Extended_Deep_Q_Learning_for_Multilayer_Perceptron.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extended Deep Q-Learning for Multilayer Perceptron"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Basics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The Reinforcement Learning Framework:\n",
    "- reinforcement learning (RL) framework is characterized by an agent learning to interact with its environment\n",
    "- at each time step, the agent receives the environment's state \n",
    "- the agent must choose an appropriate action in response\n",
    "- at the next time step, the agent receives a reward and a new state\n",
    "- the agent has the goal to maximize the expected cumulative reward\n",
    "\n",
    "\n",
    "### Policy:\n",
    "- we can distinguish between deterministic and stochastic policy\n",
    "##### Deterministic policy:\n",
    "- a mapping from state to action\n",
    "    - $\\pi : S \\to A$\n",
    "    - for each state $s \\in S$ it yields the action $a \\in A$ that the agent will choose while in state $s$\n",
    "####  Stochastic policy:\n",
    "- a mapping from state to action probability \n",
    "    - $\\pi : S \\times A \\to [0,1]$\n",
    "    - for each state $s \\in S$ it and action $a \\in A$ it yields the probability $\\pi(a|s)$ that the agent chooses action $a$ while in state $s$\n",
    "\n",
    "### State-Value Functions:\n",
    "\n",
    "- state-value function for a policy $\\pi$ is denoted $v_\\pi$\n",
    "- $v_\\pi = \\mathbb{E_\\pi}[G_t | S_t = s]$\n",
    "    - for each state $s \\in S$, it yields the expected return if the agent starts in state $s$ and then uses the policy $\\pi$ to choose its actions $a$ for all remaining time steps\n",
    "    - $\\mathbb{E_\\pi}$ is the expected value of a random variable, given that the agent follows policy $\\pi$\n",
    "    \n",
    "### Action-Value Functions:\n",
    "\n",
    "- action-value function for a policy  $\\pi$ is denoted $q_\\pi$\n",
    "- $q_\\pi(s, a) = \\mathbb{E_\\pi}[G_t | S_t = s, A_t = a]$\n",
    "    - for each state $s \\in S$ and action $a \\in A$ tuple, it yields the expected return if the agent starts in state $s$, takes action $a$, and then follows the policy for all remaining time steps\n",
    "    - $\\mathbb{E_\\pi}$ is the expected value of a random variable, given that the agent follows policy $\\pi$\n",
    "    \n",
    "### Temporal Difference:\n",
    "- $TD_t(a_t, s_t) = r_t + \\gamma max_a(Q(a, s_{t+1})) - Q(a, s_t)$ <p>\n",
    "$r_t$: reward $r_t$ obtained by playing the action $a_t$ in the state $s_t$<br>\n",
    "$\\gamma$: discount factor(hyperparameter)<br>\n",
    "$max_a(Q(a, s_{t+1}))$: Q-value of the best action played in the state $s_{t+1}$<br>\n",
    "$Q(a, s_t)$: Q-value of the action $a_t$ played in the state $s_t$\n",
    "    \n",
    "- use the temporal difference $TD_t(a_t, s_t)$ to update $Q(a, s_t)$ \n",
    "    - $Q(a, s_t) = Q(a, s_t) + \\alpha * TD_t(a_t, s_t) $ \n",
    "    - the action value function is updated after every time step"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Deep Q-Learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![alt text](images/Generalization-of-Q-learning-with-neural-network.png \"Q_Learning_Network\")\n",
    "[Image Source](https://www.researchgate.net/profile/Jason_Gu2/publication/251898588/figure/fig1/AS:358420172558336@1462465436051/Generalization-of-Q-learning-with-neural-network.png)\n",
    "\n",
    "- in deep Q-learning, we use an artificial neural network as q value function\n",
    "    - the input to the network is the environment states\n",
    "    - the output of the network is a q-value for all possible actions\n",
    "    - we can now select the action based on this q-values\n",
    "        - for example by using an $\\epsilon$-greedy policy\n",
    "    - the network is updated by using the temporal difference method"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Implemented Extensions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fixed Q-Targets:\n",
    "- in vanilla Q-Learning, we are updating a guess for the q values with a guess\n",
    "- this can potentially lead to harmful correlations\n",
    "- we can break these correlations by using fixed Q-Targets meaning we are using two separated networks\n",
    "    - new updaterule: $\\Delta w = \\alpha * (r_t + \\gamma max_a(Q(a, s_{t+1}, w^-)) - Q(a, s_t, w)) \\nabla_w Q(a, s_t, w)$\n",
    "    - $w^-$ are the weights of a separate target network that are not changed during the learning step\n",
    "        - target Q-Network weights are updated less often(or more slowly) than the Q-Network\n",
    "        \n",
    "### Experience Replay:\n",
    "- deals with the problem that the sequence of experience tuples can be highly correlated\n",
    "- by learning from this experience tuples in sequential order, we run into the risk of getting swayed by the effects of this correlation\n",
    "- we can handle this problem by saving the last $m$ transitions\n",
    "    - $m$ is a large number\n",
    "    - a transition (or experience) is often defined as a tuple in the form of (state, action, reward, next_state, done)\n",
    "- the pack of the last transition is called experience replay(or replay buffer)\n",
    "- from this replay buffer, we take random batches of transitions to make updates to the network\n",
    "    - this randomness breaks the correlation\n",
    "    - this also allows us to learn more from individual tuples multiple times\n",
    "\n",
    "### Gradient Clipping:\n",
    "- deals with the problem of exploding gradients\n",
    "    - gradient gets exponentially large from being multiplied by numbers larger than 1\n",
    "- gradient clipping puts the gradients in a range between a min and max value to avoid this problem\n",
    "\n",
    "### Double Deep Q-Learning:\n",
    "- addresses the problem of overestimation of the Q-Values\n",
    "- the Q-Values depend a lot on what actions have been already tried and what states have been explored\n",
    "- that means that choosing the right max value can be error-prone especially in the early stages\n",
    "    - that is because the Q-values are still evolving and we might need more information to figure out the best action\n",
    "    - we are basically choosing the max value among a set of noisy numbers which leads to the overestimation\n",
    "- to solve this problem, we select the best action using one set of parameters $w$ but evaluate it using a different set of parameters $w^-$\n",
    "    - we can use our already introduced target Q-Network for this\n",
    "- that means we have two separate function approximators that must agree on the best action\n",
    "    - if $w$ picks an action that is not the best according to $w^-$ then the estimated Q-Value is not that high\n",
    "    \n",
    "\n",
    "### Dueling Networks:\n",
    "\n",
    "![alt text](images/dueling_dqn.png \"Dueling_Q_Learning_Network\")\n",
    "[Image Source](https://coach.nervanasys.com/algorithms/value_optimization/dueling_dqn/index.html)\n",
    "\n",
    "- the core idea of dueling networks is to use a specialized Dueling Q Head to separate $Q$ to an $A$ (advantage) stream and a $V$ (state value) stream\n",
    "    - the streams may share some layers at the beginning then branch off with their own fully connected layers\n",
    "    - this allows the network to differentiate actions from one another better\n",
    "- the reason for this is that the value of most states do not vary a lot across actions\n",
    "    - we can start learning the state-value even if only a single action has been taken at this state\n",
    "- this leads to faster learning, but we still need to capture the difference that actions make in each state\n",
    "    - the advantage stream realizes this"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import gym\n",
    "import json\n",
    "import os\n",
    "import helper\n",
    "from dqn_agent import Agent\n",
    "import sessions\n",
    "\n",
    "# ignore gymn warnings\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Environment: Lunar_Lander_v2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- LunarLander-v2 defines \"solving\" as getting average reward of 200 over 100 consecutive trials"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[33mWARN: gym.spaces.Box autodetected dtype as <class 'numpy.float32'>. Please provide explicit dtype.\u001b[0m\n",
      "Episode 100\tAverage Score: -214.30\n",
      "Episode 200\tAverage Score: -152.37\n",
      "Episode 300\tAverage Score: -96.496\n",
      "Episode 400\tAverage Score: -65.861\n",
      "Episode 500\tAverage Score: 84.633\n",
      "Episode 600\tAverage Score: 169.72\n",
      "Episode 700\tAverage Score: 192.23\n",
      "Episode 758\tAverage Score: 200.30\n",
      "LunarLander-v2 Environment solved in 658 episodes!\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAEKCAYAAAA8QgPpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJztnXeYG8X5x7+vdP3c7XPBvWGwKQaMC9VgA6YE00JLgADBJECAlF+oCS0kJCGQQAjgJPQeCCWhGkyPOxhjG4PPBWzce72q+f2xO6vRamZ3Je1KOvv9PM89J83O7o7avPPWISEEGIZhGCZTYoUeAMMwDNMyYQHCMAzDZAULEIZhGCYrWIAwDMMwWcEChGEYhskKFiAMwzBMVrAAYRiGYbKCBQjDMAyTFSxAGIZhmKwoKfQAoqRTp06iT58+hR4GwzBMi2LWrFnrhBA1fv0KJkCIqALABwDK7XE8L4S4iYj6AngGQAcAnwA4TwjRQETlAB4DcBCA9QDOEkIs9bpHnz59MHPmzAhfBcMwzK4HEX0dpF8hTVj1AI4WQuwPYCiAcUQ0EsDvAdwthBgIYCOAi+3+FwPYKIQYAOBuux/DMAxTIAomQITFNvtpqf0nABwN4Hm7/VEAp9iPx9vPYR8fQ0SUp+EyDMMwLgrqRCeiOBHNBrAGwCQAiwBsEkI02V2WA+huP+4OYBkA2Mc3A+ioueYEIppJRDPXrl0b9UtgGIbZbSmoABFCNAshhgLoAWA4gL113ez/Om0jrRa9EGKiEGKYEGJYTY2vD4hhGIbJkqII4xVCbALwHoCRANoRkXTu9wCwwn68HEBPALCPtwWwIb8jZRiGYSQFEyBEVENE7ezHlQDGAvgCwLsAzrC7XQDgZfvxK/Zz2McnC94Ni2EYpmAUMg+kG4BHiSgOS5A9J4T4LxHNB/AMEf0GwKcA/mn3/yeAx4moFpbmcXYhBs0wDMNYFEyACCHmADhA074Ylj/E3V4H4Lt5GBrDMEworN1aj1lfb8S4fboWeiiRUBQ+EIZhmF2RCx+Zjh89MQvb65v8O7dAWIAwDMNExPKNOwEADU2JAo8kGliAMAzDRETcznWuZwHCMAzDZEI8JgVIc87XWrpue9GZwliAMAzD5MCmHQ0wZRSU2AKkrjF3DWT0ne/h4kdn5HydMGEBwjC7IYmEwO/fWIA1W+sKPZQWzarNdRh66yTc//4i7fF4XAqQ3DSQpmZLAE1dXFy50yxAGGY3ZOqS9bj/vUW45vk5hR5KUfLWvFX41Utzffut21YPAHhl9grt8ZKYNcWaBEhzQmDD9gast69joqG5OH0oLEAYZjdEWlzCMK0UEiEEpi/ZYDQhZcuEx2fh8an+W2KUlVhT6E6DgEj6QPTv8x/eXIADb5uEg37zNt6Yu8p4HxnFVRYPNmU35kngsABhmN0QWZlUpNcjbVE8M2MZznxwCt6cZ558o0RO7Dsb9AIk6QNJPT55wWpMXrA6RWjMXJpunpLnSQFUGvffwWLNljoMvOF1PDkt0J5QOcEChGF2R+x5SC7c9/rV67j5lXmFG49CU3MCt/5nPlZu3okFq7agz7Wv4upnPtX2XbzW2lLomhc+z+getWu24b0v1wQaiw4ZVSUn9p2NzdhW35SmCUkNpM6lgVz0yExc9MjMFMHiFuXzVmzGXr96A2/PX416W1MsLUlO2VvqGrWaxrebrNyTZ2cs83t5OcMChGF2Q8iWIHLSqmtM4JH/LS3YeCSJhMC/Zi3HQx8vwS+fn4P73rWc0y/NXqGdzGP2BL15Z2NGjuqxd72PHzzsH9Gk8z28PPtbDLrxDSxau83RQLbWNWGfm97EnW99iVlfb8CgG1/HT5+dbdRAJCbNBQD+V7seAPDxonVoaLb6lSomrP1ufgtDb3kLH3yVuu9RY7PwvXZYsABhmN2QWNKGVVRc+cynuO7fljaxflsDGpWV+/b65IT43zkrsHlno5Oo5+apad9g6brt6HPtq7j/PUsINTUnMPK37+DFT5drz1mxaSfWbEmNSqvX+Ij+85nlMK9dsy1NA7jv3UU4/f4pqG9K4MVPv03zgTQ0JfDN+h1OfznZA5ZSmEgILFq7Dec/NB1TF1sCpE1FqeOrKovH8O6CNc41tjc04/yHpqM5ITBj6QbsbGjGmQ9OAQDUhZB74kchq/EyDFMg5G7QiTzviPDWvFWoKI3jiD31m739d85K5/GOhqaUCXprfSPaVpWids1WXPHUpxg3pCsGdG7lHJevZcm67bj+xc/Rr1M1AOCxKUvx49H9sWJTHVZtqcONL+qjqw65YzIAYOkdJzptqvP73QVrICCcybyiNO5boqTE1hh22AmAt/53Hp6Y+o1zXH19G3c04oqnP8Frn6f6c9pUlqb4QC58JF1zunfyQvz57YWoKos7bfnQQFiAMMxuiFy450t8XPTIDLSrLMW/P/0WQOokLdlS15jyfOn6HViqrNa32ZPwum0NAIA35q0CFLeNlIXSL7LDnkAr7Un1va8sn0dFaRzbXZOraga65LGZzmPp6/j7B4tx+2tfAACG9W7v3KdLmwrP1y2jplZtqcOmHQ146dPUcN+mRPITeOETvWYEAI/a5kX1/VCZ++1mAMnXDLAJi2EYDz5auA6fLduU1bmOBStHDWRHQxM+/Wajb7/JC9Y4wsPEmi3euRArbOfwjgZ9OQ/5XiyyBUirCmt9XFUWx3MzluHXL1vSZv32hpTz1m+rx/kPTXeeT5q/2nksNQwpPICkaeiW/8xPmbB1NNsC4uGPl2LorZMcIZgJi9Zuwyuf6fNM3PdROW9Un4zvlSksQBimhfL9f07D+Ps+9uzz5rxV2mRBacLKVQO59T/zcerf/oflG/UrYxMblUk8kRC4552Faf4HNxc9YmkGqi9E5dx/TMOEx2bi359Ygqp2ja2J1Dfjly/oEya/94+pmPjhYuM9dfkbc7/d4jy+4UXv6K+mRO75GE9N+8a3T5NGgFx4aJ+c7+0HCxCG2YW59PFZeHZmejinNGFp5p2MkCGjX6zcauyT0Nzk3H9Mcx5PW7IBd036Cj99brbv/eoam40aCAC8NX81FqxKHcviddudx+cM75Vy7OPa9XjwfbMAeWPuKqzabBZsflV2Zyz1187CYPPOxrS2itK4pme4FHJP9J5E9C4RfUFE84joKru9AxFNIqKF9v/2djsR0T1EVEtEc4jowEKNnWFaGu4wUsdy5TJhzf12s6MJNCcE3pq3ymjmWr2lDj3aVwEAlqzbZrz31rr0Cf+LlVuwbIOltUgfxWofExYA7PWrN7TXC8r+PdpmdPyv79Zi5O/eyfp+Ovbq2jrU6wHAnOWb09oqd2UBAqAJwM+FEHsDGAngciIaDOBaAO8IIQYCeMd+DgDHAxho/00AcH/+h8ww+aGhKRGqE/TSx2elRAxJoSCQ6gc56d6PcMzdHwAAHv54CSY8Pguvfr4SbrbVN2HEb9/B09Mt84rJrAQAywzmrV/86zMIIZxciaCs9jF1eXH0Xp09j4/Zu0tKJFPYjB5Ug18cOyiUaxF5C6MgWeu5UjABIoRYKYT4xH68FcAXALoDGA/gUbvbowBOsR+PB/CYsJgKoB0RdcvzsBkmL5xwz4fY+9dv5HwdWavp/a/WYs8bX3fapVVJiHQzljSHrLRNN9J5reIOX5UazpvzVqWV5JBCxs20JRtwyn0f46R7P0pp/9VJg3H6gT2Mr0nu8mfiH+cPS2vbu1sbPPnDEejsEzVVVRZHrw5Vnn1yobwkhrGDu+Cxi4Z79jvIjvTyomN1OZ6+ZGRa+7MTRuLFyw5x/FxRUhQ+ECLqA+AAANMAdBFCrAQsIQNALhm6A1CNucvtNobZ5ZAOYBNBo6eqDatpmTOxYXsDxt71ftrxX/zrM0coTFm0Hn2ufRXzV2zB1MXrtfeevnQDFq3dhksfn4UzHpiScuyr1VtxQK922nF8pjG9HLlnDdpUmjMM/ATIPt3TzVSDurTCoQM6eZ4HAOWlcfSMUIBIv0RfO0fFRBAB0qVNOdpXl6FLm/KU9q5tK3BAL//zw6DgAoSIWgF4AcDVQogtXl01bWnfZCKaQEQziWjm2rVrNacwTMsnaHnvqjL9RCwFyLebdmKJ4mSWPD9rOZ60o3/e/dL6Hf3k6U9w9sSpePXzlWnRRZ9+swlj/pQURF+v344l67bj2007MWPpRgzqEtzuX14Sww4Pk5g74uvSI/ulPG9dkf6aYwFX45WlcfS2BcjYvbsEOueX4wbhtAPNa9n+NdU49QDreLmtEZaXek+9Qcx0HVtZgsNdUbm8JHrfh6SgAoSISmEJjyeFEP+2m1dL05T9X1Y8Ww6gp3J6DwBpwdFCiIlCiGFCiGE1NfpsV4Zp6QQtw15p0ECySf9YtNYSNMs27IRXdGq7qlIc+cf3cNSd7+GIP7wLADgwgxVxeUkMPdpXGo9v3JGMOLpgVG/07pC6mo9rfComc86ofh3x7i9GO6VdKkvj6N3REiBeY1CpLI3jD6fvh8MHpms4vzttX7x8xWGOViNNiuVx70le5pcc5qE1ta0sBZBuTiwvyd+0XsgoLALwTwBfCCHuUg69AuAC+/EFAF5W2s+3o7FGAtgsTV0Ms7tRH7BwYKlh/4hcSpjECGj2OL9Kif6RCW5H+TivVcpL4vjx6P544uIRvn2JCG55ESPCG1cfjid/OMKZTNU+7/1iNG44YW8AQIdWZejbqdrxA1WWxXDKAd3x21P3NZrd3JTGYyiJx9C+qizt2B7tKtGqvARtbK1ITvZV5d4C5NcnDcbVYwfizIN7Gvu0Kreu6dYG8xG+KymkBnIogPMAHE1Es+2/EwDcAeAYIloI4Bj7OQC8BmAxgFoAfwdwWQHGzDB55cR7PtRWofXLPwCAZRt2YO1WfWhsLvkfRPrcDklVuc6EFPz65aXWhHzYwE6oaV3u299tnooRsFfXNjh0QCcn30Xt0qdTNTq1tiZ7dwRYRWkcrStKce6IXjhmcJdAobCyj07JkZFQXdtazvsN2xvt9hjOHdEr/QSbPdpV4uqxe6LU442Tvih3EmHZ7qCBCCE+EkKQEGI/IcRQ++81IcR6IcQYIcRA+/8Gu78QQlwuhOgvhNhXCDHT7x4M09KZt2IL1mq2O1XzOqYv0e+Tffgf3nW2XHWTmwZC2sxniS4MVmdWMqHuuveXs4b6jsXtHVUFinzsFjJSJrvHpa7eq8pKMOlnR3je//xRvXHS/t1S7nHifsngUKkBdm5tCRD18zBVEgaSL8nLdSP9YO6PMpP3OlcK7kRnGMYbXcVXVQM5e+KUtON+5FoDS1d7SaLzN2QSUhpTJ0Cf04jShYP6VB5z37/ZNvu4J3G3xuE37lvH7+M4rWXP3h2q0KHa0nCkAOlmayCdWiU1Kq+JPqk5mfu4vxcPnncQHvh+fvOruRovwxQ5OgGiRmFlY47KpUQTEXlqMDr/TJBF8VVjBmK7q9ggaSTIOcN74unpy+zj6TJGnXTlI/f9pQblnsTdAiSTxTwp2o58f6SJrGeHKvz9/GE4uE8ymMArMowMmpPKsUO6pjwf1KU1+viEB4cNayAMU+To/B1ePggT6mSZiwnrtv/Ox6yvzTWetmjqMgUJoz1yUA1uPGlwSpvutN+dtp/zWmIxQsxjFpPnu+8/om8HAMB39t8jpd0dtRY0/Nfqm/wv317VH3HM4C5opzjaDfEN2mu6ufb4vXCya+xRZtCbYA2EYYocrQDJYv5XzVZBz4/HSGuuuvedhcZzNuxoSGsLMhHr+pjOihHQbB8PspJ3dxnQubV2TxJ3BFMm3gTnHhoNREcsgHpjem06YWEK2Y4S1kAYpsjRmbDcGoQsTOh9TvJxUB9IF0MUVA+PbG1djkqQhbzOqWzyATimLZ/rxgwaiIkKV4JfJr4bOaYYwUlxNoVRA95O9OT9Te3pB0xJo1HCAoRhigx35JQu69wtQA63E/Ykat0rHUE1EFPtKFNJddOiWp3A22gyxQFoTVHmCdT+D/IUDjGDBmKizDXhZxPQRCCnRIanAMlBA1FbZe2ufEZfSViAMIwHk+avNobJRsWFD6fuea1zSnspEO7S7annWScG9YG46yxJNu1I93MAwHDbt+BGndve/vmRuGrMQE2fTExYydW+l3Aw+UDM/d0RXZlPypYPxHp/vSriZmvWc7f/+7JD8O/LDslwlOHAAoRhPLjksZk488HMw2Qz5bXPVzpVcN3lz+sCmLBUdJsLSeRpwQWIXgPZrBEgPzikj3HFrU54nVtXYD/Nvhz6EiT6cakJgoF8IMYeFof076htz0oDoaSGV5KjBmJ6aeqpnVqVZ1QqJkxYgDBMgVm+cQcue/IT/OTpTwGkT3Y6jcLLBOUWICfu1w1XHj0AADDpi9W48aXPA9fC6lit10C2Gvb2Nq3YYzHCOCXsVNdNLwi8V+CWCUvbxe6XvL8Xj1w4HJ/ffGz63bPQQIgIAv4aSC4CJA+V2gPBUVgMU2BkToZpVz99FFZwDaSiJO5oBpc+PgtA8OKG7mifTq3KsFfXNviodh0AYN/ubbFX19b416zliBFpnfNynvzb9w50fAO6iVk3nwbRQLz0C+nY9ptwy0pi2hIg2UzURMDoPTvjjXmrUOIRY5yLCSsfe30EgTUQhikwcuKqt6OX3JODzgfilQfiNi+VxChtItRFaemoSAsNJRy/b1KTuGX8ELSyneIx0gs2x18RI2fVrZv+tCYsw7iSEbMBNZAsJ9xszosR4c9nD8UH/3eUZ12qIJcO4gMpJCxAGKZIMO3x0ZwQeHzKUlz37zlOm5cJa5NLA4nHKU0obQ+4Xa6uNPj3RvROXpsoRUDoFCOtc1yrgWQQxqv4NrxW48mMbmMXT7KLwrLySXp1zH1jKnNUW86XDgUWIAxTYKS93KQVJATwq5fnOeU7rDazBFm4ZmvKc10y2w6DD8ON337l8VhSAyDSR4fp/R2atkw0kBQnunl8arhvNmRzXpjagVmAhnaLnGABwjAFRmoT9U0JNDUn0qYsnbDwSgScvyJ1Y09rks9OA/Fz9MaInIk/pjiP3X3c6CZmfSKh/r6OCStgHki2K/ZsfSBhESSvppCwAGGYAiP9Gc0JgQE3vI5GlylLJyy8TFjuWlQ6H4gpEdCNV6VbACiJk3aSvvHEvZVrpF83qFbipwHECJ4xukGq2npfP7sorLBgJzrDMEZmfb0hLYvc7QvRCQudVvLV6q3a/vFYLG2O3e6x53jquQE0EMVRLYdVGo+h2nbA6/JYtAIkgzwQ9Xgu0Ux+ZKWBBOwXJJQ6SCZ6IWEBwjAF5OXZK3z76ITFEnt/cpXv3PuRtn9uGoj3cdU8ZuU/2OfFyDGT6Yox5mzCUu4ZxDyVvRM9Gx9IdvfSYU4kLA4RwgKEYYoM93zrDtltaErgT5O+SjtP5oukayCUNmEH9YH4TVSWcFJCc4U8z++6we5lMmFRgD4p/UOMwvLbMjZIld2gBMlELyQFFSBE9BARrSGiuUpbByKaREQL7f/t7XYionuIqJaI5hBRfrfeYpg84fZ5uAVCk89uUO7ztRpIwCgstwnLPW+pkyVRMqLMr9KsNow3g2KKzjkBNZBsfQa6847fp6ump3JOVnfSY/aBhHiTHCi0BvIIgHGutmsBvCOEGAjgHfs5ABwPYKD9NwHA/XkaI8PkFbfJZ/329P01vHCbsHR5IDuCaiCG2VkKFlVQqLf1W4XrJsBMamGlHA9kwgpvxjUJx1wd9jrYie6BEOIDAO5Sp+MBPGo/fhTAKUr7Y8JiKoB2RNQNDNOC0TlS3RrH09O/8Tzud35JLN3IY0padGOaLGV+iFtrEI4JK0D0VFpbBiYsJ7/D+14ioEktE7IRjlnfi30gGdNFCLESAOz/ne327gCWKf2W220pENEEIppJRDPXrl0b+WAZJp/UNzV7JhHe8foC1K5JrakVj8XSJrX6ptyisGRtrRJNhJd1nt+VNSasDJzo6vEgk2mY862veS5EI5axOGVxyI+iFCAmdG9Z2i9JCDFRCDFMCDGspqYmD8NimOzRJd55Mfau9yE8lIcH3l8EILUKrFYDCVgLy5QHYqpp5URh+fpAdPfS9DNfwblPLjWlsiEsDSTIZ2+6VZEoIEUpQFZL05T9f43dvhxAT6VfDwD+MZAMU8QELasuWbZhZ6C9PE7ctxtOO8BS0OMxSpv0dBV+dcQI+MnRA9C6PLVwd7e26fuECCQd+P4mLE0YbxY+EDlGP8L0GfhrV+HBPpDMeQXABfbjCwC8rLSfb0djjQSwWZq6GGZ3ojmAAIkROYJGp4HUa/Yt1xGPEX5+7CD85ZyhKe2PXjQcfzh9P7SvLktpD6yB6Nq05wRxWPtPpmGafPxMWJkuCrzgPBAPiOhpAFMADCKi5UR0MYA7ABxDRAsBHGM/B4DXACwGUAvg7wAuK8CQGabgBNFA1KS+uGbfV9WJ3qN9pfE6McdUlXp+lzYVOPPgnva9ku1yaH6r9KDzn68PBN7CQZqJ8lHgMAqKPRO9oBtKCSHOMRwao+krAFwe7YgYpvgJVgJD3VaVQK60DzVUuEubCizfuFN7HbnazrSqbTYmLB2mXrI9Vggneh492KyBMAxjJBtrRxANRN0d0M/R7DUVycky0IQlRGATVlD8Vvvk+9rI6RcWYcmPnGphFYf8YAHCMC0NXW0pN+rmTla2tnnGCVYOPbMZK0gRxiD49fIL402asALdLhD5XP0XuwmLBQjDFJBsHK5BTViq/d9rwvGaD51wXTL3Tbm6PbhAGeQKXdukR3V5XUdNJAxCqJN+SKG1FxzSB2cc1MOzT7Hke5hgAcIwLYygJixZMovSfehpfc3HktcA/CfshONED54rcdGhfTH1+jS3p30/fxNWsHLuvl0CY7pfpouBVuUluPO7+3v2KZZwXRMsQBimoGSuggQxYVlXTppvvCZiT+GSgQ9EHZVfqKt6PS+BGCgTPcAsFmp2eGhX8oc1EIZhQiWg/FD6kees5yUcpCAIagKSQsvX+a089qsurD8/GV4cZTl3Hfn0gZjexxBTTXKCBQjD5IlnZ3yDz5Ztyvk6XvuhO30gUgoJek16mfhAgp6fiQnLS6Py9YGQTx5IwOKOmZBPq1KxayAFzQNhmN2Ja174HACw9I4Tc7pOUA1EiKQ24DUPefpAHBMWnGt531Ne03ts6nWamr0EiJ8pLJifIIiZKyj59EuY7lUscoU1EIYpINlEYQX3gVhoEtFT8NRAnETCAD4QoQgQPw1EeexVmsU3jBf5L6bIPpAkLEAYpoURJApLiGS/nKKwYv59Uu4b4Jru41mZsJQH+c7KzmdynzHiK/xbZQULECYrXp79LZ6a9o1/xxZMEF9D7veI7hzZzzJh+Yfq6o9JZ7UZ9Zh8z/y3tE0+bvISIH5hvD5jk7RcH0hxqyDsA2Gy4qpnZgMAzh3Rq8AjiY48yI+sCKSBKP0Ifias8KKwJH4+B1UwnLCPeWNRv9sGHVeo+4HkcU4vcvnBGgjDmAgyUReCIOXcgdQIJC8h4elgDxCF5dxPMaz4JwAmH5+4n4cAMZ4fbFxRbGlbqGq8vz5pcN7uGxQWIAxjIGi0Uy5kuiMhAKzbWu9/XRG8lIlnHkgmxRQVMi1lYu4YznWKfSVvQhV8Fx3WF4f071i4wWhgAcIwBrKZ3PPBhMdnBep3x2n74YyDemBEvw4+2ebmY+5y7jrUY0GVtqCreD9NJujWuWFoDZWl8ZRrmgjze8OlTJhI+Gb9Djzy8ZJCD2OXJkwLVsKgzkRpJevZoQp3fnd/lMZjPqVMgkdhBZ08/aOwvK+THJt/H6+3MMxEQnmJ/FbjTX1ebFZVFiAtlDMfnIKb/zMf2+ub/DsXASs370Rjc+YlKwpJqAKkwL98rwlbPdSpVeoWtcly7v73UPNAfE1YAbMpzD4Q+T+oEz1QN59r+EekhQ1rICFDROOI6EsiqiWiaws9nkKxeWdjoYfgyY6GJmeM50ycilG/m4xfvTS3wKPKjDAnfZPjOzqxknrloHkg9517YMqxIImE8phIu6uZ4L6L4gnjjWoqb13hHwzbr6baGkORyZMWJUCIKA7gPgDHAxgM4BwiKr7QhDyQLFpX4IEYGP3H97D/LW8BAKYsXg8AeH3uqkIOKWPCFCBZ1AsMmWB5IDLqav8ebbHgtnEZRWGZ7vb70/dNPx5UgLiey31DVHNSoE8pjN9JRCasD/7vKDxx8QhMM5S0f+qSEXju0lGh3jMsWpQAATAcQK0QYrEQogHAMwDGF3hMBSEfEUK5sEYTKVTf1FyAkWRPrm9xn2tfxa3/mQ8geOhtVKhzntuco67ynWNEqLCdxlZ70LDZ9Nc5om965FBgE5bS7d1fjMYbVx9uPO5FmBqI+VLZ3aN9dRkOG9gJXQybah3SvxM6tSoHAIzsZ72XXdvq++abliZAugNYpjxfbrftdsgfarE51byobyr4MjwjRAjDfcgOdMi3D8R9O3Vqc1fKVSd9Z6J1XcDLhzCgcysAwJ5dWjlCV51kdRNuYCe6MvK+narRrqosZZx+l5FlUsLwgZCvOS/6z/iKowbgw18ehf41rSK/VxBamgDRfXIpnxoRTSCimUQ0c+3atXkaVv6Rv+8WJD9S5qTHp36NG1/6vHCDCUCY4ZhqFNYTU7/GgbdNsu6Rpw8wlqJlZB6R5aUxHDekK1698jCcMrS79gupu19g57Chm1pmvneHKpwzXF8RQfoOTKv7bCik1TgWI/TsUFXAEaTS0gTIcgA9lec9AKxQOwghJgohhgkhhtXU1OR1cPkk4WggLUmEJPnVS3PxxNTcamlt3tGITTsaQhpROmGaCdWCgTe+NBcbtjcYQ3vDIE0D8dEIfjy6v/f1fITpkD3agogw0k50a1NZ6tk/cB6hSYAo2kAsRvjdafvikP4d8ZOjB6T0u+2UffD4xcOxZ5fWAe/oP5Zir5CbT1paLawZAAYSUV8A3wI4G8C5hR1SYRCu/y2Fr1ZvxZot+kzqmUs3oKqsBIP3aBPoWsfc/T7WbK3PeX8NE6E60TWXahYibwsAv1pY14zbC9eM2wuz7Q2v3KNywnN9pv6bvzMEFx/WF51bJ1f8ehNWbmG8jgaitD11ycjF9JJ9AAAgAElEQVS0fv1rWoVu7okytPb1qw53EhZbAi1KgAghmojoCgBvAogDeEgIMa/AwyoIjgmrCNwKDU0JnDVxCq4/YW8c3KeDZ99j7/7AeOyMB6YACL7hknTUr95SF6qJQpLL3O7WLnTCaNOORvz702+zvkeMzFqSW2NQJ363EFCF2B62c/a4IV211/WbO8tKYmkTtlZY5BjGG/P1R4SPvFOUGsje3YItnoqFlmbCghDiNSHEnkKI/kKI2ws9nkKTr3Ib/6tdhw++0vuUlm3cgU+/2YRfPj8nL2Nxs2F7NGYst3bw29e+wGG/nxzoXHfUlW7PizVb67IfHDKMLArYtXObCsy5+Vhc5mPSygSt/MjNBYKSeAGT+oo1dr4AtCgNhEknXy6Qc/8xDYBeO6gus75GW+vCSW78cOFaVJXFcVBvb21G0hBRdJf7rZ34weLA57oFhk6A5PrZWQIkWIkUVdj4zX9tKtL9F7kMVWf2CmzCMnRzMuTzuAT200Cq7N9BabzFrcuzhgVIC6eYfCBb65rw+JSlGKfs7+BexVeUxlDX6D3hn/fP6QCCm7KiKJHyzfod2LQze83GLTB0wmJnY255MWEpIJmYgbJZe+sm3KDXMQmapA8kf9pAm8pSrN/egIoSvY/iZ8fsiVblJTj1gN0ns4AFSIhIu7ffftBhEtQJu2lHAyYvWIPTDuyB/9Wuw7wVW3DJEf1CGYO079c3JfCrl+fh1c9XOsfcE+mgrm3wme2ozYSm5gTiMf2+FiYNZGdDM256ZS7617TCpUdmZpI54o/vZjxGlZdmp/o2dImEOxtyEyCZmLBSorBcxyJ35OfgRDfh7HiYR2vSr78zGAtXb8XRe3XWHq8uL8FPj9kzfwMqAnYfXSsP9L/hNZxwz4d5vWfQn/5Vz8zGz577DIvXbsO5/5iG21/7Iut7qg5iIdK9MJt2JE1Z7u1K2yh1f6Sc/XDhWoz63Tue9xxww+u4/kV93kiDQQO5860v8dzM5fjd6wvyWshxxtINuOHF1LpfOhPWrK835nSfTNYpXiv1kgAmlxL7Zm19wnOD3jvXfTx0UVhR06aiBBOO6F/028zmExYgISIEsGDV1rzfMwgrN+8EYJ5sM0FdTSdEesSR+gNzRx81NQsM690eV44Z6Jx7+6tfYOVms0NZXv/p6cu0x1UNJJEQ+NNbX2L1lrqUciq6CTwqNmqc+rpV/l/eWZjTfbLWQFznlcb9r9OzQxVu+s5gPHjesMD3lOSikJsEXzKRMPrJPHmL/Gs9xQ6bsFo4hUgkVCdj3cSsOjYbm9OdyaXxGMrsSatZCCeiRsfPnp3tG+qqCsXZyzfh3sm1uHdyrXHMuSKE8Jy43FoXEE0tLK+50303rzmvJKAn+sJD+wbq50b3XuXsRLcFSD6+/6WxGBqaExx8pYE1kBZO0J+PLhEs2x+fOkEmhMC/Zi1POR5XfmkfLkwN/W1KJFASJ8TtSas5IVL6uwmSJ6Gap0zZ3U3N4U00OgHhdzwKDSiTK6p93e+2lwAPA93Vcw3jlUPOh2bpDhlmQZKENZAWTi4LsIbmBMoNESVeNDenaiD3uEwx6orziqc+TTm2aO12HNCrnWNTb0qItOJ+JsoMtnrVhGX6cTeFWE+9qVnAK1m4yWUmPPfvU9G7Y3Vo95d4TZ6ZfC+iDjvV1sIKeK5J01MXIFEjv6uF2FCq2An8zSGiw4joQvtxjV1OhCkwQRMJk6VPkv2zrY6rTsY604yXPNi8sxElMXKERnOz8DShlJUkj5ls9Q3NXutrOebgE41bALhp1AgjIQT+OnkhzvvntLQw5f8tWo+np+dW90tHJqVWhMdbVBJ11GAkpUys//mociwFbJj7q+8qBNJAiOgmAMMADALwMIBSAE8AODS6oTFByPT3o67Yvlm/A3WNzRjmU37EbRZSr6EzGflNDvEYOWaBpkTCUwNpW1mKtbYzXEYLNTQlUgShqoGs36avsyUFyM6GZixauw37dG+r7bdmSx1aaXaIm/vtZudxs8Yc9sD7i3HnW18BAD5cuM74esLEy/TnXlioz91nRa2B5JSJ7hOFlY/guqQJizUQN0G/OacCOBnAdgAQQqwAkHt5SyZnMl1/qZP/Sfd+5NSf8uKBDxalPG/0daJ7/8Qam4UyAXg70dWw0dJ4DNvqm7Dnja/jyqeTpjEpQJ6ftRwTHp+lvc7G7Q2YuXQDLnpkBk669yPt5lYfLlyL4b99B6/OWZl27KR7P0qO36WBLNuwA79/Y4HxNUTBFUcNQI/25rLe3Tw2HFK1OiBYFFYu6H0gQZ3o3rWwwjRNmpAaMlcySSeoAGkQlsdVAAARhW/QZbIiqCNc9jOZci55bCaufuZT7bE/vPFlynN1xa8zYXmtjAFrZ0LVB+KlsVSXJ7WBsjjh6/XbAQBvzlvttEsnuqlWFwB8/5/TcMYDU5ztdesakq/hi5VbsLWu0cmAn75kg+f43ULzuD+bC0RGxQ8O7eN5/KoxqQlt6sd0xMDUbQ6C5IHkQhR5E/L7kx8TlgwZtp7nM/u92An6zXmOiB4E0I6ILgHwNoC/RzesXZvG5oTvJBUU9+/n4kdm4E9vfanvDFNNJoFJ81fjpdkrNGeko67e/cJ4dTQ1ixQnqJcNXj1UWhLTZm83NCVQ39SMVz4zj19NbgSSZUQ272jE8X/5EGc9ONU55ucucUd07cgxo9zEoxcNx8DO+lLkrVIEa/INv+vM/fHhL49K0zIkhw/slLaqj9oHEsWKPZZXE5atgSBVkDABBYgQ4k4AzwN4AZYf5NdCiHujHNiuzO9fX4AzH5ySYlfPFrcAeWfBmrQcCMvub63cdeGsmTiYgVQNRFdGxG/FqQqN/85ZiY0em0Kp1yqNx7BdM1k3Nifw4VeZ+R2Wb9yBd75YjdV2Rdz5K7c4x/y0OvX9WmfwuYTBkXvWYIhhb5TykhguPsyKY+nevtJpP2Vod+2OdSk+dNfHE7kPJIIVu9Ryo9yUSyK/q/mqfN2S8HWiE1EcwJtCiLEAJkU/pF2bNVvr8I+PlgCwIpJyJciX+vlPknkaOo2hzqOon66/Gr2li+TyEyBq6K6X70AIkaqBxGPY2dCU1m97QxN++NhMz3u6kb4f3QrfL+lPjdK6JuQS9t8f2Stlp0aTP4mIcObBPXHmwT1x8l8/8u2vCsW854FEcPm4YgKNGqnNycUXayBJfJceQohmADuISB+2wmTElEXrncdB8x+8CGICVm+jczp6hfPqakjVK2Gq9ZrKun4vy89slbx3asZ3WZywvT5d2K3dmr0WsHDNtrQ2P5OUnLQ27WjAOwvWZH1vADj74J4pz90hzX7+JCCzSDwiStdAIq6JHqUAiSLD3438rsrfAvtAkgRNJKwD8DkRTYIdiQUAQogrIxnVLoxqQgoyiTYnBOoam1OcySpBfj7qBJGpBqJb4ak+EF00k9+qsLHZO3RX7af2ql2zDZ8uSy9AuCYHAaLDb591+Rn+T1kMZEOP9pW44/T98MyMZI0vt+/C/T7NveU43zwVHalpIC4fSORRWPrrH9K/I84c1lN7zA8ZnVdu8PWEifSByO81ayBJggqQV+0/JkfUCdw0iX7w1Vqc/9B0fPB/R+GeyQvx/Kzlxr0xgkRhqROEFfWU6ih2ayCrNtdh6frtGNmvo3ay8vOB+G3w1Bww+7ypWaT8WLc3NKeYdwCgXVUpVm7KbWc/N26Hu5vx932Exb87ERWluU1eOlOfe1GhmqQ6VJelOM+zgZA+AfbtFG1Qpemj1u1hHpQrxwxE28pSnJaHvTeciEFpwor8ji2HoE70RwE8DWCW/feU3ZYVRPRdIppHRAkiGuY6dh0R1RLRl0R0nNI+zm6rJaJrs713oQlis/237bOY+fUGPG/XmTIJikAaSDxVA3FPXG4NZNxfPsDZE6fie/+Yiq11qT4HIYSvD6QugAAJUl69oTnh60/p16kaq7aEI0B+dsye6FBd5uubkh9hhUc9E9OkecrQPTz7uF9uEBNWIJQvirzkSft1w9s/OxIH9Gofzj0MRJG5XVEax6VH9o88BBlI/n4cExarIA6B3n0iGg1gIYD7APwNwFdEdEQO950L4DQAKQH0RDQYwNkAhgAYB+BvRBS3Hfn3ATgewGAA59h9WxzqxGmq4yMnTfWw+lid4IJoIKUuDcT9/XcLAbkC/7h2Pd77MtXGnxD+Jqx6n532GhOJQKGv9U3NvgJEnfxuOXmI7zW9qCyNo7I0jk0aAdKhuizl+fQlG/DG3FUpbS9dnizMILf5PbBXO4zsl8z0/+1p+zqP9TWiUtuCaGpBAilkH+uW5Fx7gCFMOExa+nR73sjeAIC9u1kRcS399YRJUPH9JwDHCiGOFEIcAeA4AHdne1MhxBdCCF2ywngAzwgh6oUQSwDUAhhu/9UKIRYLIRoAPGP3LRqCJvSpE6dJgJAjQFKr3koeeD+ZGa7e1jSGkhQfSCJtkvLygUjh0s82czQlEikmKq0G4rpeDyXMFLBKgQQRINvqm5x9TEwc1DspQNz28MHd9CGwJirK4qgojaW8vvG2xjB6zxo8fvFwp/3MB6fgsSlfp5zfvV3ydUqf1U3fGYJedljt4QM7OftmA0lNwGucqgA5ZajeXHPJ4cF3llRNWPnaGKmlL9jHDu6CpXeciK52dn9Lfz1hElSAlKoTvhDiK1j1sMKmOwB116DldpupPQ0imkBEM4lo5tq15szksDEJgxWbdqZMqNvrkyYhswZiP1AO73nj685kqpo11Cu4zWO1a7aiz7WvYs7y5Baybr8C4B2FJY9J524i4R/G6257/OIRqFTMPc1CoKrMvwrwuD9/6OSvmOil5DyUu3wS7ud+VJbGUamMq1+namdyLyuJ+Zq2VGWhuty6Tl1js2MCkQLu3BG9ACSFw5M/HIHhdj2yNBOW3ecnRw/AjSfurb3v+KHdjQmHEnVtIW+Rr4lwVzP57GqvJxeC/sJmEtE/iWi0/fd3WL4QI0T0NhHN1fx5aQ66T0Z4tKc3CjFRCDFMCDGspqZG1yUSTOGEh9wxGT96IvlWbVfyGNQJX52c5MpwshIiKkTyubrSVm/rThKcNN/qr2ZoN2tMWNvq0nMrJB/XWgl65bYAaHb5QHS+jC2uibZvp2pcOWag87x1RalxNd27YxUe+P5BxvG46dym3HnsLk2vK/9eXRbHpJ/qra+VpXFUlSY1hJeuONSZwMtLYujjU5Jd1RakBrKzsdkRvvL/90dYJhH5ObevLsMhAzoCSP+iyz4VpXHPGmMvXX4opl8/xnjc2Q9GCePlrVmZXAkqQH4MYB6AKwFcBWA+gB95nSCEGCuE2Efz97LHacsBqHF9PQCs8GgvGnQ13WSW7HtfJjUhXR2puyd9hf1veQurbWewtDi9/UWy3hOQ/MGrK2vV/u0u8iePqYpJU0KkmbBWeWwnK0NVK0qSpUfqfaKwdNnicu4rL4nhkQsPRixG2uifUf06omeHyrR2E+0qk74JtwmrXOPkJiK0qypLaweAyrIYKhQNpLwkhnkrrAz1AV1aY5/ubT2Fm7oyHbdPVwCW+a7M0UCsa0tzpHYCd7U5Zct9gi+qy0vQuY25gKJzeeW+UVdxZ3Z9ggqQEgB/EUKcJoQ4FcA9ADLficifVwCcTUTl9n4jAwFMBzADwEAi6ktEZbAc7a9EcP+s0WkgurYUJ7qtMci9sZMrd+uX7c79kL93dWXtpYHIY6r/RKeBrPDxMwDJiKPmhEhxkgeJplI5f1RvzyqyREkHdBDUvAm3BqLLESBKrfCrUllagkpFOJfFYxjR1zItjd27MwBgj3bmSVqdkM8b2RuzbhyLAZ1bO4JMBjM4AkQZnvurIkuYSHNlrglzuu1AWANhciXoL/UdAGMByLTdSgBvATgkm5sS0akA7gVQA+BVIpothDhOCDGPiJ6DpeE0AbjczoQHEV0B4E1YgushIcS8bO4dFTp/hq5NneTdPgv5TE5ErcpLtHZ3dWX9xNSvcev4fRCPUVrOhnSqC5cGok4cRAiUR1GuaCA7FQHil/Phxt9+TMakScCahJ+ZMAoNTQlMWZRa/yrNB6ITILCETmVpPOV1AEBN67KUsFAiwtnDe+HUA7s7wqnSM3Q3+driMULHVpZ5rdrWauTnvWeX1ti7Wxvc/J1k1Jj8iAjA/FuPcwIfjh3SFfdMrsWYvboY7xsE+V0gSn4GbMtnciWoAKkQQjg1H4QQ24jIvIz0QQjxIoAXDcduB3C7pv01AK9le8+o0ZkYdKWmGzzCeGV3ORG5Hc2kmIEkT077Bvt2b4uzh/dKufaMpRuc64kUDSQ1u7t3hyrfSCcgKbQSQqRUxG0w7DVeVRZPibQKun4mgmey3M+PHeREXo3q3zF1jGn7XKQLEOlHaFtZmi5AWlWgTmN+UzWbSg/nv+oDUYVJlf165OddURrH61cdbryOGqm1T/e2xiTSTJAaZBtF+2ITFpMrQU1Y24noQPnETv7zn3V2Az5btgnb6ptShMWw37yN2jVbtUmDKSYs4dZAUksluBPVTCUhpM9B1W6++8AUZ9J2+0DUy/TuWI0VHj4QiZycr35mdkqehMmEVdO6POW548T1uQ8BnhneOse4FCjuXA1dDoVsuf7EvfG0KxO6TWWJb3ixlwaiLujVx44GYhC2+eDwgZ1w44l745aTh7ATnQmNoBrI1QD+RUQrYC0m9wBwVmSjaiFsq2/C+Ps+xtF7dcYdpycTxNZtq8dT05bhyjED0s5pbBZoXV6CrfVNaHY7vV0aSJoJzP69u9vLlO1hVaRQUx3tzc0iRR3o26kaHyxMOvl1+20ASQEiN2SSmOoyGet8+cxZqolFOw6NcHnsouFYu7U+zbeiG4O89sn77+H0SdY4IuzwSYKs8vDPpJiwlMdyIeDpLxKpi4ewISL80M4XkQsR1kCYXPHUQIjoYCLqKoSYAWAvAM/C8k28AWBJHsZX1Gyts1bic7/dnBaFVVkW0/pAGpsTjjnIvSJ1R+e4J5xfPj8Hx939QZppLFlqIbX9z28vtK+bev+tSi5Kv5rqFB/JqX/7OP2Fwly2w1SaJdtKw36VTt2OcsAKNuijiejSRSW5hyVf1z/OtyrqSBPWXWfub7i/9V630eybrgoNVZhIQWbK+wFUH0j0s3pya1aWIExu+JmwHgQgS5OOAnA9rJIiGwFMjHBcRY8QAnV2KfPKsniaOaqyNJ42YZz/0HR8uHCdY6JpTgi8MTe5/3ZSA7H+6ybnL1dvxTUvfJ7SJgXIu1/qS4urQ3vgg8Upx9ylLBas2qq9hqnqqWlSjGdZItxvTjPttKfjuwf10N0h5dkPD+8LADhsYCcAwKCurQEAh/TvpL1mLEZ46pIReO//jkpp//Ho/kYTVqk95kafUFz3eVHBUVhMWPiZsOJCCLn36lkAJgohXgDwAhHNjnZoxUt9UzMG3fgGjrdj/StK4mlO9IrSdKEi9+yWdvT6pgR+9MQnadcnR4AEi3AqLYnhm/U70vYuT5Ichztqyi85TmKauE0aiNt8FHQ3N78prTQDzUbnjHefftWYgfjJ0QMdjemO0/fFeaN6O2UrdLiFy/Trx6RpO+rqXpbT9yrDnodtLRySPpD83TMX+nSswsmGxFOmsPgt5+JEJH+FYwBMVo7lVle6BbC1rhF9rn0Vf3et2qXm8bpdTK+iNN1cVVEax2fLkiVEttQlHc8ykmfJutQyHW4fSFCn65VPf4pPvknfJ0PitfDt2qYikLnJtO3pU9O+0babrulnovEzq2RiGtOFA7svT0Qp16wqK8HBfTogE/zGJHNHhvU2V711AigyunN2yPfYK7O9mHjv/47Cz47Zs9DDYDT4CYGnAbxPROtgRV19CABENABA7ht6FznrtlnWuyemfY2R/Tpi3fZ6HDWoc1q/5Rt3YtnGHWntqnZxyn1J30KFbcdfuCbVXJSMwpI+kODL0s899ldXcwDcK91YjFAWj2Fnwtt5HGTzK6/+Ya2wM9n8qKwkhk6typzPEYjGx+DeRdBNv5pWeOfnR3pqe8lSI2GOTE++a2Exuy6e33w7J+PnAB4BcJhIJhTEAPwk2qEVHvn7EgL4zl8/woUPzwCQnvOxfnsDzvvn9JQ2t7lisVIUUEYSfVybGtHk9oG4o7S82LjdvIveRrs8uwyB3b9H6u7EQVb17j7uCrtuzHt5e9/H73imvpWZNx6DKdcdjda20zuKRXeQIfWvaRXofc6LY5vDeJmQCLIn+lQhxItCCHUr26+EEOnG+10Mk30/SFkJr42jTLkEW+ua8Lf3ah1vQSZ5A+s8BIhE+jHctaDc85pufG4T1m3j9/G8V6YaS1CyuW63tpV4ZoKV8xHFBO2ngQQhnxkiHMbLhMUu78fIBelwdjuA/QrbAcAWjwq3pmzm37w6HwtWbcVediSQuziiF+sC7AteXhLHVjSlCQP3rm5lJbG0LG336tkvGirbMF4/2ZztdeV5USy6sx2TSmGc6CxBmNyIfj/IFowsDaL+uO94fQG+Wr3NcEaSe+wCiToqNLkMALDB1iJknkcmGsjabUEEiPVxu+c790SiEw6lLt+DyakuMWkKuU5ZQSfr0w9MDeGVORpRzJlhaFvuKgRRIofLeSBMrrAG4oGjgah5FO8vwsQPFhnOCIZpoyMpQOTkHGT/dMmmHf4mLOmATt+0yDU+jQBx+x7cAsWvf9AdG/0IKkBuHZ+6va30yUThRA8zmikfiYQSNmExucICxAPdft+Ad1hsEEyrdykw5B4UmRAkYktOvjEiPPyDgx1TWjwCDcR9OKwoo6Crfbeg4cijJO5wcYbJFjZheZBpqfKgZBKKGup9FT/AUXt1xsh+VjVb9wpaRmupze4J2W/yMTmWg66wJ56n37gpqAbiHl6yVH6RTpp59IG4tw1gmGxhAeJB0oQV7q/bveLPF9Ks5LZ9mxzkaiKeWyD4h9u6M9Ez49ghXbXtQSOe3O9xPvMsssGphZWH8anb20bJAb3aRXr9QtK7YxUnN4JNWJ44TvSQrxsjwhVHDcBf360N9bq/O21fnDK0O8be9T6+3ZRebd/RQFztJnNPq/ISbLWjydymI38NJJrJKagGkibAZDJl6CNqeeTLYf/shFEZ71jZUnjfVQttd4U1EA/q7ZIlKwPsl5EJMYrmx9u+qgyVZXGceoBVN+jE/bqlHFd9ICntrudyk6gUDcRldjONX87bYYS26ggqmNyra1l1t19NK133rPj+yF6hXSuvAs7ZmyXau5WVxDx3l2RaPgURIET0RyJaQERziOhFImqnHLuOiGqJ6EsiOk5pH2e31RLRtfkYZ31EqyeiaH66Mttazp1uwaD6QFRMk32nVsmEQ7fpyG8eT9cAoL13ENRz/ATTwM56AdGzQxUe/sHBxjLt2fCbU/YNZbdAIL8mtnyay5hdm0JpIJMA7COE2A/AVwCuAwAiGgzgbABDAIwD8DciihNRHFYZ+eMBDAZwjt03UqJyoseIIrE/y/BbeW33ZGvSQNzPDx/YCVcePQC/P30/py3d8W8qVaK/t+6sbN4BPwHy7KWj8PyPRmmPHbVXZ7SuKNUeKxbyEcbL5jwmLAqiXwoh3lKeTgVwhv14PIBnhBD1AJYQUS2A4faxWiHEYgAgomfsvvOjHGd0AiSaoJukgLCeuycIJw/EcJ6kNE742bGDnA2zAJ0PxHssQcq5B30P1BgGPxNWh+oydKjOrJpuMZDPUibFHlDAtByKwQdyEYDX7cfdASxTji2320ztkbCjoQl/nbzQs0R6LsRiFEk4qTQzOddOM1XFtO3uMN64RovINIzXmFyX4+uOyrdSaApiwmIdhMmRyAQIEb1NRHM1f+OVPjfA2iL3SdmkuZTwaNfddwIRzSSimWvXrtV18aWuMYE73/oK05ds8O+cBUTRmA/SNZDkXb43opezendP/mnahaZfWqSWST7I/q4O3x/ZG8P7dsB5I3un9c2EMAoXFiPHDukCABjVv2Pk92INhAmLyExYQoixXseJ6AIAJwEYo5SJXw6gp9KtB4AV9mNTu/u+E2Fvtzts2LCsLANysqxrbEZpnLBfj3aY9XV42kiMCAkK32iRLFWSPjMcN6QrHp/6tXXcdczkbFeFhp/fxI1bA+nUqhzPXar3TWRCvEBJmFEzsl/H0BzyfgTdHZJh/ChUFNY4ANcAOFkIoe7E9AqAs4monIj6AhgIYDqAGQAGElFfIiqD5Wh/JarxyQm0vimBklgs9JwGK4w3/InQ7SR3Ry+ZNJBDBqSueuXkrwqWTIcbpH8201hU+SW7E/lKJGR2fQplD/grgNYAJhHRbCJ6AACEEPMAPAfLOf4GgMuFEM1CiCYAVwB4E8AXAJ6z+0aCGnFUEiffuk+ZYkVhhXpJAFAEhPVcvUU8lty61X3vq8YMxDMTRmLvbm2svpotT1Whs1fX1r4FBKMqGbKr+kAKAb+TTK4UKgprgMex2wHcrml/DcBrUY5LotrZy+Kx0GtXWXkg4f985aStm7xLFA3EvfIkIozs19EpmKibpNVrvnH1EVi5OT3T3bqW/T/AeLMK4+VVc2jwW8nkyq7pkcwRdf4siVPojtsYRVPIzl2u3W3CStbC8r6OXoCkPvcTgFFpIGGWTt9dCbu2G7P7wgJEAxE5q/GSWMx37wsvOrcuT2uLyoSV5gNB8j6WAJH3D3YdFbfW4neNIBM9T2OFIRnGyzC5wQLEgJxES+OUtuVrNtdRsTSQKPNAkm2ldluKD8QwdXjtE5H2MoxhvOR1mCkC2InOhAULEANyMi6Jx1Bqz57H2bH6maCbjDP54ZqcxsP7pGdbx10+jlgs2VYSiyXbfW6vi3QKHMZLPsfTuzJ5Jp/b5zK7NixADJQ4JixyHrujsUybHqno3CeZ1MIqM2g/t54yJK3NHYUFkJLTEXwvbJ35yS0Q/EYfhqvivV+Mdh4fOzhz4c3ocVCnZNQAABZ4SURBVDSQwg6D2QXgWssGSmJJoVESd5UIsQlSGlwXNZSJE71NZQl2NqZvraszQ8WTEiLZ5kRWxXy1Arky1Y3Z3WS6lmzN1geiOnjVUuD3f/+gXXZviXzjvMOsgjA5wgLEgGP6iZNjwnLPiUFyEkyr+aA/3TYVpVi9pT6tXffbd28YRZRsK1Hqb/luBqUJGnC/Dt9M9BAmJ/eWuvFYPOdrRs0nvzqm6Ff2rIEwYcECxID0gZTGYk74q99GTESplWN1fWS/oCasNpX68uO6s6VAU6NsnMisGGnDe1XCcKI7hwO8PH8zWMub4jpUl/l3KjjsA2HCgX0gBhwfSDwZ0psWyup693QZ66akvKAmLNOObjoB5LSJ5AQhBWEiIRQNRH8vaSprpblnuhNdfw1ZDDCMuaklCpCWQFID4feXyQ0WIAaSJqxkJrqfCatUM6vqJvpYDIGXf9VlerON1+mqEiTH3pwQvk50uf9526p0rcd9iu4an910rHFHwGwg/nZGAlfjZcKCf6IGSh0TVjITPW0rWFdDaYlOA0m/tpcP5Nrj90p5XlVm0EAM56f2IVx/wt5oXVGCrm0rtEUWVbbstDaQal+VboYJooGobWFMTqyBRIMTxlvgcTAtHxYgBlKc6M4qPrWP27EsTVgpGzEZTE2myfFHR/ZPeV5l1EDMP391hXnckK74/ObjUFEad84xia96ewfGdhq/S5A8kDAS01TtiauWRANrIExYsAAxoOZ+yDDehMtD7n4uczYIwMWH9QVgisIK/uM1buzncY6p1pG8lt/ErHPcB5nMw57wWQOJBt6RkAkLFiAGUvJAYkk/gor7uRr+2qdTNQBTRFNwJ3rPDlUAgMtGp2om2UQ5+ZmwRg+qARC0FpamDygtCi0XWH5Eg+BiWExIcBivgRKlhpQ0TTUrs+OIvh3QpXWF65zkL1I+MiUSBl39dWxV5uxUd9J+e+CEez60r+9hwpJjMPgtTCv7iecN0yYtqudKdJdQ28JY3bIGEi387jK5whqIAbkKj1OylElC0TjuOecAow9ERVfKhIgC/3rVSbSvrdVY1zCfkzBoAUkfiJ6ykhjaGvJOgvlAzGPy4sHzDsKEI/r53pMJB97SlgkL1kAMOKG7sWRElmqy0s1tZdooLJMJK9jkqGoRKSt8rzBeoU8Uc9JEAt05lSC1sLLVOgZ0boVVm+s098zqcowfXI2XCQnWQAyo+4c7GohiwtJNlikmLA9zUSy4ApIyicZShIn/FdxjzMWsZBJGKjEKd48PnuCigV0gTFgURIAQ0W1ENMfeD/0tItrDbiciuoeIau3jByrnXEBEC+2/C6IeY1zxgcgoLFUD0a2OdSYsbSJhBhtKxUwaSLDTXWOx/mfj6E7TQHzCeHnuL15MGirDZEqhNJA/CiH2E0IMBfBfAL+2248HMND+mwDgfgAgog4AbgIwAsBwADcRUfsoB6hqIDLDvFmZeHUTqNaEpbf1BDZhmTUQ8zmmOH/5NBMbeKdW5WhXVRrInBTVfHSIXR6FCYdkkEVBh8HsAhTEByKE2KI8rUbyOz0ewGPCWiJNJaJ2RNQNwGgAk4QQGwCAiCYBGAfg6ajGGE8xYSXrSUl0E6puIyZtSCyC/3hVQaVeKogW4b5FNhrItOvHaK/lRkaKZcI1x++Fnz47G93bVRr7fHTNUehYnb4tMJM9XAuLCYuCOdGJ6HYA5wPYDOAou7k7gGVKt+V2m6ldd90JsLQX9OrVK+vxyQireCy1npRzH82PTxuFZTD1BLXvm/we7iRGFdOOc9n4FJwKvwGlTibC6bghXTH/1nGefXq0rwp+QSYQrIEwYRGZCYuI3iaiuZq/8QAghLhBCNETwJMArpCnaS4lPNrTG4WYKIQYJoQYVlNTk/X4pekpFqNkFJbqRNe8c0Gr8QLZOdGD4jeJZ+Po9hI+OtMdU7wEXQwwjB+RaSBCiLEBuz4F4FVYPo7lAHoqx3oAWGG3j3a1v5fzID2I6aKwUjSQdEoDbMTknB/UiW4432sOMCUSRoWpYjBTnOT7+8HsuhQqCmug8vRkAAvsx68AON+OxhoJYLMQYiWANwEcS0Ttbef5sXZbZMgM8jgpxRSF6gPRhPHaGkinVkmbvbaYouF8HaZ+2fhAMjk3E0wVg5kihXckZEKiUL/8O4hoEIAEgK8B/Mhufw3ACQBqAewAcCEACCE2ENFtAGbY/W6VDvWoUHfyk5N4IkAi4R/P2A8j+3XEBwvXWucr/Ub164gpi9db5wcch8mEJf0cMUrPPDcJiKgWnKaKwUxxYvKRMUymFCoK63RDuwBwueHYQwAeinJcKlK9j1FSmKgTtRQqowfV4L0vLWERJ8J3h1kWOOlkV01QalmTbPJAVKSQiMcIieZUiZGcIAznhlzK4hKlDMk+3dsAsLLLmeKEo7CYsGDvpw9xpeyIu/ouADxy4XCcZQsNncM81z3RTd06t7HMZJcfNSDt3uY8kGgmjDOHJd1Wpx7QHW//7AiMHtQZD194MF6/6vCMrsX+3ejh/UCYsGDjtQ8pJiyDD0RqFloBknMUlr5nVVkJlt5xIlZvqcOf316YkoNy7vBemLp4PS46tK/+ohFO0kSEAZ1bAwCOGtQ5uhsxWXPN8Xth444GHLln9lGKDAOwADGi1rKSQsBcTFHmjOi1DYlqOsgmD8RrnGoIcfvqMjx+8Qhj32Jc5PNqOH/07VSNZy8dVehhMLsAbMLyIR5L7mtuisKSD03ahhui4Pkdfv3kOEq0NVNc9w12S09qWnNWOMMwFqyBGJCTrVp63ZQHIh/r/B3G62dRykSHNKuV6DYeMZBtItmrVx6GLm0q/DsyDLNbwALEgGrCKi+1wlTVzZZ0e3OYkv4k6rwd1ITlp9U02RFYuiRGN7maiYbs0Tbl+aAurXH8vl1zuyjDMC0WFiA+xGOE7u0qcdv4IThmcFeM/N07AFIFgBNSq5mhdYt9QnilTNpXlQEAfuzaM92LsCKd3vzpEeFciGGYFgkLEB/kBH7eqD7GPtKypfoh/Fb7YTnRK8vigSvhtoS4f95ulWFaDuxE9yHIdCZ9Cupkf9yQruhXU52SZKdOjkGd6FFEJxXjFF38oo1hGDesgRjIZLUuHdlqMd4O1WWY/PPROV9fp4H07GDeP8MLDpVlGCZMWICEgDRhBS2QCMo+kXDa9WNyrj3F2d4Mw4QBCxAfgky2yVBab7GgRlQFr4WV+pzDaBmGKRbYB2IgE3OPWtjQizu/uz9+cEgfjOjbMbAJK4o9G9hRzTBMGLAGEgJSA/HLA+nWthI3nzzE6htSJnomSGHEJiyGYcKANRAfgmRtJzzyQEyEFcabCexDZxgmTFiAGMhksk1GYYU/RUdxTYZhmDBgARICIkIBEqYLhMN4GYYJk4IKECL6BREJIupkPyciuoeIaoloDhEdqPS9gIgW2n8X5GFsAIIl3SUS1v9MBEjQgoZhmrAyvTfDMIwXBXOiE1FPAMcA+EZpPh7AQPtvBID7AYwgog4AbgIwDNacPouIXhFCbIx6nEHm2mZNJrofms0NHTq1Kse6bfUZX9MPVkAYhgmTQmogdwP4JVIX+eMBPCYspgJoR0TdABwHYJIQYoMtNCYBGJf3ERsQAfNAUs7x0G2mXHe08zgKFwjrHwzDhEFBBAgRnQzgWyHEZ65D3QEsU54vt9tM7ZETyIQlM9Ezme09LqzuLhhmHkgUOSUMw+y+RGbCIqK3Aeg2i7gBwPUAjtWdpmkTHu26+04AMAEAevXqFWisueI40Q0T9JtXH5G2X4eXCUslEg2EVRCGYUIgMgEihBirayeifQH0BfCZvSLuAeATIhoOS7PoqXTvAWCF3T7a1f6e4b4TAUwEgGHDhmU9VWayWG+WeSCGTZ0GdW2d1hY0GzxUH0gLUEBYuDFMyyHvJiwhxOdCiM5CiD5CiD6whMOBQohVAF4BcL4djTUSwGYhxEoAbwI4lojaE1F7WNrLm/keuwk/DUR/TrB+kURhsReEYZgQKLZSJq8BOAFALYAdAC4EACHEBiK6DcAMu9+tQogN+RhQsEx0q08mc30iaBhviCJeDq8oV/ktQT1iGCaFggsQWwuRjwWAyw39HgLwUJ6GhYP7dMDDHy9N2wdch5yQM9EWgs7hYWogNa3LAQB7tMtuPxGGYRgVzkQ3cMK+3TD9hjEY1b+jb9/mROYaSCESCY8b0hUTzzsIPzoy+P7p+eK4wV1Q07ocPzikT6GHwjBMQAqugRQznVsH23vDKeeehQ+kojSGusaEsV+4pUwIxw7RBcYVns5tKjDjBm3cBcMwRQprICEQtJy7ihQgh/TvhNvGDzH2i8KJzjAMEwYsQELAcaJncU6MgPNG9TH242K8DMMUK2zCCgGZFJhJpvcRe9bg2MFdcOOJgz37sQbCMEyxwgIkQyYc0Q+fLduU0pZNOfeK0jgmnj/Mtx/LD4ZhihUWIBly/Ql7p7U5tbAimOy5fhXDMMUK+0BCIOkDCW+yH7t3l9CuxTAMEwWsgYRA0gcS3jXv+94B2LyzMbwLMgzDhAwLkBAIa0vbspKkQlheEkfn1vGcrscwDBMlLEBCIJHFjoRuHr1oOPp1qg5rSAzDMJHDAiQEZCmTXBSQI/esCWk0DMMw+YGd6CFQUWqZmjLakZBhGKaFwxpICDx43kF4Yda3bIJiGGa3ggVICPRoX4Wrxg4s9DAYhmHyCpuwGIZhmKxgAcIwDMNkBQsQhmEYJisKIkCI6GYi+paIZtt/JyjHriOiWiL6koiOU9rH2W21RHRtIcbNMAzDJCmkE/1uIcSdagMRDQZwNoAhAPYA8DYR7Wkfvg/AMQCWA5hBRK8IIebnc8AMwzBMkmKLwhoP4BkhRD2AJURUC2C4faxWCLEYAIjoGbsvCxCGYZgCUUgfyBVENIeIHiKi9nZbdwDLlD7L7TZTexpENIGIZhLRzLVr10YxboZhGAYRChAiepuI5mr+xgO4H0B/AEMBrATwJ3ma5lLCoz29UYiJQohhQohhNTVcHoRhGCYqIjNhCSHGBulHRH8H8F/76XIAPZXDPQCssB+b2o3MmjVrHRF9HWQcBjoBWJfD+VHD48sNHl/uFPsYeXzZ0TtIp4L4QIiomxBipf30VABz7cevAHiKiO6C5UQfCGA6LA1kIBH1BfAtLEf7uX73EULkpIIQ0UwhhP++swWCx5cbPL7cKfYx8viipVBO9D8Q0VBYZqilAC4FACHEPCJ6DpZzvAnA5UKIZgAgoisAvAkgDuAhIcS8QgycYRiGsSiIABFCnOdx7HYAt2vaXwPwWpTjYhiGYYLDmejeTCz0AHzg8eUGjy93in2MPL4IIbkdK8MwDMNkAmsgDMMwTFawANFQLHW37CTLNUQ0V2nrQESTiGih/b+93U5EdI895jlEdGDEY+tJRO8S0RdENI+Iriqm8dn3rCCi6UT0mT3GW+z2vkQ0zR7js0RUZreX289r7eN9oh6jfd84EX1KRP8ttvER0VIi+tyuWTfTbiumz7gdET1PRAvs7+KoYhkfEQ2iZL2/2US0hYiuLpbxhYIQgv+UP1hRXosA9ANQBuAzAIMLNJYjABwIYK7S9gcA19qPrwXwe/vxCQBehxXyPBLAtIjH1g3Agfbj1gC+AjC4WMZn35MAtLIflwKYZt/7OQBn2+0PAPix/fgyAA/Yj88G8GyePuefAXgKwH/t50UzPlhRkp1cbcX0GT8K4If24zIA7YppfMo44wBWwcqvKLrxZf26Cj2AYvsDMArAm8rz6wBcV8Dx9HEJkC8BdLMfdwPwpf34QQDn6PrlaZwvwyp2WazjqwLwCYARsBK3StyfN6ww8VH24xK7H0U8rh4A3gFwNKyEWiqy8ekESFF8xgDaAFjifg+KZXyuMR0L4ONiHV+2f2zCSidw3a0C0UXYSZj2/852e8HGbZtSDoC1wi+q8dnmodkA1gCYBEu73CSEaNKMwxmjfXwzgI4RD/HPAH4JIGE/71hk4xMA3iKiWUQ0wW4rls+4H4C1AB62TYD/IKLqIhqfytkAnrYfF+P4soIFSDqB624VGQUZNxG1AvACgKuFEFu8umraIh+fEKJZCDEU1kp/OIC9PcaR1zES0UkA1gghZqnNHmMoxHt4qBDiQADHA7iciI7w6Jvv8ZXAMvHeL4Q4AMB2WCYhE4X6jZQBOBnAv/y6atqKeu5hAZKOVz2uYmA1EXUDrJIwsFbWQAHGTUSlsITHk0KIfxfb+FSEEJsAvAfLttyOiGQSrToOZ4z28bYANkQ4rEMBnExESwE8A8uM9eciGh+EECvs/2sAvAhLCBfLZ7wcwHIhxDT7+fOwBEqxjE9yPIBPhBCr7efFNr6sYQGSzgzYdbfslcPZsGp0FQuvALjAfnwBLN+DbD/fjuQYCWCzSNYbCx0iIgD/BPCFEOKuYhufPcYaImpnP64EMBbAFwDeBXCGYYxy7GcAmCxsY3QUCCGuE0L0EEL0gfU9myyE+F6xjI+IqomotXwMy44/F0XyGQshVgFYRkSD7KYxsMogFcX4FM5B0nwlx1FM48ueQjthivEPVjTEV7Ds5TcUcBxPwyp33whrdXIxLJv3OwAW2v872H0J1q6NiwB8DmBYxGM7DJZ6PQfAbPvvhGIZn33P/QB8ao9xLoBf2+39YBXprIVlVii32yvs57X28X55/KxHIxmFVRTjs8fxmf03T/4WiuwzHgpgpv0ZvwSgfZGNrwrAegBtlbaiGV+uf5yJzjAMw2QFm7AYhmGYrGABwjAMw2QFCxCGYRgmK1iAMAzDMFnBAoRhGIbJChYgDGOAiJpd1VQ9KzMT0Y+I6PwQ7ruUiDplcd5xRHQzEbUnIt69k4mcQu2JzjAtgZ3CKoMSCCHEA1EOJgCHw0pCPALAxwUeC7MbwAKEYTLELj3yLICj7KZzhRC1RHQzgG1CiDuJ6EoAPwLQBGC+EOJsIuoA4CFYCXo7AEwQQswhoo6wkkZrYCUIknKv7wO4Elap8mkALhNCNLvGcxasqtH9AIwH0AXAFiIaIYQ4OYr3gGEANmExjBeVLhPWWcqxLUKI4QD+Cqt+lZtrARwghNgPliABgFsAfGq3XQ/gMbv9JgAfCasg4CsAegEAEe0N4CxYBQ2HAmgG8D33jYQQzyK5b8y+sLLuD2DhwUQNayAMY8bLhPW08v9uzfE5AJ4kopdgldgArPIvpwOAEGIyEXUkorawTE6n2e2vEtFGu/8YAAcBmGGVHkMlkoX33AyEVQIDAKqEEFsDvD6GyQkWIAyTHcLwWHIiLMFwMoBfEdEQeJfr1l2DADwqhLjOayBkbTXbCUAJEc0H0M3eA+UnQogPvV8Gw2QPm7AYJjvOUv5PUQ8QUQxATyHEu7A2i2oHoBWAD2CboIhoNIB1wtpDRW0/HlZBQMAqtHcGEXW2j3Ugot7ugQghhgF4FZb/4w+wih4OZeHBRA1rIAxjptJeyUveEELIUN5yIpoGaxF2juu8OIAnbPMUAbhbCLHJdrI/TERzYDnRZUnvWwA8TUSfAHgfwDcAIISYT0Q3wtoRMAarKvPlAL7WjPVAWM72ywDcpTnOMKHD1XgZJkPsKKxhQoh1hR4LwxQSNmExDMMwWcEaCMMwDJMVrIEwDMMwWcEChGEYhskKFiAMwzBMVrAAYRiGYbKCBQjDMAyTFSxAGIZhmKz4fwMKtGu/r6/sAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best 100-episode performance: 200.2989085178099 +/- 59.337490722745\n"
     ]
    }
   ],
   "source": [
    "with open(os.path.join(\".\", \"configs\", \"Lunar_Lander_v2.json\"), \"r\") as read_file:\n",
    "    config = json.load(read_file)\n",
    "    \n",
    "env = gym.make(config[\"general\"][\"env_name\"])\n",
    "env.seed(config[\"general\"][\"seed\"])\n",
    "agent = Agent(config=config)\n",
    "\n",
    "if config[\"train\"][\"run_training\"]:\n",
    "    scores, best_mean, best_std = sessions.train(agent, env, config)\n",
    "    helper.plot_scores(scores)\n",
    "    print(\"Best 100-episode performance: {} +/- {}\".format(best_mean, best_std))\n",
    "    agent.save()\n",
    "else:\n",
    "    agent.load()\n",
    "    score = sessions.test(agent, env, config)\n",
    "    print(\"Episode score: {}\".format(score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Environment: Acrobot-v1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Acrobot-v1 is an unsolved environment\n",
    "    - it does not have a specified reward threshold at which it's considered solved"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[33mWARN: gym.spaces.Box autodetected dtype as <class 'numpy.float32'>. Please provide explicit dtype.\u001b[0m\n",
      "Episode 100\tAverage Score: -487.30\n",
      "Episode 200\tAverage Score: -343.34\n",
      "Episode 300\tAverage Score: -201.39\n",
      "Episode 400\tAverage Score: -178.74\n",
      "Episode 500\tAverage Score: -126.87\n",
      "Episode 600\tAverage Score: -117.44\n",
      "Episode 700\tAverage Score: -101.06\n",
      "Episode 800\tAverage Score: -104.88\n",
      "Episode 900\tAverage Score: -98.000\n",
      "Episode 1000\tAverage Score: -92.94\n",
      "Episode 1100\tAverage Score: -92.24\n",
      "Episode 1200\tAverage Score: -85.74\n",
      "Episode 1300\tAverage Score: -80.71\n",
      "Episode 1400\tAverage Score: -80.63\n",
      "Episode 1500\tAverage Score: -80.23\n",
      "Episode 1600\tAverage Score: -78.84\n",
      "Episode 1700\tAverage Score: -77.44\n",
      "Episode 1800\tAverage Score: -74.68\n",
      "Episode 1900\tAverage Score: -75.61\n",
      "Episode 2000\tAverage Score: -72.52\n",
      "Episode 2100\tAverage Score: -74.84\n",
      "Episode 2200\tAverage Score: -75.34\n",
      "Episode 2300\tAverage Score: -76.96\n",
      "Episode 2400\tAverage Score: -75.15\n",
      "Episode 2500\tAverage Score: -72.47\n",
      "Episode 2600\tAverage Score: -73.20\n",
      "Episode 2700\tAverage Score: -72.83\n",
      "Episode 2800\tAverage Score: -69.09\n",
      "Episode 2900\tAverage Score: -69.94\n",
      "Episode 3000\tAverage Score: -71.69\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best 100-episode performance: -67.08 +/- 7.4130695396711355\n"
     ]
    }
   ],
   "source": [
    "with open(os.path.join(\".\", \"configs\", \"Acrobot_v1.json\"), \"r\") as read_file:\n",
    "    config = json.load(read_file)\n",
    "    \n",
    "env = gym.make(config[\"general\"][\"env_name\"])\n",
    "env.seed(config[\"general\"][\"seed\"])\n",
    "agent = Agent(config=config)\n",
    "\n",
    "if config[\"train\"][\"run_training\"]:\n",
    "    scores, best_mean, best_std = sessions.train(agent, env, config)\n",
    "    helper.plot_scores(scores)\n",
    "    print(\"Best 100-episode performance: {} +/- {}\".format(best_mean, best_std))\n",
    "    agent.save()\n",
    "else:\n",
    "    agent.load()\n",
    "    score = sessions.test(agent, env, config)\n",
    "    print(\"Episode score: {}\".format(score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Environment: CartPole_v1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- CartPole-v0 defines \"solving\" as getting average reward of 195.0 over 100 consecutive trials."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[33mWARN: gym.spaces.Box autodetected dtype as <class 'numpy.float32'>. Please provide explicit dtype.\u001b[0m\n",
      "Episode 100\tAverage Score: 21.19\n",
      "Episode 200\tAverage Score: 15.60\n",
      "Episode 300\tAverage Score: 12.51\n",
      "Episode 400\tAverage Score: 14.74\n",
      "Episode 500\tAverage Score: 83.52\n",
      "Episode 600\tAverage Score: 193.96\n",
      "Episode 622\tAverage Score: 195.04\n",
      "CartPole-v1 Environment solved in 522 episodes!\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best 100-episode performance: 195.04 +/- 7.854832907197962\n"
     ]
    }
   ],
   "source": [
    "with open(os.path.join(\".\", \"configs\", \"CartPole_v1.json\"), \"r\") as read_file:\n",
    "    config = json.load(read_file)\n",
    "    \n",
    "env = gym.make(config[\"general\"][\"env_name\"])\n",
    "env.seed(config[\"general\"][\"seed\"])\n",
    "agent = Agent(config=config)\n",
    "\n",
    "if config[\"train\"][\"run_training\"]:\n",
    "    scores, best_mean, best_std = sessions.train(agent, env, config)\n",
    "    helper.plot_scores(scores)\n",
    "    print(\"Best 100-episode performance: {} +/- {}\".format(best_mean, best_std))\n",
    "    agent.save()\n",
    "else:\n",
    "    agent.load()\n",
    "    score = sessions.test(agent, env, config)\n",
    "    print(\"Episode score: {}\".format(score))"
   ]
  }
 ],
 "metadata": {
  "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}