR1J4CI3AvMAx6JZmAJyea2MMYkKT9tFsKhkx3t99aZUDa3\nhTEmSflJFs8An4rIXSJyF/AJ7lmLsERkgIgsF5GVIjK6lO1pIvKyt/1TEWnjrU8VkWdFZJGILBOR\nW32XKNZsbgtjTBLy08D9AHAVsMVbrlLV8eGOE5EawGPAOUAnYIiIdArZbRiwVVXbAQ8C93rrLwLS\nVLUrbkyq3wQSSdwLntvCGGOSRLltFt4H/hJVPQFYcITn7g2sVNVV3rmm4MaUWhq0zyDgLu/1VOBR\ncZN9K1BXRGoCtYG9wI4jvH5sBDdy20RIxpgkUW6y8MaDWi4irVR1zRGeuzmwNuh9HtCnrH1UtVhE\ntgONcIljELABqAP8XlW3hF5AREYAIwCysrLIzc09whAjT/bv59TUVPL+8x9WNW9e6j6FhYVxEWu0\nJHP5rGyJK5nLVxVl89Mb6ihgiYh8hhtxFgBVHRi1qNxdyX7cgIVHAXNF5L3AXUpQDBOACQA5OTna\nt2/fKIZ0BLp3p1V+Pq3KiCc3N5e4iTUKkrl8VrbElczlq4qy+UkWf6rgudcBLYPet/DWlbZPnlfl\nlAnkA5cB73hDom8Ukf8COcAqEkF2Nrz+upvbQqzjmDEm8ZXZwC0i7UTkZFX9IHjBfePP83Huz4H2\nItJWRGoBlwLTQ/aZDvzaez0YeF9VFVgDnOHFURc4EfjmSAoWUza3hTEmyZTXG2o8pTcqb/e2lUtV\ni4GRuHGllgGvqOoSERkrIoEqrElAIxFZCdwIBLrXPgZkiMgSXNJ5RlUTZwwNe5LbGJNkyquGylLV\nRaErVXWR326sqjoDmBGy7s6g10W4brKhxxWWtj5hBOa2+PJLOO+82MZijDERUN6dRYNyttWOdCBJ\nxea2MMYkmfKSxXwRGR66UkSuAb6IXkhJwob9MMYkkfKqoUYB00RkKAeTQw5ulrwLox1YwsvOhqlT\nYccOqF8/1tEYY0yllJksVPVH4Kci0g/o4q1+S1Xfr5LIEl3w3BannBLbWIwxppLCPmehqnOAOVUQ\nS3IJ7hFlycIYk+BsLu1osbktjDFJxJJFtNjcFsaYJGLJIpqys2HxYti3L9aRGGNMpZTZZiEiBbih\nwg/bBKiqWhefcILntujSJfz+xhgTp8rrDVWvKgNJSsGN3JYsjDEJzHc1lIgcLSKtAks0g0oaxx8P\naWnWbmGMSXhhk4WIDBSRFcD3wAfAauDtKMeVHGrWhK5dLVkYYxKenzuLu3FDhH+rqm2BM4FPohpV\nMgn0iNLSmn+MMSYx+EkW+1Q1H0gRkRTvIb2cKMeVPLKzIT8f1oXO+2SMMYnDz0x520QkA/gQmCwi\nGwmaXtWEEdzI3aJFbGMxxpgK8nNnMQjYBfweeAf4DrggmkEllW7d3AN61m5hjElgfpLF0UAtVS1W\n1WeBiYCvbrUiMkBElovIShEZXcr2NBF52dv+afCkSiLSTUTmicgSEVkkIun+ihRnbG4LY0wS8JMs\n/g2UBL3f760rl4jUwE2Peg7QCRgiIp1CdhsGbFXVdsCDwL3esTWBF4BrVbUz0BdI3MegbdgPY0yC\n85Msaqrq3sAb73UtH8f1Blaat9LdAAAdOklEQVSq6irvmCm4Kq1gg4BnvddTgTNFRICzgYWq+rV3\nzXxV3e/jmvEpOxu++87NbWGMMQnITwP3JhEZqKrTAURkELDZx3HNgbVB7/OAPmXto6rFIrIdaAR0\nAFREZgJNgCmqel/oBURkBDACICsri9zcXB9hVb2GKSl0A7589lm2d+1KYWFh3MYaCclcPitb4krm\n8lVF2fwki2txvaAexY0LtRb4VVSjcnGdAvwE17g+W0S+UNXZwTup6gRgAkBOTo727ds3ymFVUIcO\ncOut9BCBvn3Jzc0lbmONgGQun5UtcSVz+aqibH4mP/oOONHrPouqFvo89zqgZdD7Ft660vbJ89op\nMoF83F3Ih6q6GUBEZgA9gdkkoqZNoUkTa7cwxiSs8kadvVxVXxCRG0PWA6CqD4Q59+dAexFpi0sK\nlwKXhewzHfg1MA8YDLyvqoHqpz+ISB1gL3A6rgE8MdncFsaYBFdeA3dd72e9MpZyqWoxMBKYCSwD\nXlHVJSIyVkQGertNAhqJyErgRmC0d+xW4AFcwvkKWKCqbx1h2eKLzW1hjElg5Q1R/qTX/XWHqlbo\nW72qzgBmhKy7M+h1EXBRGce+gOs+mxyC57YwxpgEU27XWa+76pAqiiW5BQ/7YYwxCcZPb6j/ej2h\nXiZoTChVXRC1qJJRhw6Qnm5jRBkD/O9/rka2Zcvw+5r44OehvGygMzAW+Ie33B/NoJKSzW1hDABz\n5kCnTm4UnFtvhUK//StNTIVNFqrar5TljKoIrqo89RRs21YFF7K5LUw19+STcPbZrjf5JZfAuHHu\npvuFF+zPIt75mSkvU0QeEJH53vIPEcmsiuCqwvLlcO21bnDYqD/c6c1tkbbZzwPwxiSP4mK44Qb3\nt3b22TBvHjz3HHzyiauVveIKOPlkmD8/1pGasviphnoaKAAu9pYdwDPRDKoqHX88fPyxa0444wy4\n5RbXaSkqvEbujJUro3QBY+LP1q1w7rnwyCNw000wfTrUr++29enjEsbTT7vh03r3hmuugY0bYxuz\nOZyfZHGcqo7xBgRcpap/Bo6NdmBVqXdv+PJL+M1v4P773fvFi6Nwoa5dQcSShak2vv0WTjzR3bVP\nmuT+vmrUOHSflBS46iq37403wrPPuqqp8ePtsaR44idZ7BaRUwJvRORkYHf0QoqNunXhiSfgP/9x\nPTVycuDBB6GkJPyxvnlzW1iyMNXBe++5O4ctW2D2bLj66vL3z8x0yWTRIjjpJPj976F7d3j33aqJ\n15TPT7K4DnhMRFaLyA/Ao7jBBZPS+ee7X9af/cx9yzn7bMjLi+AFsrMtWZik9/jjMGCAa4/47DM4\n9VT/x55wAsyY4b647d3r/hZ//nNYtSp68Zrw/PSG+kpVuwPdgK6q2iMwz0SyOvpoeP11mDjR1ad2\n7QpTpkTo5NnZ1F6/Hu67D3Yn3Q2aqeb27YPf/tYt55wD//0vtG175OcRcV/cliyBv/3N3aV06gS3\n325dbWPFT2+oG73BBK8BrvHeDxOR7OiHFzsirqHtq6/cN50hQ2Do0Ah0sR0xgvw+feCPf4T27V2/\n3eLiiMRsTCxt3eoSxOOPu44ir79+sCG7otLSYPRo155x0UXw17+6v8cXX7SutlXNTzVUDq7aqbm3\n/AYYAEwUkT9EMba40K4dzJ0LY8fCyy+7LrZz5lTihI0bs2jcONfi17IlDB8OXbrAq6/ab79JWMuX\nu/aJuXPhmWfcjXNoQ3ZlNGsGzz/v7lSOOcZ9cTv1VNcxxVQNP8miBdBTVW9S1ZuAXsDRwGnAlVGM\nLW7UrAl/+tPBLrZnnhmBLrann+5OOG2a6w4yeLDrNlKpTGRM1Xv3XZcotm2D99+HK6+M3rV++lPX\nBvLUU+5uo1cv14tx06boXdM4fpLF0UDwx+I+IEtVd4esT3qldbFdtKgSJxRxLXeLFrmO5hs2uIc9\nBgywr0wm7qm6ZyfOPRdat4bPP3cP1kVbSgoMG+aSxahR7k+nQwd4+GHrahtNfpLFZOBTERkjImOA\n/wIvikhdYGlUo4tDgS62b755sIvtAw9UsottjRoHO5r/4x/ur65nT9dQYj2nTBzatw+uu849lX3+\n+a56qHXrqo2hQQP3t/f11/CTn8Dvfueee33vvaqNo7rw0xvqbmAEsM1brlXVsaq6U1WHRjvAeHXe\nee7BvXPOcU+l9u8Pa9dW8qTp6a6/7qpVrtvH9OnQsSP83/+5zGRMHMjPd13Kn3zSNT6/9hpkZMQu\nnk6dYOZM16BeVOT+Fn/xC/j++9jFlIz83FkApOMmQXoI+MGbKrXaa9LENTlMnAiffuoavyPSxTYz\nE+65x91VDB/uLnDccXDHHbB9ewQuYEzFLFvm2ic+/tiN7fS3v7lqoVgTgUGDXFfbv/zFJY+OHeHO\nO2HnzvDHm/D8dJ0dA/wRuNVblYrPGexEZICILBeRlSIyupTtaSLysrf9UxFpE7K9lYgUisjNfq4X\nC1HrYgtuaM7HH3d/oQMHur+CY491VVVFRRG4gDH+vfOO64NRUOA6811xRawjOlx6Otx2m+ud9ctf\nwt13u7/LKVOss2Fl+flOcCEwEG/iI1Vdj485uL0pWR8DzgE6AUNEpFPIbsOAraraDngQuDdk+wPA\n2z5ijLmId7ENPflLL8GCBa5y9uab3TMaTz9tz2iYqFOFhx5yVa9t27omtZNOinVU5WvRAiZPdn+T\nTZq4L3E339yd776LdWSJy0+y2KuqCiiA17DtR29gpTf44F5gCjAoZJ9BwLPe66nAmSIi3nV+DnwP\nLPF5vZgLdLGdNw9q13ZdbG++OYKj2Pbo4b7evf++63g+bJjLSq+/bl+bTFTs3QsjRrheR4MGwUcf\nQatWsY7Kv1NOccnt8cdh+fJ6dOkC995rvaYqQjTMh4xXBdQe6A/8DbgaeElVHw5z3GBggKpe472/\nAuijqiOD9lns7ZPnvf8O6AMUAbO8a94MFKrqYbPzicgIXOM7WVlZvaZEbEyOytu9O4V//vM4pk9v\nzrHHFnL77cs49lhXeVpYWEhGZVsEVWk8dy7HPvUUddauZXunTqwaPpzt2bF/sD4i5YtT1als27en\nMmZMZ77+ugGXX/4DV131fVy0T1TUDz8UM2lSd+bObcJxxxVyyy3LOf74gliHFRGV+b3s16/fF6qa\nE3ZHVQ274D60/46bTrW/z2MGA08Fvb8CeDRkn8VAi6D33wGNvetc7K27C7g53PV69eql8ejNN1WP\nPlq1Vi3V++9X3b9fdc6cOZG7wL59qhMnqjZvrgqq55yj+uWXkTt/BUS0fHGmupRt8WLVY49VTUtT\nnTw5djFFUqB8r72m2rSpakqK6u9/r1pQENu4Kuvrr1XvvffrCh8PzFcfn+l+GrjvVdVZqnqLqt6s\nqrNEJLRtoTTrgODp2Ft460rdR0RqAplAPu7u4j4RWQ2MAm4TkZEkoOAutjffDGedBRs3pkXuAjVr\nuhb2FSvcGAuffOKqq4YOtWE6TYW89ZZrk9i1Cz74AC67LNYRRdaFF7o+I7/5jZuGoEsXeDshWkYP\ntXChG/ihe3d4/PHjol4TXdPHPv1xvaGCnVPKulCfA+29brbrgEuB0F+76cCvgXm4O5H3vUx3YEBj\nEbkLVw31qI9Y41Kgi+3TT7sHh3JzT2T4cDe9RUaG+1nWUt72wLbatUFq13ZjkAwf7iplH3oI/v1v\n9xdx++1uQB1TKaqwf7+g6nrBJRtV19Hullvc94033nANxckoM9O1Ywwd6v5kzj3XNYKPH+9GnY5n\nCxe6jjSvvuoGarzzTsjJ+ZKgaYeiosxkISLXAf8HHCsiC4M21cM9xV0uVS327gZmAjWAp1V1iYiM\nxd32TAcmAc+LyEpgCy6hJCUR1x7dty/cffcPNGjQhoICDllWr3bDLwfe++0dW6NGcFJpQL16f6N1\nz+u5et1Y+j/2BCVPTOT706+i5Pc3cezZ7ahVK5olTVw7d7oHK9euhTVrSv+5e/fpgPv/rFHD35KS\n4n/f0KVmzapbHnroeN55x31b/de/3GgFye7kk93IOuPGuRFt33nHJcwrr4y/LwSlJYlRo+CooyA3\nN/q9Ists4BaRTOAoXKN28DMSBaq6JeqRHaGcnBydnyCzvefm5tK3b9+w++3bd2jyCH4dupS1rdGW\nFQxd/3eu0GepSTGvyS+Z2vYP1OiTQ9eurjNV165uANxI/XH4LV9VKi6G9evLTgJr1rgZ3YKJuBuy\nVq3cv0+rVrBly/e0atWW/fspdSkpKX39ke4TvBQX+1/276/cv9Of/gR33RUfD9pFWrjfy2XLXM+v\njz5yQ7Q9+aTrtR5roUli1KiDSSKgMn9zIuKrgbvMOwtV3Q5sB4Z4Jzwa9yR3hohkqOqaCkVmfEtN\ndb8Qwb8UR649+/ZN4LuP/kzJ+Ic5f+YTXLTq33y8rh9/fumP3MbZgJCZySHJo1s3V5db2fkIqoKq\nG4KitAQQeL1+/eHjdzVocDARnHTSwYQQ+NmsGYfdheXm/kDfvvE7gIGrKjuyBBNYvv32M371q96x\nLkLMdOzo2mgmToQ//MH9HYwZ44bzSU2t+njKu5OIhbBtFiJyAe7huGbARqA1sAzoHN3QTKSkpsIJ\n/ZpCv7/Bjlth4kR++uCDzFw3gJ3HdWPeqX9gWq1L+HpJTV54AXbsOHhs69aHJpCuXd0InzX9tHZV\nUEmJewJ+82a35OeX/frHH10yCK2yS0s7+KF/5pmHJoGWLd1SL+yjpYlH5GC10pEqKtoV+YASTEqK\na+a74AK4/nq49Vb3POzEiW6U6aoQb0kiwM+v1D3AicB7qtpDRPoBl0c3LBM19eu7r0rXXw8vvkjd\n++7jrH9dzlmtb4cbb0TfGcaa/LosWuR+aRctcsuMGQerOGrVcoO3hd6JHHPM4VVZJSVuBrXyPvBD\nX2/ZUvYovrVqQePGbmnUyDXEDhx4eDJo0iT+6pxN4mjWzH1Yv/46jBzp7jyvv94N2Ratx2ziNUkE\n+EkW+1Q1X0RSRCRFVeeIyPioR2aiq1Yt14r3q1+5vpL33Qe/+x3y5z/TeuRIWo8cyfnnNzmw+549\n8M03hyaQ2bPd7GUBjRpB586wZUsP9u1zH/7hPvibNHHHNW7sEk7gdeBn6Ou6dS0JmKrz85+79otb\nb3XzZUyb5qYoOPfcyF0j3pNEgJ9ksU1EMoAPgckishFvnCiTBFJS3D33BRe4oUTvu8/95v7973D1\n1W7I9GOPJS3N9efu3v3Qw/PzDyaPRYtg6VJITS2hU6fyP/QbNbIPfpMY6teHxx472M32vPPgkktc\n7/SsrIqfN1GSxAHhntoD6uLGkKqJeybiBqCRnyf+qnKJ1ye4SxP3TwEvXap69dWqqanuMddLL1X9\n4gvfh8d9+SrBypa4IlG+PXtUx451IzIcdZTqpEmqJSVHdo6vv1b95S/dgAv166veeafqli2Vi6sy\nZaOyT3CLSDsROVndJEclqlqsqs8CC4AG0U9jJmY6doRJk9yDHzfd5KqpevVys8rMmmWDFppqq1Yt\n1734669db8Fhw1wHihUrwh8b/MT1rFnuTmL1avjzn+P4biJIeb2pxwM7Slm/3dtmkl2zZq5aau1a\n91T4kiVuirRevdwEATY8uqmmTjjBzekxYYKbOaBrV/dQX2mj2SZ6kggoL1lkqeqi0JXeujZRi8jE\nn8xM1/H8++/hqafcoEFDhrg+tI895t4bU82kpLg2jGXLXJPf7be771Gffuq2J0uSCCgvWZRX1VQ7\n0oGYBJCW5u67ly513UKysly/wtatXUtdfn6sIzSmyjVt6oZhe+MN1/vvpJPcMCLJkiQCyksW80Vk\neOhKEbkG+CJ6IZm4l5Li+hR+/DF8+KGba3PMGPeQww030PCTT9yIt5Ude8KYBDJwoPse9dvfuhED\nkiVJBJTXdXYUME1EhnIwOeQAtXBTrZrqTgROPdUtS5a47rZPPEG34mLXMT093VVVdezoKnkDPzt0\ncEPlGpNk6teHRx5xS7Ipb2yoH4Gfek9sd/FWv6Wq71dJZCaxdO7shiodP54Fzz9Pzzp1XGXusmVu\nXstXXjnYi0rETeYcnEACPxs1imkxjDGlC/tQnqrOAeZUQSwmGTRowI6uXd1Y7MF273b9C5ctc4+C\nB36+//6hAzs1aVJ6EmnVKjmHQjUmQURxODhjgtSu7cbz6Nbt0PX797uhYYOTyLJlMHXqoWOG16kD\nxx9/eCJp3941vBtjosqShYmtGjVclVTbtocPuLNp06F3IcuWwbx5bhjQ4OM7d4aePQ8u3btHb7Q3\nY6qpqCYLERkAPISbKe8pVR0Xsj0NeA7ohZt7+xJVXS0i/YFxuMb0vcAt1lZSDTVp4pZTTz10/a5d\n8O23LnksXuymOpsxw7WZgGsT6dDh0ATSo0dydEkxJkailixEpAbwGG4O7zzgcxGZrqpLg3YbBmxV\n1XYicilwL3AJsBm4QFXXi0gX3NSszaMVq0kwdepAdrZbAlRhwwb3OG1g+e9/D70Ladv20ATSs2f8\nT7hsTJyI5p1Fb2Clqq4CEJEpwCAgOFkMAu7yXk8FHhURUdUvg/ZZAtQWkTRV3RPFeE0iE3HDkzRr\nBueff3D95s3uziM4ibz66sHtzZsfnkCaN7fhcI0JUeYc3JU+schgYICqXuO9vwLoo6ojg/ZZ7O2T\n573/zttnc8h5rlXVs0q5xghgBEBWVlavKVOmRKUskVZYWEhGEtepx3v5ahQWkrFyJfVWrCBjxQrq\nffstddauRbyJN/Y2aEBh+/YUtG9PYYcOFLRvT1HTpiAS92WrjGQuGyR3+SpTtn79+lVuDu54ICKd\ncVVTZ5e2XVUnABMAcnJytKITlle1ykyunggSsnw7d7rBfBYsoNaXX9JwwQIa/vvfB0eGy8yEHj3Y\nUKcOTdu3dw3odeu6n8FLWetq1Iht+XxIyP+3I5DM5auKskUzWawDWga9b+GtK22fPBGpCWTiGroR\nkRbANOBXqvpdFOM0xn2gn3SSWwL27HFPpgeqr774goaLF8NHH0FhYdlTAJYmPd1/Ygl+n5npGuaD\nl9q1rZrMVLloJovPgfYi0haXFC4FLgvZZzpuQqV5wGDgfVVVEWkAvAWMVtX/RjFGY8qWlnawHcMz\nL/ANTtUlk8JCt+zcefD1kazbvPnw7eHUqgUNGhyeRAJLedsyMizRmAqJWrJQ1WIRGYnryVQDeFpV\nl4jIWNzMTNOBScDzIrIS2IJLKAAjgXbAnSJyp7fubFXdGK14jTkiIu5uIT3dzRMbKSUl7mn3wkIo\nKIDt22Hr1vKXTZtcV+KtW2HbtvLveGrUKDOZtC0ocLP6HH30oUujRlAzrmusTRWI6m+Aqs4AZoSs\nuzPodRFwUSnH3QPcE83YjIlLKSmuKqpu3YpN8FxS4pJMaELZtq3sZPP997BtG63y82Hy5MPPKeIS\nRlbW4Ynk6KMPX293L0nJvi4Yk0xSUlw7R2YmtGlzRId+8P779M3Oho0b3fLjjwdfBy9fful+bttW\n+onS08tOJKFJpnFjSE2tfLlN1FmyMMY4KSnQsKFbTjgh/P579rgqsNISSiDZbNjgqrY2boS9e0s/\nT/BdS1ZW+a9taPuYsWRhjKmYtDRo0cIt4ajCjh2H37X8+OOhrxcscK+3by/9PBkZ4RNK4HVmplWH\nRZAlC2NM9IkcrB5r3z78/kVFhyaW0pLLypVuSJfNmw/OlRKsVq1Dkshxdeq4GRy7dnWDT9apE/ly\nJjFLFsaY+JOe7uYwadUq/L7FxW7+97KSyo8/wvr1NAsMfQ8ueR13nEscXbq4n127Qrt21vOrDPav\nYoxJbDVrHqyCKsfc2bPp26qVG6l40aKDyxtvHOxunJYGnTodTB6BxRvupTqzZGGMqR5q1HBVYO3b\nw4UXHly/e7cb7j44gcyaBc89d3Cfhg0PvQMJ3JHUrx/9uAPdobdvdz3QAj+DXjfdsuXw2SkjzJKF\nMaZ6q137sCf1AVe1FXoX8txz7oM7oHXrw+9COnRw7SUBe/Yc/kFfygd+mT937Ci9TSbIMZ07R/Af\npHSWLIwxpjSNGsHpp7slQBV++OHQBLJ4Mbzzjms7AffcSMuWbuiW7dsPnWO+NMHPxjRocPAZmcDr\ncD8zM/ly3jz6RuvfwWPJwhhj/BJxH+Rt2sAFFxxcv3cvLF9+MIH88APUq+fvAz9Bnni3ZGGMMZVV\nq9bBaqgklRLrAIwxxsQ/SxbGGGPCsmRhjDEmLEsWxhhjwrJkYYwxJixLFsYYY8KyZGGMMSYsSxbG\nGGPCEg0z5kiiEJFNwA+xjsOnxsDmWAcRRclcPitb4krm8lWmbK1VtUm4nZImWSQSEZmvqjmxjiNa\nkrl8VrbElczlq4qyWTWUMcaYsCxZGGOMCcuSRWxMiHUAUZbM5bOyJa5kLl/Uy2ZtFsYYY8KyOwtj\njDFhWbIwxhgTliWLKiQiLUVkjogsFZElIvK7WMcUaSJSQ0S+FJE3Yx1LpIlIAxGZKiLfiMgyETkp\n1jFFioj83vudXCwiL4lIeqxjqgwReVpENorI4qB1DUVklois8H4eFcsYK6qMsv3d+71cKCLTRKRB\npK9ryaJqFQM3qWon4ETgtyLSKcYxRdrvgGWxDiJKHgLeUdUTgO4kSTlFpDlwA5Cjql2AGsClsY2q\n0v4FDAhZNxqYrartgdne+0T0Lw4v2yygi6p2A74Fbo30RS1ZVCFV3aCqC7zXBbgPm+axjSpyRKQF\ncB7wVKxjiTQRyQROAyYBqOpeVd0W26giqiZQW0RqAnWA9TGOp1JU9UNgS8jqQcCz3utngZ9XaVAR\nUlrZVPVdVS323n4CtIj0dS1ZxIiItAF6AJ/GNpKIGg/8ASiJdSBR0BbYBDzjVbM9JSJ1Yx1UJKjq\nOuB+YA2wAdiuqu/GNqqoyFLVDd7r/wFZsQwmiq4G3o70SS1ZxICIZACvAqNUdUes44kEETkf2Kiq\nX8Q6liipCfQEnlDVHsBOErca4xBe3f0gXEJsBtQVkctjG1V0qXtmIOmeGxCR23HV3ZMjfW5LFlVM\nRFJxiWKyqr4W63gi6GRgoIisBqYAZ4jIC7ENKaLygDxVDdwJTsUlj2RwFvC9qm5S1X3Aa8BPYxxT\nNPwoIk0BvJ8bYxxPRInIlcD5wFCNwgN0liyqkIgIrs57mao+EOt4IklVb1XVFqraBtc4+r6qJs23\nU1X9H7BWRI73Vp0JLI1hSJG0BjhRROp4v6NnkiSN9yGmA7/2Xv8aeCOGsUSUiAzAVQEPVNVd0biG\nJYuqdTJwBe5b91fecm6sgzK+XQ9MFpGFQDbw1xjHExHe3dJUYAGwCPe5kNBDY4jIS8A84HgRyROR\nYcA4oL+IrMDdTY2LZYwVVUbZHgXqAbO8z5V/Rvy6NtyHMcaYcOzOwhhjTFiWLIwxxoRlycIYY0xY\nliyMMcaEZcnCGGNMWJYsjIkDItI3GUfqNcnDkoUxxpiwLFkYcwRE5HIR+cx78OlJb/6OQhF50JsP\nYraINPH2zRaRT4LmGDjKW99ORN4Tka9FZIGIHOedPiNovozJ3tPUxsQFSxbG+CQiHYFLgJNVNRvY\nDwwF6gLzVbUz8AEwxjvkOeCP3hwDi4LWTwYeU9XuuDGYAiOh9gBGAZ2AY3FP/BsTF2rGOgBjEsiZ\nQC/gc+9Lf23cYHQlwMvePi8Ar3nzXzRQ1Q+89c8C/xaRekBzVZ0GoKpFAN75PlPVPO/9V0Ab4KPo\nF8uY8CxZGOOfAM+q6iGzkInIn0L2q+gYOnuCXu/H/j5NHLFqKGP8mw0MFpGj4cCczq1xf0eDvX0u\nAz5S1e3AVhE51Vt/BfCBN0Ninoj83DtHmojUqdJSGFMB9s3FGJ9UdamI3AG8KyIpwD7gt7iJkHp7\n2zbi2jXADYP9Ty8ZrAKu8tZfATwpImO9c1xUhcUwpkJs1FljKklEClU1I9ZxGBNNVg1ljDEmLLuz\nMMYYE5bdWRhjjAnLkoUxxpiwLFkYY4wJy5KFMcaYsCxZGGOMCev/AU4hOhXT77cpAAAAAElFTkSu\nQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "kd6UgNhyqb8B"
      },
      "source": [
        "### CNN WITH 6 LAYERS AS A SIGMOID ACTIVATION FUNCTION"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "X88q39lDqb8B"
      },
      "source": [
        "### BN+DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "REda_Yysqb8B",
        "outputId": "5458f23d-7f6b-4f58-904c-48ed00775677",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 890
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(Dropout(0.25))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_22\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_85 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_85 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_86 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_22 (Batc (None, 22, 22, 64)        88        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_86 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_87 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "dropout_28 (Dropout)         (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_87 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_88 (Conv2D)           (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "dropout_29 (Dropout)         (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_88 (MaxPooling (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_89 (Conv2D)           (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "batch_normalization_23 (Batc (None, 6, 6, 16)          24        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_89 (MaxPooling (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_90 (Conv2D)           (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "batch_normalization_24 (Batc (None, 5, 5, 16)          20        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_90 (MaxPooling (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_22 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_43 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_22 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dropout_30 (Dropout)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_44 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,158\n",
            "Trainable params: 122,092\n",
            "Non-trainable params: 66\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "iYC1s608qb8C",
        "outputId": "5037c2f3-fbff-4eaa-b4d9-40a3d626138a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 12s 202us/step - loss: 2.3139 - acc: 0.1067 - val_loss: 2.3014 - val_acc: 0.1135\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 8s 130us/step - loss: 2.3016 - acc: 0.1123 - val_loss: 2.3010 - val_acc: 0.1135\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 8s 130us/step - loss: 1.0131 - acc: 0.6374 - val_loss: 0.1495 - val_acc: 0.9534\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 8s 133us/step - loss: 0.1797 - acc: 0.9486 - val_loss: 0.0766 - val_acc: 0.9760\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 8s 129us/step - loss: 0.1295 - acc: 0.9628 - val_loss: 0.0598 - val_acc: 0.9822\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 8s 127us/step - loss: 0.1069 - acc: 0.9695 - val_loss: 0.0509 - val_acc: 0.9846\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 8s 127us/step - loss: 0.0941 - acc: 0.9728 - val_loss: 0.0509 - val_acc: 0.9834\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 8s 127us/step - loss: 0.0831 - acc: 0.9754 - val_loss: 0.0474 - val_acc: 0.9850\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 8s 128us/step - loss: 0.0815 - acc: 0.9757 - val_loss: 0.0368 - val_acc: 0.9885\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 8s 127us/step - loss: 0.0750 - acc: 0.9783 - val_loss: 0.0349 - val_acc: 0.9889\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0701 - acc: 0.9792 - val_loss: 0.0334 - val_acc: 0.9891\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0661 - acc: 0.9805 - val_loss: 0.0380 - val_acc: 0.9876\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "0CpEfIYVqb8E",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "fXb-ccqEqb8F",
        "outputId": "91cecb40-2982-45f8-bd84-b8dc6ce4c0ab",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.03795005509173498\n",
            "Test accuracy: 0.9876\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "futZ0mKPqb8G"
      },
      "source": [
        "### WITHOUT BN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "BfPI7y90qb8H",
        "outputId": "4eb7e54b-a95f-4280-981b-e9d95b10dd6e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(Dropout(0.25))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('sigmoid'))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_23\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_91 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_91 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_92 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_92 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_93 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "dropout_31 (Dropout)         (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_93 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_94 (Conv2D)           (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "dropout_32 (Dropout)         (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_94 (MaxPooling (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_95 (Conv2D)           (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_95 (MaxPooling (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_96 (Conv2D)           (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_96 (MaxPooling (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_23 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_45 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_23 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dropout_33 (Dropout)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_46 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,026\n",
            "Trainable params: 122,026\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "R70WTeIDqb8I",
        "outputId": "4f5ae7e8-6235-49af-c58a-04890a7f3d02",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 11s 176us/step - loss: 2.3127 - acc: 0.1093 - val_loss: 2.3013 - val_acc: 0.1135\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 2.3015 - acc: 0.1122 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 2.3013 - acc: 0.1123 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3010 - val_acc: 0.1135\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3010 - val_acc: 0.1135\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 6s 104us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3011 - val_acc: 0.1135\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3010 - val_acc: 0.1135\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 6s 102us/step - loss: 2.3012 - acc: 0.1124 - val_loss: 2.3010 - val_acc: 0.1135\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 6s 103us/step - loss: 2.3013 - acc: 0.1124 - val_loss: 2.3010 - val_acc: 0.1135\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "iU8WELo5qb8J",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "axKTrYtdqb8K",
        "outputId": "97168d71-6115-48f4-b57a-42204ba2f4bd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.03929063466354273\n",
            "Test accuracy: 0.9873\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLCkYY4ypY0nBGGNMnf8PZt0BHuZvpUYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Ju3ITy34qb8L"
      },
      "source": [
        "### WITHOUT DROPOUT"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "y9FKnQjwqb8L",
        "outputId": "661dd60e-4c25-4c56-d956-5ed89e7af326",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 790
        }
      },
      "source": [
        "model=Sequential()\n",
        "#Adding first layer\n",
        "model.add(Conv2D(64, kernel_size=(5, 5),activation='sigmoid',input_shape=input_shape))\n",
        "model.add(MaxPooling2D(pool_size=(1, 1)))\n",
        "\n",
        "#Adding Second Layer\n",
        "model.add(Conv2D(64,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "#Adding Third Layer\n",
        "model.add(Conv2D(32,kernel_size=(3,3),activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fourth layer\n",
        "model.add(Conv2D(32, kernel_size=(3, 3), activation='sigmoid'))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Fifth layer\n",
        "model.add(Conv2D(16, kernel_size=(2, 2), activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "#Sixth Layer\n",
        "model.add(Conv2D(16,kernel_size=(2,2),activation='sigmoid'))\n",
        "model.add(BatchNormalization(axis=1))\n",
        "model.add(MaxPooling2D(pool_size=(1,1)))\n",
        "\n",
        "\n",
        "\n",
        "#Fully Connected layer\n",
        "#Flattening\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(Activation('sigmoid'))\n",
        "\n",
        "#Softmax\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_24\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_97 (Conv2D)           (None, 24, 24, 64)        1664      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_97 (MaxPooling (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_98 (Conv2D)           (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_25 (Batc (None, 22, 22, 64)        88        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_98 (MaxPooling (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_99 (Conv2D)           (None, 9, 9, 32)          18464     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_99 (MaxPooling (None, 9, 9, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_100 (Conv2D)          (None, 7, 7, 32)          9248      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_100 (MaxPoolin (None, 7, 7, 32)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_101 (Conv2D)          (None, 6, 6, 16)          2064      \n",
            "_________________________________________________________________\n",
            "batch_normalization_26 (Batc (None, 6, 6, 16)          24        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_101 (MaxPoolin (None, 6, 6, 16)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_102 (Conv2D)          (None, 5, 5, 16)          1040      \n",
            "_________________________________________________________________\n",
            "batch_normalization_27 (Batc (None, 5, 5, 16)          20        \n",
            "_________________________________________________________________\n",
            "max_pooling2d_102 (MaxPoolin (None, 5, 5, 16)          0         \n",
            "_________________________________________________________________\n",
            "flatten_24 (Flatten)         (None, 400)               0         \n",
            "_________________________________________________________________\n",
            "dense_47 (Dense)             (None, 128)               51328     \n",
            "_________________________________________________________________\n",
            "activation_24 (Activation)   (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_48 (Dense)             (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 122,158\n",
            "Trainable params: 122,092\n",
            "Non-trainable params: 66\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "3-fjZZfMqb8M",
        "outputId": "bc4d1846-e2cd-4a50-ba15-76efea177885",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        }
      },
      "source": [
        "#Compiling with Cross-entropy-loss and Adam Optimizer taking Accuracy as a metric\n",
        "from keras.optimizers import Adam\n",
        "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
        "history=model.fit(X_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(X_test, y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/12\n",
            "60000/60000 [==============================] - 13s 221us/step - loss: 0.9769 - acc: 0.6535 - val_loss: 0.1304 - val_acc: 0.9664\n",
            "Epoch 2/12\n",
            "60000/60000 [==============================] - 7s 125us/step - loss: 0.1054 - acc: 0.9695 - val_loss: 0.1018 - val_acc: 0.9700\n",
            "Epoch 3/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0687 - acc: 0.9803 - val_loss: 0.0758 - val_acc: 0.9770\n",
            "Epoch 4/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0519 - acc: 0.9853 - val_loss: 0.0518 - val_acc: 0.9839\n",
            "Epoch 5/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0431 - acc: 0.9874 - val_loss: 0.0447 - val_acc: 0.9855\n",
            "Epoch 6/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0363 - acc: 0.9895 - val_loss: 0.0366 - val_acc: 0.9889\n",
            "Epoch 7/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0310 - acc: 0.9906 - val_loss: 0.0520 - val_acc: 0.9837\n",
            "Epoch 8/12\n",
            "60000/60000 [==============================] - 7s 125us/step - loss: 0.0289 - acc: 0.9910 - val_loss: 0.0363 - val_acc: 0.9881\n",
            "Epoch 9/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0242 - acc: 0.9927 - val_loss: 0.0392 - val_acc: 0.9881\n",
            "Epoch 10/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0214 - acc: 0.9935 - val_loss: 0.0295 - val_acc: 0.9913\n",
            "Epoch 11/12\n",
            "60000/60000 [==============================] - 8s 126us/step - loss: 0.0188 - acc: 0.9944 - val_loss: 0.0320 - val_acc: 0.9900\n",
            "Epoch 12/12\n",
            "60000/60000 [==============================] - 8s 125us/step - loss: 0.0169 - acc: 0.9949 - val_loss: 0.0393 - val_acc: 0.9873\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "A7phV_cXqb8N",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "def plt_dynamic(x, vy, ty, ax, colors=['b']):\n",
        "    ax.plot(x, vy, 'b', label=\"Validation Loss\")\n",
        "    ax.plot(x, ty, 'r', label=\"Train Loss\")\n",
        "    plt.legend()\n",
        "    plt.grid()\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ii9Igv6xqb8P",
        "outputId": "e758e436-5cc9-4d4b-bb0c-d8cf4f1477ac",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "score = model.evaluate(X_test, y_test, verbose=0) \n",
        "print('Test score:', score[0]) \n",
        "print('Test accuracy:', score[1])\n",
        "\n",
        "fig,ax = plt.subplots(1,1)\n",
        "ax.set_title('Epoch vs Loss')\n",
        "ax.set_xlabel('epoch') ; ax.set_ylabel('Categorical Crossentropy Loss')\n",
        "\n",
        "# list of epoch numbers\n",
        "x = list(range(1,epochs+1))\n",
        "\n",
        "\n",
        "vy = history.history['val_loss']\n",
        "ty = history.history['loss']\n",
        "plt_dynamic(x, vy, ty, ax)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test score: 0.03929063466354273\n",
            "Test accuracy: 0.9873\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl8FeW9+PHPNxshe9gimwQFE/Yt\ngtYtCCLVKrXXDcFercqtv7Zu1avtta21y1WvF6nVWq1bWxdUFMtVLCglotaCgIqCIChh35csQCDL\n9/fHTMIhZJkkZ3K27/v1mtc5M2fOzPdhme955nnmeURVMcYYYwDiQh2AMcaY8GFJwRhjTB1LCsYY\nY+pYUjDGGFPHkoIxxpg6lhSMMcbUsaRgTCNEREWkX6jjMKY9WVIwEUFEikXkkIiUByyPhDquYBKR\nXDcRJYQ6FhO77B+fiSQXqeo7oQ7CmGhmNQUT8UTkGhH5QEQeEZESEVktIuMCPu8hInNEZK+IrBOR\nGwI+ixeRn4rIVyJSJiLLRKR3wOHHi8haEdkvIo+KiDRw/h5uLaZTwLYRIrJbRBJFpJ+IvOvGtltE\nXmpFGTuIyAwR2eouM0Skg/tZFxF5w41xr4i8JyJx7md3isgWt2xrAv9cjGmIJQUTLcYAXwFdgF8A\nrwVcpGcCm4EewKXAb0XkXPez24DJwAVABvA94GDAcb8FnAoMBS4Hzq9/YlXdCnwI/FvA5quAWapa\nCfwKmA9kA72A37eifP8FnAYMB4YBo4G73c9+7JavK5AD/BRQEckDfgicqqrpbuzFrTi3iSGWFEwk\ned39NVy73BDw2U5ghqpWqupLwBrgQvdX/xnAnapaoaqfAE8C33W/dz1wt6quUcenqron4Lj3qep+\nVd0ILMS5KDfkBZzkglubuNLdBlAJ9AF6uDG834qyTwHuVdWdqroL+CVwdcDxuwN93PK/p86gZtVA\nB2CgiCSqarGqftWKc5sYYknBRJJvq2pWwPKngM+26LGjO27AqRn0APaqalm9z3q673vj1DAasz3g\n/UEgrZH9XgVOF5HuwNlADfCe+9l/AgIsEZGVIvK9Js7XmB5u3LVqywfwP8A6YL6IfC0idwGo6jrg\nFuAeYKeIzBSRHhjTBEsKJlr0rHe//0Rgq7t0EpH0ep9tcd9vAk5u68lVdR/OLaIrcG4dzaxNUqq6\nXVVvUNUewH8Af2hFV9etOLWNWrXlQ1XLVPXHqnoScDFwW23bgaq+oKpnut9V4P5WF9LEBEsKJlp0\nA25yG3YvAwYAc1V1E/BP4L9FJFlEhgLXAc+533sS+JWI9BfHUBHp3MoYXsC5LXUpR28dISKXiUgv\nd3UfzsW5ponjdHBjrV3igBeBu0Wkq4h0AX5eWwYR+ZbbmC1ACc5toxoRyRORc90G6QrgUDPnNca6\npJqI8n8iUh2w/raqXuK+Xwz0B3YDO4BLA9oGJgN/xPllvQ/4RUDX1uk4993n4zRSrwZqj9lSc3CS\nzEZV/TRg+6nADBHJdGO7WVW/buI45fXWzwN+jdMQvsLd9oq7DZxyP4LT0LwP+IOqLnQT4H04CbIS\nJzlOa2XZTIwQm2THRDoRuQa43r1NYoxpA7t9ZIwxpo4lBWOMMXXs9pExxpg6VlMwxhhTJ+J6H3Xp\n0kVzc3NDHYYnBw4cIDU1NdRh+CKaywbRXT4rW+RqS/mWLVu2W1W7NrdfxCWF3Nxcli5dGuowPCkq\nKqKwsDDUYfgimssG0V0+K1vkakv5RGRD83v5ePtIRJ4WkZ0i8nkjn4uIPOyOWrlCREb6FYsxxhhv\n/GxTeBaY2MTn38R56KY/zgM1j/kYizHGGA98SwqqugjY28Quk4C/uCNT/gvIcgcTM8YYEyK+dkkV\nkVzgDVUd3MBnb+AMS/y+u74AZ3jj4xoMRGQa7uP5OTk5o2bOnOlbzMFUXl5OWlpjg2pGtmguG0R3\n+byUTURITU0lPj6+naIKDlWlgXmQooaX8lVXV3PgwAHqX9vHjh27TFULmjtHRDQ0q+oTwBMABQUF\nGikNSdHc6BXNZYPoLp+Xsq1fv5709HQ6d+4cURfZsrIy0tPTm98xQjVXPlVlz549lJWV0bdv31ad\nI5TPKWzBGcu+Vi+ODmdsjAmhioqKiEsIxqnhde7cmYqKilYfI5RJYQ7wXbcX0mlAiapuC2E8xpgA\nlhAiU1v/3vzskvoizry1eSKyWUSuE5Hvi8j33V3mAl/jzBj1J+D/+RULAB98AHfdBTashzHGNMrP\n3keTVbW7qiaqai9VfUpV/6iqf3Q/V1X9gaqerKpDGmpgDqply+D++2HHDl9PY4xpu7FjxzJv3rxj\nts2YMYMbb7yxye917+50YNy6dSuXXnppg/sUFhY2+wDsjBkzOHjwYN36BRdcwP79+72E3qR77rmH\nBx98sM3H8VPsjH2Un++8rlkT2jiMMc2aPHky9XsZzpw5k8mTJ3v6fo8ePZg1a1arz18/KcydO5es\nrKxWHy+SxE5SyMtzXlevDm0cxphmXXrppbz55pscOXIEgOLiYrZu3cpZZ51FeXk548aNY+TIkQwZ\nMoS//e1vx32/uLiYwYOdnvCHDh3iyiuvZMCAAVxyySUcOnSobr8bb7yRgoICBg0axC9+8QsAHn74\nYbZu3crYsWMZO3Ys4Ayvs3v3bgCmT5/O4MGDGTx4MDNmzKg734ABA7jhhhsYNGgQEyZMOOY8zWno\nmAcOHODCCy9k2LBhDB48mJdeegmAu+66i4EDBzJ06FBuv/32Fv25ehERXVKDondv6NjRagrGtNAt\nt8AnnwT3mMOHg3vta1CnTp0YPXo0b731FpMmTWLmzJlcfvnliAjJycnMnj2bjIwMdu/ezWmnncbF\nF1/caAPrY489RkpKCl988QUrVqxg5MijI+r85je/oVOnTlRXVzNu3DhWrFjBTTfdxPTp01m4cCFd\nunQ55ljLli3jmWeeYfHixagqY8aM4ZxzziE7O5u1a9fy4osv8qc//YnLL7+cV199lalTpzb7Z9HY\nMb/++mt69OjBm2++CUBJSQl79uxh9uzZrF69GhEJyi2t+mKnphAXB6ecYjUFYyJE4C2kwFtHqspP\nf/pThg4dyvjx49myZQs7mmgrXLRoUd3FeejQoQwdOrTus5dffpmRI0cyYsQIVq5cyapVq5qM6f33\n3+eSSy4hNTWVtLQ0vvOd7/Dee+8B0LdvX4YPHw7AqFGjKC4u9lTOxo45ZMgQ3n77be68807ee+89\nMjMzyczMJDk5meuuu47XXnuNlJQUT+doidipKYBzCylCRlg1Jlw09YveT5MmTeLWW29l+fLlHDx4\nkFGjRgHw/PPPs2vXLpYtW0ZiYiK5ubmt6pe/fv16HnzwQT766COys7O55ppr2tS/v0OHDnXv4+Pj\nW3T7qCGnnHIKy5cvZ+7cudx9992MGzeOW2+9lSVLlrBgwQJmzZrFI488wj/+8Y82nae+2KkpgNPY\nvH49tOEv3hjTPtLS0hg7dizf+973jmlgLikpoVu3biQmJrJw4UI2bGh6ROizzz6bF154AYDPP/+c\nFStWAFBaWkpqaiqZmZns2LGDt956q+476enplJWVHXess846i9dff52DBw9y4MABZs+ezVlnndWm\ncjZ2zK1bt5KSksLUqVO54447WL58OeXl5ZSUlHDBBRfw0EMP8emnn7bp3A2JvZqCKqxbB4OPG47J\nGBNmJk+ezCWXXHJMT6QpU6Zw0UUXMWTIEAoKCsiv7VnYiBtvvJFrr72WAQMGMGDAgLoax7Bhwxgx\nYgT5+fn07t2bM844o+4706ZNY+LEifTo0YOFCxfWbR85ciTXXHMNo0ePBuD6669nxIgRnm8VAfz6\n17+ua0wG2Lx5c4PHnDdvHnfccQdxcXEkJiby2GOPUV5ezpQpU6ioqEBVmT59uufzeqaqEbWMGjVK\nW23ZMlVQnTWr9cdogYULF7bLeUIhmsumGt3l81K2VatW+R+ID0pLS0Mdgq+8lq+hvz9gqXq4xsbW\n7aNTTnFerbHZGGMaFFtJIS0NevWybqnGGNOI2EoK4LQrWE3BGGMaFHtJIT/fqSnYwHjGGHOc2EsK\neXlQWgrbt4c6EmOMCTuxlxRsYDxjjGlU7CUFGxjPmLC3Z88ehg8fzvDhwznhhBPo2bNn3XrtIHnN\nufbaa1nTgh9/Tz75JLfccktrQ44asfXwGji9j1JSrKZgTBjr3Lkzn7ij8N1zzz2kpaUdNyJoXb/6\nuIZ/2z7zzDO+xxmNYq+mYAPjGROx1q1bx8CBA5kyZQqDBg1i27ZtTJs2rW746/vuu69u3zPPPJNP\nPvmEqqoqsrKyuOuuuxg2bBinn346O3fu9HzO5557jiFDhjB48GB++tOfAlBVVcXVV19dt/3hhx8G\n4KGHHqob1trLCKnhKPZqCuC0KyxeHOoojIkMoRg7uwmrV6/mL3/5CwUFBQDcd999dOrUiaqqKs4+\n+2xWrVrFwIEDj/lOSUkJ55xzDvfddx+33XYbTz/9NHfddVez59q8eTN33303S5cuJTMzk/Hjx/PG\nG2/QtWtXdu/ezWeffQZQN4T1Aw88wIYNG0hKSvJlWOv2EHs1BXDaFYqLoY2jGBpj2t/JJ59clxAA\nXnzxRUaOHMnIkSNZs2ZNg8Nfd+zYkW9+85tAy4a1Xrx4Meeeey5dunQhMTGRq666ikWLFtGvXz/W\nrFnDTTfdxLx588jMzARg0KBBTJ06leeff57ExMS2FzYEYremUDsw3pAhoY7GmPAWqrGzG5Gamlr3\nfu3atfzud79jyZIlZGVlccUVVzQ4/HVSUlLd+/j4eKqqqtoUQ+fOnVmxYgVvvfUWjz76KK+++ipP\nPPEE8+bN491332XOnDn89re/ZcWKFcTHx7fpXO0tdmsKYO0KxkS40tJS0tPTycjIYNu2bSxYsCCo\nxx8zZgwLFy5kz549VFVVMXPmTM455xx27dqFqnLZZZdx7733snz5cqqrq9m8eTPnnnsuDzzwALt3\n7z5mnudIEZs1hdqB8awHkjERbeTIkQwcOJD8/Hz69OnDaaed1qbjPfXUU8yaNatufenSpfzqV7+i\nsLAQVeWiiy7iwgsvZPny5Vx33XWoKiLC/fffT1VVFVdddRVlZWXU1NRw++23k56e3tYitjvRCBvu\noaCgQJcGY/a0E0+Es8+G555r+7EaUVRURGFhoW/HD6VoLhtEd/m8lO2LL75gwIAB7RNQEJWVlUXk\nhdgrr+Vr6O9PRJapakEjX6kTm7eP4OgYSMYYY+rEblLIy7OB8Ywxpp5mk4KIXCYi6e77u0XkNREZ\n6X9oPsvLg7Iy2LYt1JEYE5Yi7daycbT1781LTeFnqlomImcC44GngMfadNZwYAPjGdOo5ORk9uzZ\nY4khwqgqe/bsITk5udXH8NL7qNp9vRB4QlXfFJFft/qM4SKwW+rYsaGNxZgw06tXLzZv3syuXbtC\nHUqLVFRUtOmCGO68lC85OZlevXq1+hxeksIWEXkcOA+4X0Q6EA1tET17Qmqq1RSMaUBiYiJ9+/YN\ndRgtVlRUxIgRI0Idhm/ao3xeLu6XA/OA81V1P9AJuMPXqNqDDYxnjDHH8ZIUugNvqupaESkELgOW\n+BpVe7FuqcYYcwwvSeFVoFpE+gFPAL2BF3yNqr3k5cGGDTYwnjHGuLwkhRpVrQK+A/xeVe/AqT1E\nvtqB8dauDXUkxhgTFrwkhUoRmQx8F3jD3eZpTFgRmSgia0RknYgcN3i5iJwoIgtF5GMRWSEiF3gP\nPQhsYDxjjDmGl6RwLXA68BtVXS8ifYG/NvclEYkHHgW+CQwEJovIwHq73Q28rKojgCuBP7Qk+Daz\ngfGMMeYYzSYFVV0F3A58JiKDgc2qer+HY48G1qnq16p6BJgJTKp/eCDDfZ8JbPUceTCkpDgD41lN\nwRhjAA+jpLo9jv4MFAOC09D876q6qJnvXQpMVNXr3fWrgTGq+sOAfboD84FsIBUYr6rLGjjWNGAa\nQE5OzqiZM2d6LF7zht5xB4mlpSx7/PGgHbNWeXk5aWlpQT9uOIjmskF0l8/KFrnaUr6xY8d6GiUV\nVW1yAZYBeQHrpwDLPHzvUuDJgPWrgUfq7XMb8GP3/enAKiCuqeOOGjVKg+pHP1JNS1OtqQnucVV1\n4cKFQT9muIjmsqlGd/msbJGrLeUDlmoz121V9dSmkKiqdTfdVfVLvDU0b8GpVdTq5W4LdB3wsnvc\nD4FkoIuHYwdPfj6Ul8PW9r1zZYwx4chLUlgqIk+KSKG7/AnwMsvNR0B/EekrIkk4Dclz6u2zERgH\nICIDcJJC+w62UtsDyRqbjTHGU1K4Eee2zk3usgr4fnNfUufZhh/iDJHxBU4vo5Uicq+IXOzu9mPg\nBhH5FHgRuMat5rSf2tFSrbHZGGOaHxBPVQ8D090FABF5CbjCw3fnAnPrbft5wPtVwBktiDf4evSA\ntDSrKRhjDK0f7fT0oEYRSiLOLSSrKRhjTBQMgR0MtVNzGmNMjGv09lETU24KHoe5iBh5efDCC3Dw\noPNAmzHGxKim2hT+t4nPouteS21j89q1MGxYaGMxxpgQajQpqGrszFEZODCeJQVjTAyzNgWA/v2d\nBmdrVzDGxDhLCmAD4xljjMuSQi2bmtMYY5pPCiLymohcKCLRnUBqu6W28wPVxhgTTrxc6P8AXAWs\nFZH7RCTP55hCIz8fDhyALfXH7DPGmNjhZZKdd1R1CjASZ06Fd0TknyJyrYhEz/MKNjWnMcZ4a1MQ\nkc7ANcD1wMfA73CSxNu+Rdbeap9VsHYFY0wMa3ZAPBGZDeThzMt8kapucz96SUS8DKEdGbp3dwbG\ns5qCMSaGNZsUgIdVdWFDH6iXqd0ihYj1QDLGxDwvSeFDEbkNOBNQ4H3gMVWt8DWyUMjLg/feC3UU\nxhgTMl7aFP4CDAJ+DzwCDMS5lRR98vNh40anF5IxxsQgLzWFwao6MGB9oYis8iugkKrtgbR2LQwf\nHtpYjDEmBLzUFJaLyGm1KyIyBm9zNEcem5rTGBPjvNQURgH/FJGN7vqJwBoR+QxQVR3qW3TtrV8/\nGxjPGBPTvCSFib5HES46doTcXKspGGNiVrNJQVU3iMgw4Cx303uq+qm/YYWQTc1pjIlhXgbEuxl4\nHujmLs+JyI/8Dixkap9VqKkJdSTGGNPuvNw+ug4Yo6oHAETkfuBDnC6q0Scvz5mrecsW6N071NEY\nY0y78tL7SIDqgPVqd1t0soHxjDExzEtN4RlgsTsGEsC3gaf8CynEAgfGO++80MZijDHtzEtD83QR\nKcIZ5gLgWlX92NeoQumEEyA93WoKxpiY1GRSEJF4YKWq5gPL2yekELOB8YwxMazJNgVVrcZ5UO3E\ndoonPOTlWU3BGBOTvLQpZAMrRWQJUDdSnKpe7FtUoZafD889B+XlzhwLxhgTI7wkhZ/5HkW4qe2B\n9OWXMHJkaGMxxph25KVL6gWq+m7gAlzgd2AhZVNzGmNilJek0FC/zG8GO5CwYgPjGWNiVKNJQURu\ndEdCzRORFQHLeuAzLwcXkYkiskZE1onIXY3sc7mIrBKRlSLyQuuKEWTJydC3rzU2G2NiTlNtCi8A\nbwH/DQRe0MtUdW9zB3a7sz6KU9PYDHwkInNUdVXAPv2BnwBnqOo+EenWijL4wwbGM8bEoEZrCqpa\noqrFqjoZ56JeiTNHc5rHLqqjgXWq+rWqHgFmApPq7XMD8Kiq7nPPubM1hfCFDYxnjIlBzfY+EpEf\nAvcAO4DaK6QCzU2u0xPYFLC+GRhTb59T3HN8AMQD96jq3xuIYRowDSAnJ4eioqLmwm6z7kDeoUN8\n+MorHM7JadUxysvL2yXWUIjmskF0l8/KFrnao3xeuqTeAuSp6h6fzt8fKAR6AYtEZIiq7g/cSVWf\nAJ4AKCgo0MLCQh9CqUcEHnqI07OzoZXnKyoqol1iDYFoLhtEd/msbJGrPcrnpffRJqCkFcfeAgSO\nPd3L3RZoMzBHVStVdT3wJU6SCL3aZxWsXcEYE0O81BS+BopE5E3gcO1GVZ3ezPc+AvqLSF+cZHAl\ncFW9fV4HJgPPiEgXnNtJX3uM3V85OZCZaT2QjDExxUtS2OguSe7iiapWue0R83DaC55W1ZUici+w\nVFXnuJ9NEJFVOPM03OHTbaqWE7EeSMaYmONl6OxfAohIiqoebMnBVXUuMLfetp8HvFfgNncJP/n5\nsGBBqKMwxph242WO5tPdX/Kr3fVhIvIH3yMLB3l5zrScZWWhjsQYY9qFl4bmGcD5wB4AVf0UONvP\noMJG7RhIX34Z2jiMMaadeEkKqOqmepuqG9wx2lgPJGNMjPHS0LxJRL4BqIgkAjcDX/gbVpjo1w/i\n4qwHkjEmZnipKXwf+AHOE8pbgOHuevTr0MEZGM9qCsaYGOGl99FuYEo7xBKebGpOY0wM8dL76AER\nyRCRRBFZICK7RGRqewQXFvLzYe1aGxjPGBMTvNw+mqCqpcC3gGKgH3CHn0GFlbw8OHQINtVvazfG\nmOjjJSnU3mK6EHhFVVszDlLkqu2WareQjDExwEtSeENEVgOjgAUi0hWo8DesMGLdUo0xMaTZpKCq\ndwHfAApUtRI4wPGT5USvbt0gK8tqCsaYmOClofkyoFJVq0XkbuA5oIfvkYULGxjPGBNDvNw++pmq\nlonImcB44CngMX/DCjP5+VZTMMbEBC9JoXZIiwuBJ1T1TVowhHZUyMuDrVttYDxjTNTzkhS2iMjj\nwBXAXBHp4PF70aO2B5LdQjLGRDkvF/fLcSbDOd+dO7kTsfScAlgPJGNMzPDS++gg8BVwvjuTWjdV\nne97ZOHk5JMhPt7aFYwxUc9L76ObgeeBbu7ynIj8yO/AwooNjGeMiRFehs6+DhijqgcAROR+4EPg\n934GFnasB5IxJgZ4aVMQjp1Up9rdFlvy8pyB8apjY34hY0xs8lJTeAZYLCKz3fVv4zyrEFvy86Gi\nAjZudG4lGWNMFPIyn8J0ESkCznQ3XauqH/saVTgK7IFkScEYE6WaTAoiEg+sVNV8YHn7hBSmapPC\n6tUwcWJoYzHGGJ802aagqtXAGhE5sZ3iCV9du0J2tvVAMsZENS9tCtnAShFZgjNCKgCqerFvUYUj\nGxjPGBMDvCSFn/keRaTIz4d580IdhTHG+KbRpCAi/YAcVX233vYzgW1+BxaW8vLg2WehtBQyMkId\njTHGBF1TbQozgNIGtpe4n8UeGxjPGBPlmkoKOar6Wf2N7rZc3yIKZzYwnjEmyjWVFLKa+KxjsAOJ\nCDYwnjEmyjWVFJaKyA31N4rI9cAy/0IKY0lJcNJJVlMwxkStpnof3QLMFpEpHE0CBTizrl3id2Bh\nywbGM8ZEsUZrCqq6Q1W/AfwSKHaXX6rq6aq63cvBRWSiiKwRkXUiclcT+/2biKiIFLQs/BCwgfGM\nMVHMy9hHC4GFLT2wO0TGo8B5wGbgIxGZo6qr6u2XDtwMLG7pOUIiPx8OH4YNG5xbScYYE0X8nGt5\nNLBOVb9W1SPATGBSA/v9CrgfqPAxluCxHkjGmCjm5Ynm1uoJbApY3wyMCdxBREYCvVX1TRFpdN5n\nEZkGTAPIycmhqKgo+NF6lFhSwhnAujfeYHPHpjthlZeXhzRWP0Vz2SC6y2dli1ztUT4/k0KTRCQO\nmA5c09y+qvoE8ARAQUGBFhYW+hpbszp1ol91Nf2aiaOoqIiQx+qTaC4bRHf5rGyRqz3K19QwF2WA\nNvQRoKra3DgPW4DeAeu93G210oHBQJGIAJwAzBGRi1V1qYfYQ8d6IBljolSjSUFV09t47I+A/iLS\nFycZXAlcFXD8EqBL7bo7kc/tYZ8QwGlXeOutUEdhjDFB57mhWUS6iciJtUtz+6tqFfBDYB7wBfCy\nqq4UkXtFJLKH3c7Ph+3boaQk1JEYY0xQNdum4F7A/xfoAewE+uBc5Ac1911VnQvMrbft543sW9h8\nuGEisAfS6NGhjcUYY4LIS03hV8BpwJeq2hcYB/zL16jCnY2WaoyJUl6SQqWq7gHiRCTOfZgt/J88\n9tNJJ0FCgjU2G2OijpcuqftFJA1YBDwvIjsJmJYzJiUm2sB4xpio5KWmMAk4CNwK/B34CrjIz6Ai\ngnVLNcZEIS9JoRuQpKpVqvpn4E84zxjENhsYzxgThbwkhVeAmoD1andbbMvPhyNHoLg41JEYY0zQ\neEkKCe6AdgC475P8CylC2MB4xpgo5CUp7Ap82ExEJgG7/QspQtR2S7V2BWNMFPHS++j7OL2OHsEZ\n92gT8F1fo4oEnTs7i9UUjDFRxMskO18Bp7ndUlHVct+jihTWA8kYE2WaGiV1qqo+JyK31dsOgKpO\n9zm28JeXB2++GeoojDEmaJpqU0h1X9MbWUx+PuzYAfv3hzoSY4wJiqaGzn7cnWe5VFUfaseYIkdg\nD6QxY5re1xhjIkCTvY9UtRqY3E6xRB7rgWSMiTJeeh994PY8eomAMY9UdblvUUWKvn2dgfGsB5Ix\nJkp4SQrD3dd7A7YpcG7ww4kwiYnQr58lBWNM1PDSJXVsewQSsfLy7PaRMSZqNPtEs4hkish0EVnq\nLv8rIpntEVxEyM+HdeugqirUkRhjTJt5GebiaaAMuNxdSoFn/AwqouTl2cB4xpio4aVN4WRV/beA\n9V+KyCd+BRRxAqfm7NcvtLEYY0wbeakpHBKRM2tXROQM4JB/IUWY2mcVrF3BGBMFvNQUbgT+7LYj\nCLAXuMbPoCJKp07Qtav1QDLGRAUvvY8+AYaJSIa7Xup7VJHGeiAZY6JEs0mhkQHxSoBlbsIweXnw\nf/8X6iiMMabNvLQpFODMqdDTXf4DmAj8SUT+08fYIkd+PuzcCfv2hToSY4xpEy9JoRcwUlV/rKo/\nBkYB3YCzsbYFh03NaYyJEl6SQjfgcMB6JZCjqofqbY9dNjCeMSZKeOl99DywWET+5q5fBLwgIqnA\nKt8iiyR9+zrjIFlNwRgT4bz0PvqViLwFnOFu+r6qLnXfT/EtskiSkOA8uGY1BWNMhPNy+wggGWey\nnd8BG0Skr48xRaa8PKspGGMinpcB8X4B3An8xN2UCDznZ1ARyQbGM8ZEAS81hUuAi3En2FHVrdgc\nzcfLy4PKSli/PtSRGGNMq3mbsd7kAAATVUlEQVRJCkdUVXEm1sFtYPZERCaKyBoRWScidzXw+W0i\nskpEVojIAhHp4z30MBM4MJ4xxkQoL0nhZRF5HMgSkRuAd4Anm/uSiMQDjwLfBAYCk0VkYL3dPgYK\nVHUoMAt4oCXBhxUbGM8YEwW89D56UETOw5lHIQ/4uaq+7eHYo4F1qvo1gIjMBCYR0I1VVRcG7P8v\nYGoLYg8v2dnQrZvVFIwxEc3L2Ef3q+qdwNsNbGtKT2BTwPpmYEwT+18HvNVIDNOAaQA5OTkUFRU1\nF3ZIDD/hBFi8mE/c+MrLy8M21raK5rJBdJfPyha52qV8qtrkAixvYNsKD9+7FHgyYP1q4JFG9p2K\nU1Po0NxxR40apWHrhhtUu3atW124cGHoYvFZNJdNNbrLZ2WLXG0pH7BUm7m+qmrjNQURuRH4f8BJ\nIrIi4KN04AMP+WYL0DtgvZe7rf55xgP/BZyjqpE9bEZeHuzaBXv3OvMsGGNMhGnq9tELOLdz/hsI\n7DlUpqp7PRz7I6C/+6DbFuBK4KrAHURkBPA4MFFVd7Yk8LAU2APp9NNDG4sxxrRCo72PVLVEVYtV\ndbKqbsCZglOBNBE5sbkDq2oV8ENgHvAF8LKqrhSRe0XkYne3/wHSgFdE5BMRmdPWAoWU9UAyxkQ4\nLw3NFwHTgR7ATqAPzkV+UHPfVdW5wNx6234e8H58C+MNb7m5kJRkPZCMMRHLy3MKvwZOA75U1b7A\nOJxGYVOfDYxnjIlwXpJCparuAeJEJE6dZwsKfI4rcuXnW03BGBOxvMynsF9E0oBFwPMishN3HCTT\ngLw8mDPHGQfJGGMijJeawiTgIHAr8HfgK5yJdkxD8vKckVJtYDxjTARqNCmISD8ROUNVD6hqjapW\nqeqfgeVAVvuFGGFsak5jTARrqqYwA2e8o/pK3M9MQ2q7pVq7gjEmAjWVFHJU9bP6G91tub5FFOmy\nsiAnx2oKxpiI1FRSaOoWUcdgB+K3rVvhs8+gurodTmZTcxpjIlRTSWGpO3/CMUTkemCZfyH5469/\nhaFDITMTCgvhzjvh1Vdh82ZwxuQLIuuWaoyJUE11Sb0FmC0iUziaBAqAJJwpOiPKVVdBjx6wZAks\nXgwPPXS012j37jBmDIwe7bwWFEBGRhtOlpcHu3eTUFISlNiNMaa9NJoUVHUH8A0RGQsMdje/qar/\naJfIgqx3b7j6amcBOHwYPvnkaJJYsgRef935TAQGDDiaJEaPhiFDIDHR48ncHkgpmzY1s6MxxoQX\nLzOvLQQWNrdfpOnQwbngjxkDP/qRs23vXvjoo6NJ4o034Nlnnc+Sk2HkyGNrFLm5TgI5jtsDKWXj\nxvYoijHGBI2XJ5pjRqdOcP75zgJOW0Nx8bG1iccec249AXTt6iSI2iRx6qnuNAruwHhWUzDGRBpL\nCk0Qgb59neWKK5xtlZXw+edHk8TixTB37tHG6v79YfToeGZ0yeOE2XOo+s5lJHxzAkyYAH36hK4w\nxhjjgSWFFkpMhBEjnOX733e2lZbC0qVHk8Q//gEXbXuc63mSCbPn03v2LAD2dj2Fg2dMIOvyCaR9\nqxDS00NXEGOMaYAlhSDIyIBzz3WWWps3n86zz6bxVOWT7Fy0mq4fz2f0rvkUvv40qa8/QiUJfNnl\nG+waPoGkb03gpEtHckLP+NAVwhhjsKTgm1694Mwz91BYKMAAYAA7d97MB4sPs3vOP+n4/nz6r59P\n4Tt3wzt3s+eWTsxJHs/6/hM4UjiBfmN7M2KEc8epwcZsY4zxgSWFdtStG0y4qANcNBYYC/w3pet2\nsuUvC9C/z+Osz+dz8Wcvw2fwxe/z+RsT+CDtfA6MOoeBp6YyYoTTA6p/f4i3SoUxxgeWFEIso183\nMu6dDPdOdlqrV66k8s35dH99Pj9Y9gQ3lz9M5aJE3l90JvN0Ag8ygS87DmfYiLi6JDFiBAwa5MwE\naowxbWFJIZyIwODBJA4eTNadt0FFBbz/Ponz51M4bz5jV/yE+/gJ5dqFD9eex6vLJvCzw+exlZ4k\nJsLAgc5DdkOHOsuQIc7T2nb7yRjjlSWFcJacDOPHw/jxyAMPwPbt8M47pM2fz3nz53Pe4Rf5I1DS\naxArcs5jWcUg/vVWLo8/l8smenOEDnTufHyiGDQIUlNDXThjTDiypBBJTjgBpk51FlVn2Nf588mc\nP5+zFj3GWYcP1+2qIhzI6MG2hFy++iyXT9/P5V9VfZhJLhvIJfHkE8kf1qEuUQwdCiedBHFe5uIz\nxkQtSwqRSuToz//bb3emAN2yxXkEu7gYKS4mrbiY/sXF9C/+gPP3z0QIGDf8K9hR3IN1r+WygT68\nTC5bk3KJ65tLxtBcepx2IgNHJjNkCHTuHKpCGmPamyWFaJGQ4PRf7dMHzjnnuI+lqsqZVMJNGhQX\nk1NcTJevihm17l8k7XiFuCNVsAZneQW20p0v6cOOjrkcPiGXxP65ZA7LpdeZuVRT4dRWrMGiTWpq\nnMEZO0bcDCUmWllSiBUJCXDiic5y9tl1m+PdherquqSh64sp+6yYmhXF9Pi6mJO2L6HT+lkkrq+C\n+cD/QB5wmCRK4rIpjc/mQFI2Bztkczg5iyOp2VSmZ1OTkY1mZSOdsonvkk1i1yyScrJJ7p5Nak4a\nGZlCRobzYHe0drEtKYGNG2HTJmcJfF+7HDnijJlV+9dz4onOqL6B6927R++fkQkvlhSMIz7euRL1\n7o2cdRYZwDFTSlRXU7lhK5ve38COxcVsWbyCzvFCQuk+kg7so8OhfXSq2E5K+WrSt+8jo2Y/cTQ+\ne1ElCewni51ks4ZsyuKyKE/M5mByNhXJ2U5iScumJjMbzcwivlMmSZ3TSe6aTsdu6aR2SyWrUxyZ\nmc4MqJmZkJbWvhWXiopjL+4NXfzLyo79Tnw89Ozp/FGfeip85ztO7Fu3Ot8pLoZFi2D//mO/l5Dg\nfC8wUdRf2jQHiDEuSwrGm/h4Ek/qzUkn9eak755JUVERhYWFje9fUwOlpVTu3MfBLfuo2L6fim37\nqNy5j6rd+9A9+2DfPuJK99O5bB/dy/fR4dB6OlbsI7V0H/Ha9LypNQgHSKWMdMpIZ7v7WpGYzpGk\ndI4kZ1DVMR1Nc5a4TGdJyE6nQxdnSclxlvQe6WR07UBmltChg3P86mpp8mK/aRPs2nV8XN26ORf8\nvDyn45ibZ+t+/Xv9xV9aevR89ZcPPoCXXnKakQJlZjadNHr0cJJLc2pqnIEfKyudWozX14a2VVU5\nNcFOnZy2qdrXzEyr+dRXXQ3l5c7ffVmZ81p/SU9Ppan/dsFgScH4Iy4OsrJIzMoi85S+ZLbku6pw\n4ADs23d0KS3lyO5SKnaVUbG7jMo9ZVTtK6O6pIyE0jI6lZfR5UAZCYc2klhRRoeyMjruK6NDTYWn\nU1aSQBnp7CCdA/HppFWnso4UDpJCPCnk0JHOiSkUZKSQlJVCcm4KKaemkN4thczuKWT3TKFzb+cz\nUhpYOnZs0VUwI8PpOjxoUMOfV1c7PZQbShobN8KHHzrzgwSKi3NqG3Fxp5KY2PjFvT3mMReB7Gwn\nSdRPGI1t69zZ+XMJpx5yqk6bUO1Fu7GLuZft5eXNn++mm1r0P6lVLCmY8CPi3AtKS3N+XruS3KVF\nd0kqK53/bWVl1JSUcWhnGQe2O68Vu8qo3OssNSVlaEkplJcRf7CM5MN76Z1aQwo7Sa45SFLlQeIq\nDsLBg7D+IHxV0/JydejgJIeGEkZCQuNLYuJx2+ITEujpLqfXbj8xAU46uk9FVQJ7SxPYvd9Zdu5N\nYMeeBLbuOURGTndqOnREk52Fjs6rpDjvk5LjSEx0npJvy2t8vHPh27sX9uxxXgPf177u3AmrVzvv\nm5rFNi7OSSaNJZEdO3qxdKlTQ6mt7dQuXre1ZN9Dh45O69uU+HgnodW2oWVkOPH27Xvstvr71N/2\n8cfbgFNa/m+vBSwpmOiWmOhcRbKziQNS3aU5RUVFDGusnq7q/KQ+eLDh5dChxj9rbP+KiqNXnaqq\nli2NSAZ6uMtxPm/mDyApyUlWQVi6JyUdzRK9EqFv4rGZI3BJSqKSREoOJrKn1Fn2lsQfl0RqX7dt\nc+Y32bu3tv2m33FFqc2rDS2NfZaS0vTntUtysnMrrKkLeUaG80cRjPau+PjG2+mCxZKCMS0l4vzq\n79DBSTihpOo0AjSVNAITTWUlyz78kFEDBzrJqDVLWZnz0742oQUu2vaLViLQxV0Ap3pQL3Ecs56W\nCJ2SqElIZF/5ITK7dEaSEo8uTV3V27rU1tLi4xt+XxUPpQlwoJHPw+lemMuSgjGRTMS5yMTHU9dK\n3oyykhJ8aa2srUHVTxSBjRaBSzC3HTlCXGUlcVWVJFANBypgfwP3hRpbgpDMWq2xpNLAtq6XXebP\n311gOH4eXEQmAr/D6Qr/pKreV+/zDsBfgFHAHuAKVS32MyZjjE8Ca1BZWSEJ4dPmesU1prraewKp\n39BQXX3sa2Pvg/B5VTv0O/YtKYhIPPAocB6wGfhIROao6qqA3a4D9qlqPxG5ErgfuMKvmIwxpkG1\nta3k5FBH0qR9RUW+n8PPG1qjgXWq+rWqHgFmApPq7TMJ+LP7fhYwTsTGTTDGmFDx8/ZRT2BTwPpm\nYExj+6hqlYiUAJ2B3YE7icg0YBpATk4ORe2QLYOhvLw8YmJtqWguG0R3+axskas9yhcRDc2q+gTw\nBEBBQYG26p5hCDT71G8Ei+ayQXSXz8oWudqjfH7ePtoC9A5Y7+Vua3AfEUkAMnEanI0xxoSAn0nh\nI6C/iPQVkSTgSmBOvX3mAP/uvr8U+IdqKPuGGWNMbPPt9pHbRvBDYB5Ol9SnVXWliNwLLFXVOcBT\nwF9FZB2wFydxGGOMCRFf2xRUdS4wt962nwe8rwAu8zMGY4wx3oXfM9bGGGNCRiLtFr6I7AI2hDoO\nj7pQr3ttFInmskF0l8/KFrnaUr4+qtq1uZ0iLilEEhFZqqoFoY7DD9FcNoju8lnZIld7lM9uHxlj\njKljScEYY0wdSwr+eiLUAfgomssG0V0+K1vk8r181qZgjDGmjtUUjDHG1LGkYIwxpo4lBR+ISG8R\nWSgiq0RkpYjcHOqYgk1E4kXkYxF5I9SxBJOIZInILBFZLSJfiMjpoY4pWETkVvff4+ci8qKIhPeM\nMs0QkadFZKeIfB6wrZOIvC0ia93XEE+i3TqNlO1/3H+XK0Rktoj4Mr2dJQV/VAE/VtWBwGnAD0Rk\nYIhjCrabgS9CHYQPfgf8XVXzgWFESRlFpCdwE1CgqoNxxiOL9LHGngUm1tt2F7BAVfsDC9z1SPQs\nx5ftbWCwqg4FvgR+4seJLSn4QFW3qepy930ZzoWlZ2ijCh4R6QVcCDwZ6liCSUQygbNxBmpEVY+o\n6v7QRhVUCUBHd5j6FGBriONpE1VdhDOQZqDA2Rz/DHy7XYMKkobKpqrzVbXKXf0XznQEQWdJwWci\nkguMABaHNpKgmgH8J1AT6kCCrC+wC3jGvTX2pIikhjqoYFDVLcCDwEZgG1CiqvNDG5UvclR1m/t+\nO5ATymB89D3gLT8ObEnBRyKSBrwK3KKqpaGOJxhE5FvATlVdFupYfJAAjAQeU9URwAEi9/bDMdx7\n65NwEl8PIFVEpoY2Kn+5c7NEXZ97EfkvnFvUz/txfEsKPhGRRJyE8LyqvhbqeILoDOBiESkGZgLn\nishzoQ0paDYDm1W1tlY3CydJRIPxwHpV3aWqlcBrwDdCHJMfdohIdwD3dWeI4wkqEbkG+BYwxa8J\nySwp+EBEBOe+9BeqOj3U8QSTqv5EVXupai5OQ+U/VDUqfnGq6nZgk4jkuZvGAatCGFIwbQROE5EU\n99/nOKKkEb2ewNkc/x34WwhjCSoRmYhz2/ZiVT3o13ksKfjjDOBqnF/Rn7jLBaEOynjyI+B5EVkB\nDAd+G+J4gsKt/cwClgOf4fzfj+ghIUTkReBDIE9ENovIdcB9wHkishandnRfKGNsrUbK9giQDrzt\nXlP+6Mu5bZgLY4wxtaymYIwxpo4lBWOMMXUsKRhjjKljScEYY0wdSwrGGGPqWFIwph2JSGG0jSxr\nooslBWOMMXUsKRjTABGZKiJL3IeEHnfnjygXkYfcOQkWiEhXd9/hIvKvgHHus93t/UTkHRH5VESW\ni8jJ7uHTAuZseN59wtiYsGBJwZh6RGQAcAVwhqoOB6qBKUAqsFRVBwHvAr9wv/IX4E53nPvPArY/\nDzyqqsNwxhmqHb1zBHALMBA4CecJeGPCQkKoAzAmDI0DRgEfuT/iO+IMrFYDvOTu8xzwmjsHQ5aq\nvutu/zPwioikAz1VdTaAqlYAuMdboqqb3fVPgFzgff+LZUzzLCkYczwB/qyqx8xsJSI/q7dfa8eI\nORzwvhr7f2jCiN0+MuZ4C4BLRaQb1M372wfn/8ul7j5XAe+ragmwT0TOcrdfDbzrzri3WUS+7R6j\ng4iktGspjGkF+4ViTD2qukpE7gbmi0gcUAn8AGfSndHuZztx2h3AGaL5j+5F/2vgWnf71cDjInKv\ne4zL2rEYxrSKjZJqjEciUq6qaaGOwxg/2e0jY4wxdaymYIwxpo7VFIwxxtSxpGCMMaaOJQVjjDF1\nLCkYY4ypY0nBGGNMnf8PZt0BHuZvpUYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ZGE-nD-tXfxl",
        "outputId": "b639558b-87ba-4222-8b60-c05daabc8145",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 386
        }
      },
      "source": [
        "from prettytable import PrettyTable    \n",
        "x = PrettyTable()\n",
        "x.field_names = [\"No of layer\",\"With/Without Dropout/BN\",\"activation function\", \"Accuracy %\"]\n",
        "\n",
        "\n",
        "x.add_row([\"3\", \"BN+DROPOUT\"   , \"RELU\"     ,\"98.6\"])\n",
        "x.add_row([\"3\", \"DROPOUT\"   , \"RELU\"     ,\"98.47\"])\n",
        "x.add_row([\"3\", \"BN\"   , \"RELU\"     ,\"98.48\"])\n",
        "\n",
        "x.add_row([\"3\", \"BN+DROPOUT\"   , \"SIGMOID\"     ,\"98.53\"])\n",
        "x.add_row([\"3\", \"DROPOUT\"   , \"SIGMOID\"     ,\"97.79\"])\n",
        "x.add_row([\"3\", \"BN\"   , \"SIGMOID\"     ,\"97.88\"])\n",
        "\n",
        "\n",
        "\n",
        "x.add_row([\"5\", \"BN+DROPOUT\"   , \"RELU\"     ,\"98.52\"])\n",
        "x.add_row([\"5\", \"DROPOUT\"   , \"RELU\"     ,\"98.32\"])\n",
        "x.add_row([\"5\", \"BN\"   , \"RELU\"     ,\"98.67\"])\n",
        "\n",
        "x.add_row([\"5\", \"BN+DROPOUT\"   , \"SIGMOID\"     ,\"98.73\"])\n",
        "x.add_row([\"5\", \"DROPOUT\"   , \"SIGMOID\"     ,\"97.73\"])\n",
        "x.add_row([\"5\", \"BN\"   , \"SIGMOID\"     ,\"97.79\"])\n",
        "\n",
        "\n",
        "\n",
        "x.add_row([\"6\", \"BN+DROPOUT\"   , \"RELU\"     ,\"99.28\"])\n",
        "x.add_row([\"6\", \"DROPOUT\"   , \"RELU\"     ,\"99.3\"])\n",
        "x.add_row([\"6\", \"BN\"   , \"RELU\"     ,\"98.88\"])\n",
        "\n",
        "x.add_row([\"6\", \"BN+DROPOUT\"   , \"SIGMOID\"     ,\"98.76\"])\n",
        "x.add_row([\"6\", \"DROPOUT\"   , \"SIGMOID\"     ,\"98.73\"])\n",
        "x.add_row([\"6\", \"BN\"   , \"SIGMOID\"     ,\"98.73\"])\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "print(x)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "+-------------+-------------------------+---------------------+------------+\n",
            "| No of layer | With/Without Dropout/BN | activation function | Accuracy % |\n",
            "+-------------+-------------------------+---------------------+------------+\n",
            "|      3      |        BN+DROPOUT       |         RELU        |    98.6    |\n",
            "|      3      |         DROPOUT         |         RELU        |   98.47    |\n",
            "|      3      |            BN           |         RELU        |   98.48    |\n",
            "|      3      |        BN+DROPOUT       |       SIGMOID       |   98.53    |\n",
            "|      3      |         DROPOUT         |       SIGMOID       |   97.79    |\n",
            "|      3      |            BN           |       SIGMOID       |   97.88    |\n",
            "|      5      |        BN+DROPOUT       |         RELU        |   98.52    |\n",
            "|      5      |         DROPOUT         |         RELU        |   98.32    |\n",
            "|      5      |            BN           |         RELU        |   98.67    |\n",
            "|      5      |        BN+DROPOUT       |       SIGMOID       |   98.73    |\n",
            "|      5      |         DROPOUT         |       SIGMOID       |   97.73    |\n",
            "|      5      |            BN           |       SIGMOID       |   97.79    |\n",
            "|      6      |        BN+DROPOUT       |         RELU        |   99.28    |\n",
            "|      6      |         DROPOUT         |         RELU        |    99.3    |\n",
            "|      6      |            BN           |         RELU        |   98.88    |\n",
            "|      6      |        BN+DROPOUT       |       SIGMOID       |   98.76    |\n",
            "|      6      |         DROPOUT         |       SIGMOID       |   98.73    |\n",
            "|      6      |            BN           |       SIGMOID       |   98.73    |\n",
            "+-------------+-------------------------+---------------------+------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "fZQRVQYEnP19"
      },
      "source": [
        "FROM ABOVE WE CAN CONCLUDE THAT RELU ACTIVATION FUNCTION  WORKS BETTER THAN SIGMOID."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "EjACmgsh8xW4"
      },
      "source": [
        "IN THIS ASSIGNMENT WE HAVE USED DIFFERENT ACTIVATION FUNCTION,NO OF LAYERS TO COMAPRE OUR PERFORMANCE."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "i1M6K4bQ9F20"
      },
      "source": [
        "Batch Normalization works more better than dropout for Relu Activation Function."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "ap4pCd3W9N2Q"
      },
      "source": [
        "AS LAYERS INCREASE PERFORMANCE ALSO INCREASES."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "WHL-eitx9Htm",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}