assignments/P1-Intro/L8-Lab-Taxi/lab-taxi.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Solve OpenAI Gym's Taxi-v2 Task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Python 3.6.9 |Anaconda, Inc.| (default, Jul 30 2019, 19:07:31) \n",
      "[GCC 7.3.0]\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "print(\"Python %s\" % sys.version)\n",
    "import time, copy, math\n",
    "from collections import defaultdict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OpenAI Gym 0.14.0\n"
     ]
    }
   ],
   "source": [
    "import gym\n",
    "print(\"OpenAI Gym %s\" % gym.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NumPy 1.16.4\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "print(\"NumPy %s\" % np.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "matplotlib 2.2.2\n"
     ]
    }
   ],
   "source": [
    "import matplotlib\n",
    "print(\"matplotlib %s\" % matplotlib.__version__)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from monitor import interact"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use the code cell below to create an instance of the [Taxi](https://github.com/openai/gym/blob/master/gym/envs/toy_text/taxi.py) environment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "env = gym.make('Taxi-v2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<TimeLimit<TaxiEnv<Taxi-v2>>>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "env"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Agent:\n",
    "\n",
    "    def __init__(self, nA=6, alpha=0.01, gamma=1.0, epsilon=0.005):\n",
    "        \"\"\" Initialize agent.\n",
    "\n",
    "        Params\n",
    "        ======\n",
    "        - nA: number of actions available to the agent\n",
    "        \"\"\"\n",
    "        self.nA = nA\n",
    "        self.Q = defaultdict(lambda: np.zeros(self.nA))\n",
    "        self.alpha = alpha\n",
    "        self.gamma = gamma\n",
    "        self.epsilon = epsilon\n",
    "        \n",
    "    def _generate_policy_from_epsilon_greedy(self, state):\n",
    "        # else\n",
    "        policy = np.ones(self.nA) * (self.epsilon / self.nA)\n",
    "        # if a maximizes Q(s,a)\n",
    "        max_action = np.argmax(self.Q[state])\n",
    "        policy[max_action] += (1 - self.epsilon)\n",
    "        return policy\n",
    "    \n",
    "    def _generate_action_from_policy(self, policy):\n",
    "        max_q = float(sum(policy))\n",
    "        probs = [q / max_q for q in policy]\n",
    "        action = np.random.choice(np.arange(self.nA), p=probs)\n",
    "        return action\n",
    "\n",
    "    def select_action(self, state):\n",
    "        \"\"\" Given the state, select an action.\n",
    "\n",
    "        Params\n",
    "        ======\n",
    "        - state: the current state of the environment\n",
    "\n",
    "        Returns\n",
    "        =======\n",
    "        - action: an integer, compatible with the task's action space\n",
    "        \"\"\"\n",
    "        policy = self._generate_policy_from_epsilon_greedy(state)\n",
    "        action = self._generate_action_from_policy(policy)\n",
    "        return action\n",
    "\n",
    "    def step(self, state, action, reward, next_state, done):\n",
    "        \"\"\" Update the agent's knowledge, using the most recently sampled tuple.\n",
    "\n",
    "        Params\n",
    "        ======\n",
    "        - state: the previous state of the environment\n",
    "        - action: the agent's previous choice of action\n",
    "        - reward: last reward received\n",
    "        - next_state: the current state of the environment\n",
    "        - done: whether the episode is complete (True or False)\n",
    "        \"\"\"\n",
    "        self.Q[state][action] += self.alpha * (reward +\n",
    "                                               (self.gamma * max(self.Q[next_state])) - \n",
    "                                               self.Q[state][action])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perform Exploration"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nomad agent with high learning rate to perform exploration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Episode 2500/2500 || Best average reward 8.8732\n",
      "\n",
      "\n",
      "\n",
      "Runtime 5 seconds\n"
     ]
    }
   ],
   "source": [
    "agent_nomad = Agent()\n",
    "\n",
    "agent_nomad.alpha = 0.1\n",
    "agent_nomad.gamma = 1.0\n",
    "agent_nomad.epsilon = 0.005\n",
    "num_episodes = 2500\n",
    "tic = time.time()\n",
    "avg_rewards, best_avg_reward = interact(env, agent_nomad, num_episodes=num_episodes)\n",
    "toc = time.time()\n",
    "print('\\n\\nRuntime %d seconds' % int(toc-tic))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the above until last average reward bigger than 8.0."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.36"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "avg_rewards[-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'Average Rewards during Episode')"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(avg_rewards, linewidth=1)\n",
    "plt.xlabel('Episodes')\n",
    "plt.ylabel('Average Rewards during Episode')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perform Exploitation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Settler agent with lower learning rate to perform exploitation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Episode 5000/5000 || Best average reward 9.27\n",
      "\n",
      "\n",
      "\n",
      "Runtime 5 seconds\n"
     ]
    }
   ],
   "source": [
    "agent_settler = copy.deepcopy(agent_nomad)\n",
    "\n",
    "agent_settler.alpha = 0.001\n",
    "agent_settler.gamma = 1.0\n",
    "agent_settler.epsilon = 0.00005\n",
    "num_episodes = 5000\n",
    "tic = time.time()\n",
    "avg_rewards, best_avg_reward = interact(env, agent_settler, num_episodes=num_episodes)\n",
    "toc = time.time()\n",
    "print('\\n\\nRuntime %d seconds' % int(toc-tic))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'Average Rewards during Episode')"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJztnXd8FHX6xz/PllQINfQSkC4dREBBFBQV7AV7Octx59n1zl6v6OnpWX6nYu+9iwVREFAB6b1J74FAAum7+/z+mJLZ2dnd2WRnS/K8X6/A7uzszHeSme/zfToxMwRBEAQBAFzJHoAgCIKQOohQEARBEHREKAiCIAg6IhQEQRAEHREKgiAIgo4IBUEQBEFHhIIgCIKgI0JBEARB0BGhIAiCIOh4kj2AWGnZsiUXFBQkexiCIAhpxcKFC/cxc360/dJOKBQUFGDBggXJHoYgCEJaQURb7Own5iNBEARBR4SCIAiCoCNCQRAEQdARoSAIgiDoiFAQBEEQdEQoCIIgCDoiFARBEAQdEQqCkGIwM3YcLE/2MIQGiggFQUgxft24H8c88mOyhyE0UEQoCEKKcdGL85I9BKEBI0JBEFKIgjumJnsIQgNHhIIgpAh/+2hZsocgCCIUBCFVeH/BtmQPQRBEKAhCqsLMyR6C0AARoSAIKUq1X4SCkHhEKAhCilLtDyR7CEIDRISCIKQg2V43qnwiFITEI0JBEFKAQCDYVJSb6RZNQUgKIhQEIQV4cfbGoPeZHjeqRCgISUCEgiCkAF8t2wUA6NoyFwDgdZOYj4Sk4KhQIKIbiWgFEa0kopssPiciepqINhDRMiIa7OR4BCFV2Xe4EgDw15N74ppRXeB1uyT6SEgKjgkFIuoL4BoAwwAMADCRiLqbdjsFQHf151oAzzk1HkFIZXYVVwAATu7bFndP6IMMj0t8CkJScFJT6A1gLjOXMbMPwE8AzjLtcwaAN1hhLoCmRNTWwTEJQlqw42A59pdWJXsYQgPESaGwAsBoImpBRDkATgXQ0bRPewDG3P7t6jZBaDBsKyoL2XawrBq3f7g0CaMRGjoepw7MzKuJ6FEA3wM4DGApAJ9pN7L6qnkDEV0LxbyETp06xXmkgpBcPl+yAwBw+/ieQds1P4MgJBLbmgIR5cZ6cGZ+mZkHM/NoAEUA1pt22Y5g7aEDgJ0Wx5nCzEOZeWh+fn6swxCElGbD3sMAgOuO7xa0fWjn5skYjtDAiSoUiGgkEa0CsFp9P4CI/mfn4ETUSv2/E4CzAbxr2uULAJepUUjDARQz865YLkAQ0p3je7XCaQPaBW3743FdMaaXLICExGNHU3gSwHgA+wGAmZcCGG3z+B+rAuVLANcx8wEimkxEk9XPvwawEcAGAC8C+HMsgxcEp5n4zGwcLHPW4VvtZ3jdwZbULI8bldWpFX304YJteGq6WdkX6hu2fArMvI0o6Kb12/zeKIttzxteM4Dr7BxLEJLBih0l2FpUhqY5GY6do8oXQIY7eH2W6XWhuLzasXPWhsenrcWekkrcOM4cWS7UJ+xoCtuIaCQAJqIMIroNqilJEBoCZBkPET+mrdoNv6n2UWV1ADPW7HX0vLFiHqNQP7GjKUwG8BSUUNHtAKZBVvdCA8KoJM/buB/tmmajY/OcuB1/5tpC5DfODNr284Z9WLfncNzOEQ98IhQaBFE1BWbex8wXM3NrZm7FzJcw8/5EDE4Qkom2MjY2QJs0ZS5G/XtG3M/RODN4fXb3hN4hgiLZiKbQMAirKRDRM7DIGdBg5hscGZEgpAhaQbrqgHMO3982FwEAKqqD3XQV1QEUHkqtPAVfLWoxBQKMv09djftO6+PAiJzl2xW70aJRBo4qaFihwZE0hQUAFgLIAjAYSo7BegADYdPRLAjpTKVPuc3X7zkEwJmeybepWcunDQwOST2qoBkA4LjH4qeV1BWPK3bfSoXPj1d+3oQvl4akH6U8k99aiFs+WJLsYSScsEKBmV9n5tehFKs7npmfYeZnAIyFIhgaJFv2l6K8KlgmLtxShO0HQksVCLUnEGB9Mk4WWpXSv328HABQqWoOo7q3jNs5dh4sBwDceUrvoO0eNRppy/6a+2rt7iT/PmqhMV304jwAwIqdxfEeTkLwuhpedwE7V9wOQGPD+0bqtgZFcZkSHnjcYzPx+LS1QZ+d89yvOPbR1FnR2aGkojqlbcRfr9iFE5+cFbK9pKI6pEuZU4REBKlCIVKfg/fmb8XLczbZPofdS2FmjP/vLKzZXWL5+ez1hfj7V6tsn9cOlT6/ft8Dka87HEu2Haz1d1MBj9vZyLNUxI5QeATAYiJ6jYheA7AIwD8dHVWKsW7PIQx4aBpW7FBWO6kWP14b+j8wDc/+uCHs5+/M26qvYhPF1v1leHf+VgAI0cYqfX48+f069H9gGl79ZXNCxuMzrYwrVbt/pJLWD321Cg/HeXJWzqlIjxlrCi0/f2XOJrwUgzCyw+PfrcXJT9UI5gDXzoQEhLYbTRc8SdIUnpq+Hg9+uTIp57YTffQqgKMBfArgEwAjVLNSg+F3tTbNATWz9aOF2y33MzsLU50np6+z3L5pXynu+nQ5Xk/Q5Ksx+rEZuPMTxVSjTSHaZLJqZwme+kHJprWqKuoERk2hotpfoymEEQp//2oVygzCbMWOYuwpqYh4jvFHtsbVx3aJOhbtnIM6NbX8fI0DpqXVuw7pfR40Mjy1mySPaNUoHkNyjII7plo+16t2lWD2+kLdv5Qonpy+Dq/+vDmh59Sw+xceBmAUlPIWRzk3nNTklg8UZ2C0xc4XS9LPmWbFrmJFQ3hh1sYoe8YP40qSmfUHVJsMjeaHgMnhe97zv+Dr5fEvmWWMy1+1qwQlFYqGWO2zvhHmbgqO1J74zBxc/87iiOdwuwgDw0z0RrTrN6+4D5RWYe+hipDJO54s316sn9frrp1QSAfz0cowfo9LX56PS1+an+DRJA87BfEeAXAjgFXqzw1E9C+nB5ZKlKsaQLgaOK3UePKsDHfYY5RUVDsSveIEZZWJ13i2H6gxVc3dWIRu6sqyyh9AIMA4aDDZmW39v20+EFZ7qwvG85z9v1/w7I8bkO11h9UUDpTWjHHayt0AgMoo3dO+Xr7bll+hykJL2bK/FIMe/h6jHp2h93Z+7ef4mZDmbNgHQAmk0H7/tTUD/X3qasdrSJm58b3F2LDXvgblovCmsflq6HBDwI7YPxXAicz8CjO/AuBkABOcHVZqcuN7oeFpgQBjrxpPHsnc2v+BafgixcLyjusRWoWzvMqPq99YkPCxbDNEb5VX+9BDFQo/rN6D699bjD++uVD//O15W/Vy0xplVeZWHXXnuZm/o3Vepp5YdqjCh8ZZnrCr3h0GH8xdnypmsKWqozUSxTYmS+2cxlyBiU/PAaA4wDfuKwUAPPBl/P0ZvgCjuLwauRnuOuVsDHzo+ziOKjqfL9mJb1fstr1/KjmVs73hF5hOY1cXNOq3TZwYSLpSqk5GRxU0QzRFwLgadopFWw9EXZEt3HIAAJBpYR92YnK1w3XvLNJfr919WHes7jtUhanLQk1D4574Kej93I1FcXdmfrp4B/aUVKJAXYXP2bAPew9VYsfB8qha377D9lfF2RnRq81U+UOd3E1yvLbPURt6tVGCDv0BRpUvgNZ5WfrfxQ57LfwpifC7VVT78Yuq5cRitmpk+DuY/755WY71I7OEw+cNO44dofAv1EQfvQ4loa1BRR91aJYd9N5Y5lhzLLZqnBX1z1iZgAfi7P/9ggvV2PBw7Fbtz1ZOw0MVyREKBw2hj49+u0Y3k/zj6/C1F82mgZ9/31fncTAzxv5nJt7/bau+LdvCLLjTwoZ/2YjOlseMJqzCrQpn3jZGf12pZ1fXHOu8IebutgrmyC0zZVW+qJFlM9bs1Z3Xa/ccQpUvgOwMN/wBti18NUHfs3VNRHsihMIXS3biopeUZ+DpCBF2GpoA0HxCs9YVosudXwftU1LhS2ggSTItzXaij94FMBxK5JEWffSe0wNLJYwr6tZ5mUF/sEtfVm4+ovAZr8//9DsAoCJBzrbVu6xj2TWaqitMq9He/0VwGNyhiuSE34Zb4XU3RLGMe2JWUNjevZ+tqPN5F245gN8LS/Hxwh36NuOkPfm4I9AiN8NyYmya7cWE/m1DtvvD3Bfa/WKlsQE1QntXcXlNyQ1fAIcrfWDmsMe97aPIvZ3v+3wlRj7yY8R9Jr9VY677ZNEObNx3GBkeFzLcLtsmpKJSRVtqbFhlVySiR0SMViBN4P68QREKl71i7VR+e95Wy+1OkEzvox1H8zEASpj5CyhJbH8lIuslUT2lojqgawvZXjf8zLoTUqtkSRGcVI98swYA8Mavmx0dpy+KU1NDW/H4/YwZa/Zi2fYau/dP64Lj4Ps9MC1+A7TBVWp45uIwtvi87GCTiTFsz8q0wcyYMuv3iLkFRvQII8PEZ9SoDpRWoUm2V59IjPgCDK+FYymcX0Ebb+92eZafa+cd8a8fdaHwyeLt6Hv/d7jlg6X4JkzElZW5zciB0uimLfP1lVf5keF2weMm2zWQtGPcM7Gm7lF5AlbbP2+ITWM0alaaadVI2yZZAOBI/klYUllTAPAcgDIiGgDgdgBbALzh6KhShF3F5Si4YyoKD1WidZ56Y5zZF8w1JQea5yrNVwjWKp8xGsTpVdKyHfZKCVRUB+B1E3yBAK587Tdc/27ksEmno6aMq+7WeUok16x11kla5lX16YY2llceUxCyf1mVH//8eg3W2yxDrZkD9xt8AkZz4fsLtiHT67Y0JVRUB3Bku1CX21fqJD17fSG+WrYTBXdMBaD4B7K9brRvmh3yHQBokVvT2Me8mv108Q6s3xv+muZtDF/I+AeLPg13f7oc57/wa9jvVPsDyPC44HW7bAnYOev34aKjOwEAjsjPxZL7TgQQ3Xw0e31hre63wkOVenLp5zGGhhsF4IUvzg35vIvqU+rY3PrvFA+WbT+oa1ZAjU8hGZq6HaHgUzuknQHgaWZ+CsFlL+otmjmiyh/QVxDN1A5cl7+qqJjnDumASUM7KuYjC/H+8++JqzJ+rY2ooZfnbMJ17yxCtZ/1OHxN67lwSugDATjvIN9iSEYzmm2s0ISzRqMoDkBtZXr9u4si7qeR6VFMRVsNY3Ibslrvm9gHWV4XKn1++PwBrNhRjO0HylDp86O00odGWR6cPah90DFfU5MAL315Pv5iyFvwWbThNGLUPsOZNAAly7hRpgebH6kJCpwU5m8Zjp/WFWL+piJc/foCvDNvK84Y2A5d83ORq/pT1u05DGYlm9+O8/aSl+fh2xW7cfmIzmic5UXTnAwM7Ng0qqZw6cvzsXpX7Il4F0z5FROfmRPz9wAEJaZZXZuLCLee2AP920fPJ6ktpz/7M656/TcAyiJJezYHJThiC7AnFA4R0Z0ALgEwlYjcAJwNe0gRBnduFrJNC1vTyhpPmbUR7y/YFlZT2Hc4uPzx4UrnHLl2mqA8YajbZO4X8Ku6umyTl4Xbx/fU9zvrf7/EcZShaKUT/nZyL2R6w9+So7q3xF2n9sY5gzvo274yhPla5Q9oD/nvhaW2xlJUGlqu2mgS6tehCdbuPoR7PluJeZuKMPGZOTj20Rl4avp6lFb5kJvp0RO8PrvumJBxBI3NH6h1MpjGxP5t4Quwbmp65Ox+tTqO9v3pq/fgrk+X4/MlO3HhUZ1w84k9AABvzt2i5y38aLMj3LLtxUGO8WyvGxVRnOAAUJvKEtrf97PFkRcVRgIBxvkv/IpKXyCstgao0Vf+AKYu3+VoEp527AqfH1nq4mR83zaOnS8cdn79kwBUAriKmXdD6cD2mKOjShFaNgptcuJxuXD1sV1wyfBOehmInq0bg4gshYI5mmfLfnuTkx12HiwPyn3QJhhztJSRUsNDqdmGdxwsD4qx/+vJPXHd8d30jmNmwRZvtDyPUd1b4mTDQ9ClZS4+/fNI/X2W1438xpn4z/kD9G0lht+vVaaxle0/ElpFVCPG+PUsjxtlVX6s3lWCi1+qifL6ctlOlFb6kJvhhksVIgM71qwszWUSqnwB+AJ1Fwqjuyu5JlqP55U7a4IMtu6PXA7kns+W66Ygq8na66aQhK6OzbPRPsL9ZeadecFRXHZ8CpGSyKLx4mx7Wfi/bNiHRVsPYP6mIpRV+dE0x4sJ/UKDBABlsfTsDCWKKV6Lum9X7A7JtdGex/IqP7K8Lozq3hL7HX72rLATfbSbmZ9g5tnq+63MbMunQEQ3E9FKIlpBRO8SUZbp8yuIqJCIlqg/V9fuMpzBGEJqtF13bpED5ppIndvH91Q0BYtj2KlrY2bFjmJbzsD/Tl+HGwz+AC2u3I7v4twhHYIKvr2gRkgBNcLFXYeHMxb2Ha5E/w5N0Ld9k6CEusfPG4BOYdpeDugYrMoXtMixtHXH46Eymo/CrWK3FZWjtNKP3EwPjipohkamTmpm4fR74WFU+xheT+TfsdHJbRUr36ut8jfXBJdm/waAh76KXFDtrblb8dLsTZbjA5SyLuZ7ukerxlHvr3A+gWxvZKGgaa61ueuaqRF1lwwPjYGxGs9FL83Tw1a37C/F+r2H0aZJsGky1xCKfP0J3QEApVGEwsItRbZCVye/tVCPnNPyOdaqpeLLq/3I9roxe/0+zN2Y+EzqsEKBiOao/x8iohLz/9EOTETtAdwAYCgz9wXgBnCBxa7vM/NA9eelWl6HIxhv/qcuUFpI5Ga6kelxB6mROZluIExIqjnGPVJkSMEdU1FR7cfEZ+bg3s+jh1eazUX92jfBtaO72irelelx4bfNNZEWxtWZZud21bIiZqyUVfn0ycxYWqJFbgZaNMrUbeXGX+/n1x2DK0YW6O/H9GylC4VKnx8Fd0zFo9+uwb/UyK+6YLT7RyuFUF7tx9mDO2DFg+MBAGsePhkAcJyphafbRRj92AxsK4rsr1n391P018O6BHcAO7JdHvp3UISjNqzLDb+TkjA5J1oDH0DJCQGshUKLRhnwm8JPM72uqJPe1a/X+La65tcIqSyvO2IOhR52G0OC3C8b9mFXcTkOqHkubqKgntoeF8EXYBTcMTWk54l2vhvfW4IqXyCkAuz0W48DoAgzLRlucxRN/5znfkWve7/F3z5ahjd/3RxxX81npM0z43q3BqBWKHawllU0IjXZOVb9vzEz55n/t3l8D4BsIvIAyAGQWnUeLPjXN6uxfLsSxVBkyAwmIqx8cDzaNslG4eFKfGiotZOb4QGBLDUFs5z438zfLfaqQUviirYisTq2L8DIzfDgUIXPUtMwhkYWmT6ftb4m2scJTeFQRTWue9va2Vta6UeOmk2qlSoe0zM/ZOVmXrca49/zsr26T0GbeJ6b+bv+oNvJSLXqL/HsRYN0YXNCr1bo1DwH/zwrvN2+g8k2naXmOJSaS4HXIhLtgqM6Bb3/8i/H6q+1icXtIiy+V4n0ORxGKBgXAwDw9fJdlmaRRpkenNinxpx3+/ieyPK4o5rkjNFND53eV3+dFUWg6KU8YiilcdFL8zDiXzU5F3/9eFnQc2GMlrrkpchJnUEFEB9SnvX7T+uDJycNwJlq8MClL9srjPf+gm249/PImhozo7zKr9+301fvwedLdkSdI5zGlkGTiAYT0Q1EdD0RDbLzHWbeAeBxAFsB7AJQzMxWQe/nENEyIvqIiCxTNInoWiJaQEQLCgutQxXjxQs/bcQ787cAUCYVoxkgV309b1OwSpeb6VZWJxZSgcFo2SgzyL5shZZlPPxfPwAAZqyNfp1mzaTaH0BupjIJ3fR+aJ0m403fq02wXN9ocMRqyVrxjCnftK8UU8PE1W87UIYcVaPq0y4PM24bg9euHKZPqBpm+7tRC2uS7dUffuN1nje0I3q2bmyrodA9arSZNumfO6QDJvZvh/FHtsH5QzvglSuOQm6mBxcd3cmyzlVuhttCkFmzZFtoPHw43OrJjmwf/DczanLGfs5acmK0Va3Gny2E9cgjWuCEXq2CzFHXHd8tbDiuFZsfmYBjDV3qsr3uiKYnTcO1qynY+ZuWV/uxaIuyGIrWB2VMzxrTpbZIufKYLji5b1tcfHSNQI6ldHskH+Ls9fvQ+75v8dWymrWysb7aY+f2x5DOzfDN8l165eJEYCd57T4ArwNoAaAlgNeI6B4b32sGJYy1C5RObblEdIlpty8BFDBzfwDT1fOEwMxTmHkoMw/Nzw8t4hZvjAuVtk2y8NZVRwd9bl515mR4VJ+CVfIUcGy3FkGRKFbc9H7kXAHLcZo1BT/rE+cQi8gpf4DRo7WSEXxkmIQpIDRBLMPtwsy19iJOwrHfVAvIqM77/BxkOzdORBpfXX8s/mFaoV85sgvuP60Pnr1oENyk2MiZOci0d9uHS7F2zyGUVvmj1nXSYvuP7qqYabQpd0zPVvj3uQOC9rWaj0qr/LYdx9FWkUa0ya+NGo7bq01jvHft8LD7a2aJDI9Lz4kwHysaXVrmWiZkRlvtRyKao1nTQBZvPWBLU46UL7H5kQm62fESterAgbLIQmFU93xcO7qr5WfG38U/I5ReMXPcYzOj7vPf6estt7fKy8Ly7cX409uLMP7JWXoehtPYuYMvBHAUM9/PzPdDKXlxsY3vjQOwiZkLmbkaSomMkcYdmHk/M2tLnBcBDLE/9Pjz+Hc14ZraxJKT6Qla7QDA8K4tgt5ne91qmYvQYzJHj6a44d3Flg6laIXttL4C24rK8Povm/Hm3C16RrLVinXL/lI0yfbi7lN747ie4YVrj9bBaSiZHheuePW3OrXv1BqgGxPVLnlpHj5fsgOVPr+eNRqOvu2b6ImCGtkZblx5TBdM7N8O7/22DYBiRrEKG2yW441aE0irNHpEviI4a2M9iyQUerVpjBcvGxrzMafecCxeveIoEBEyPC4M6tQs5B408/CZfYPqSWms2hnVHQgguGfFmQPb6dpHpg3zUTiyvG7dmWqFdty/T11tK3t42fbgSdKcH6JhpzGQFmloJ+Q0XFmScHywYFtM+4/r3Uq/TzTTUkmFDxOfmZOQvhR2rm4zAOMTmwnAjtFrK4DhRJRDipgdCyBIxBKRMQbsdPPniUYLO/tw4TZ9FZJtETdvVCUBJYEqnE8hwKwvOc8fqsTXH670Bf1xw5XUjpbEoz23o/49A6+qmdMLtxzApKEdLSdwf4DRvmk2rhndNeLkpT1E71xzNK48pkDPHSitQwVVTXWv8gf03+3m/WX4eNEOVFQH9Ljs2qI9PGVVPst8BY/bZSuPo664IzjnOzbPwYl9Wsd8zCPbNcHxvVoBAJbdfxIePuPIkH2yTPepeeIqr/Jj8psLLYv7WWHUlh87bwDm3jlWP09tCztW+gIRAy2MARJ2TF+/GhJDG2d5gnxMRuxMpFrYtdExbmaEKohzMmOrmPq/GdGL8hmZvnovyqp8WLw11MRYkYAOcHaEQiWAlWqV1FcBrABwmIieJqKnw32JmecB+AhKT+fl6rmmENFDRHS6utsNasjqUiiRSlfU4VriRoBrVkpWlSjNarXX7QqvKQAgVSrcepKSENb3/u/Q455voo7jmxWRa9hYmav+e8FAZIZ5cCt9ATTOCjYNHdcjH9ef0A2AIrR+UCMuAGDkES1x/2lH6ppOXZrvjOmpTGrXvLEA362sqXFP6rgiJa3ZoV97pbzElFkbQ5y4Vx/bRY9CseLjhdst7c1kMzgyUlayEa1kgfb7BmrCiO2S5XXDYyHQC1oET2bG1XwgwNhVXI5vV+4GwDjCYuJrZirDbdQUvG6X7t/J8rprXdhRC1wIF7JqnLyj1Veq9geC2skeqvCFLRBohxvHKiGnFx/dOag6rZG3rj4aHZtnBxVlNGLXrGanRtmmfaWW/he7dafqgp0n8VMAdwGYAWAmgLsBfAOlhPbC8F8DVJNTL2buy8yXMnMlM9+nFtcDM9/JzEcy8wBmPp6Z6x4/GCcCrKw+zhnSIfrOQNgyF+AaM0R2hltv2KJ/bLiR370m2E78xq9bIp7TqA1ozrn8RpnIcLssV8uVPn/ICnJo52bo3VbxLxwoq9ZNJ0a000TrNxwJLf9g9vp9+L8ZwYpmRbVfLy9RWwZ3UnwoL8zaiMLDFUH5DTed2AMeN8Ef5oG69cOl+HzJDgzu1BQfTR4BQFkR22mTCSCqlqMlRWkmQqMj9QP1fHXhu5tG49Urg7vkGhcFfq7x2ewtqYTX7cLdp/YO2n/EEcHmqHATbKYnsqYQqW6RpkSF61xn/L1E0rgApVaTmX2HQs2t/z63f9D7cCZQLUDD7SK9f4YZt4swvk8by4k5EGD0uvfboG0DOtTUwSq4Y6ouDMKZ3974wzD9daNMDy4fGZpzYbewY12IlKeQBwDM/Lr5B8AMw+t6R37jTAQCbCuz0lhvxgoG6+tNj4tw2GCC2byvNChOXXsO/nZyL1vjNN7gWkZydoZb1RQshEJ1ICii5x9n9cWkYR1108P3q/ZYnkcTJP+YWnvrnrEkt7G0946D5aio9oeYP2Ll8pEFetRUUWl1kE+lUaYH24rKceb/fg77fYJSDE8zr6x5+BRcOKxT2P2NRLNZnzNEsXWf2k8J79T8J4M7NQ1q7FJberZpjLZNgkNhjROPP8D6RLbtQBkyPS5cM7orNj8yQRcORITzh3bQ6yhZLQ4ARVPYVVwR1hHsCzDcLrJ8LjRBE25SNE54OVHMXB8sCG2/+q2qgRrPbZ7ANUd3C5N/6uzB9hZ/Xo9L9z0ZMVbV1aaNy0YUAKi5bq2PdqUvgGY5Xvx297ggP9noHvn6c9CpeU7I3xRITK/rSHfzTO0FEf1g+uwzR0aTIpRX+eFnjrhaee/a4bhngnG1ZV3mgg2agsflCtpnzOMzg6IxNLNAQQvrLF4zVqF7ORluZLit+whX+gJBmsLFR3dGq8ZZtu35w7s2j76TBQsi9LfdsPcwZq/fV2dNAah54G/7cCky3C6M690q6POi0qqwK1mG8sDFMo6R6uo6WhvHE3q1xvy7xuLpC5Ro7kvVrNs/j+nmWILgSQbfhS/Aeuz/oQpfkD/pGi3ahoF/nNUPyx8Yj5//dgL+GCYKJ8vrwrRVe4Laoxqp8gX0khtmNH+K1cTm8wdQXuWf+BUNAAAgAElEQVTH8K7N8YdjuqBrGKGk7Wsm3PrNaAZrkZuhR6Bled04a1B7TLl0CDY/MsGypI0VW4vK8O780L4Kxmfx3WuG46KjO+kBKlq48Cg1gVHTjDO9Ln08K9VkRy1P56QjQ2sedWmZG1bLiieRhILx12yeDVKnmakDlFf7EeDImsLwri1w9aiaB0cxH4Vi9CmYMyaBYOdtx2bZaJzlQU+bdmYrZ3HT7AxFU7B48Cp9fkvbfWaUfrDag9WhmT1hZeaFWdHr0dRVUzAzZ8M+vQy2EWPSoZGHv1oVIjSj0TjLgwuO6og/HBO9lEmrvCxd6CciU7y7IYLM72e9iF1ppd9Ss2Eo4czZGW40yfFa+i2Amiqy4Tq3VfkCYTWnI9s1Qeu8TJz61OyQif2c537B1W8sQKNML2au3YuX52yyPAYQWqBx0b0n4re7xwVluOvXZRAK2Rk1GdWVvgDuPKWX5eQbCasWo4DS/AgAju7SHMO7tsA/z+qnV/Q1+gY27yvVfWgZbhcqqgPIy/LoOVDmJMI5fzsed52qWA4y3K6kawoc5rXV+3rBqO4tccXIAgSY8cPqvZYJSuEgAKt2FofEhRs1BavJwHiTtcrLwvIHxqNrfiPcPr6nLSfkC5cGR/FmZ7jD1ryvDLMS1ibCn+84wfIcNWpv7I7mJdsOhjVLBY8h/o3Kf7EoW/7JImuhkOF2xSwU3C7CI+f0xx+POyLmsb151bCgZCkn8QUC+OfXirvuUEW15WLCro9WE95WJhQA+G7l7ohJYntKKrH3UCU2qd+/+f0luPWDpViqhpdmeAjbo7QKXa7G6zfPzcCIri3QPDcDLRtlhuTXADWJjH3b52F3cQUunDIXt36wVPWvxX7PadngZqeytgiLZnH+68fLFHOpx61rVOHKkQDKQuwPx3TB638YBq+HkutTANCKiG4holsNr7X3ibmbE0zTnAwM6tQUzMCdnyyPSVUjqunCZoTBEW+UyW8tshQ+fdrmRcyOZWZs2V9quSrzuMjSoba1qMzSJKY96G3yIucK1KZJ0Jn/V2PHj+R/ibemEA6tH4aZ0io/9h2ujGmiMEahGZOl7DCqe37Y1Xi8eEmNdTfeCwfKqi3t9baFQpTfj7mERji0e+nTxTvwsUFQe90uPHBaaMitFQ+cfiTevromsfSiYZ3wkClcVytkSVCiz3YWV+DjRdtxqMJnK3/BjFZi29yaU1swRYtYM0bb2dUYPW4XjuuRrxQUtFF6vK5E+q28CKWZTiPDa+19ShWuixdme7NV8k84CKQ730orfbhZLTOhHDL0j2+s3z7yiJY4tlvLkH0iMX9TEdbvPYxMi4nFHSb8cubaQss+s9plh/OhnKKWs66NphAOza6uEc2EZYeHz6ypszPjtjH6uI2Yi8qZiWWiSFQV2doyrk9rtG2SFXQvHCyrssxVsIycs2B/lOq9dley4TqKed0u9GmXFxS5Y0b7G2V5gifWNk2ydOeuRotGmbh5XA/8xRAGbD5OLByhhqOae3RrmsLq3dGTAyur/bXKyxncqVlcnpNohA19YOYHHT97isGI3Gs5EkQ13bqOvP87AMCTkwaqx6zZ77+TBuKm95fg+F75eGvuVv27b5pKaQCRV29arHiGxxVSktjjCg2/1ATeyCNCM2GjOdlO7dcW36zYXad2osamPUBox7SsWjygZi4d3hlZHhdu/2gZcjPdePy8AbhbDQZonKUUCtwdpfpkLBNFgorI1gm3i3RzCwAcKKtCO4uGMo0y49M3y26s/sEwJqYsrwseF+nmJCs0u7rdCfLGcd0tt0cLe7Xi2YsGYeCD03BEK1NeiPpshFtI5mQofThcRHjwy1VBnf3MmAMkNO40hRE7Rd3j4eoTXHsPOiF05egPMMAcdMwzB7XHmYPao7TSpwuF2estGo1HGYhmj8zwuPQV/PRbRgOw1hS0ui+TjgpNxmuWmxHR9HHagHbYVVyOvSW1701gTJjK8Lhw07juGNalOa58VWlBGK8VkJZz4XW5kJtZ48Bbdv9JOO/5X0OE/uz1wYUH7U4UGW5X1CKHqcD2A+VBkUIHLcxH714zHAUt7QURRPO5TLPhPwJCtU6vm1DtZxSVVumChZkjLtLisZCIlbwsL07s0zok5Fu7nrPClNrQgh4yPC6s2hVZmzhjoPUxEkXif6spTDT7fySICJUm1bmsyheiKWjkxpgqb0Z7sDM8Lj25LMOtbPO4KaQOvhbtUduuVtFq3gx8aFrE6pHG671vYh9ketw4vmfNiiheD3hfNbPZfJ1EhFP6tQ1Zydpxglux7h+n4AobUUepQrsmWbhkeCccKKvSK4BqjDiihWVMvBXDVU2zeW4GSiqqa92JzDypaiGdB0qrdV9LtBpL5uuIhlUmd23I9LhDyk3sO1yFfu2b4MlJAyN+d+7G0OAHY3HKzY9MwGmGhl7JQISCAaVMUe1tAuZwsfIqf8Rjzr97bK3PpdmFjZEkLRsrTlS3K7TOj9bD1m5pZzPRqmMeLKvG+S/8GvZzrQT50vtOCqod9ZiacRpPp2uvNo2VxkcmsryukBVqbSt+phs7iyuQ4XaHdTTbJS/Lq5dCOfnJWTjv+dC/ubGJTzjCBXFcPrJA14KjCYVo+SFmzKXYa0uWNzQ0dPJbC4PMdBpmc635mk4+so3ewCtVsFM6+xaLn6uIKLWuJA5o4aNPThoQfWcTVgvwd+dvU1Vg6++0aly7CRqosWcHAox/ntUPLRtlGhrVUEgm5++FSmSUuU2kXYw1b9bvOWRZxneXyV5vjHrRztskxxtkEjhvaEd8e9OoWo0pHN/eNNoy7DLT4w5ZoRpl54+Guk/1kVfUool2i+KFI79xJopKq7CzuEK/rwDg5w370Kl5Dm4fHz4jf9btx+PCYZ3CNhkaWtAMvdUWo+ECG7R731zNNxrawuOaUV3QP4IjOxqxVIp955rhyG+ciUfPCS77fsuJPQAAz186BN1axXYdTmNneTYUwGQA7dWfawGMAfAiEf3VuaElHq0kxVmD7KW8G7HSBko181GU77551bAoe4SnUZbS9GXBPeP0bVY+ha4tc2udkQwE17w574VfMfGZOVG/Y3yoI5nLzA1/nCLL68K8TUVBuSRGZ36kLNp05ZkLa3piaf6AWJrEWGHMWDYmoV380jxsLSqLWCCwU4scNMtRuuQZI5AGdGiCKZcOQYvcDBAR2jfNDis4AlGi5cKhJZjdPaEPvjB0rYuVTE+o1jy2Vyu9jImZ3+4eh0mmrnnheo+nAnaWjS0ADGbmwwBARPdDqX46GkpBvH87N7zEYkw0iyXmHLDWFPq0zUNRaVVEZ1mk80QKEmRWbKRWtmCvlU8hwLYbwFiR6XVj24FyBCyOozXLuXBYsBPbGK1UVx9KPMjyuPUaUaWVPuRmevTyD38eE3sCWjrQWS2ZMq53KwwtaI5HvlkTtLqvDUZns1V9uWgRXIrG5g+6PxjBpR0yPS5sP1CODs2yax0RaObvZ/XFroN173380aLtOFhWjWtH19wz7Zpm6w2swvHxn0bgnOcUc5tWIDIVsTNLdAJgDE6uBtCZmcuhlNWuNyj3dy1DUg2vbx/fE5OGdsThSl+tU7/tjCLcw2LlU/AHItdyikamx4XVu0rwyeIdermOXzbsw3vzt+I/05QSxmYBtXCLksjUrkkWcutosogHxhIf5zynlErQzGx/tVmEMN3QboOLj+6sT+an1zG6hYiCBAMzY6NB0ESLUMrNdKO0yg9fIIDWeUo4tNncmeFx4cIX5+LCF+cGbY/U3jIagzs1w4T+baPvGAUt7FQzofr8SmOnaIuuLi0VoXHNqC5olmudRJkK2BEK7wCYS0T3q1rCzwDeJaJcANHbI6URRk0hVt6cW1Pmukm2F3nZHpRW+iL6FJzCnNH87YpdKKvyW9ZesovmpJu7cb8e2XPRS/NwxyfL9ZIHAVNixRPfK8LilzvHOp69awejg3XNbqWBkd1eCOmKZt45vlcrDFBDaPu0rbu5zmhTf/qHDUH+JC0KLhyNMj04XOELah8brqS1uSPh49PWWe6XSDQz7H+nr0NppQ/d7v4GZdXRW7FqFVFT4VmIRFSdnpkfJqJvABwDZQE7mZkXqB/bacuZRnCtY4+MDwmRYi4prfQhL9tb6zDQ2mL2KUx+axHOGdyhTpqCloPxkUVBOc1Wa85mLa9DpzYnsAphXLT1IJ422N3rG8b7QOs5kWsRmVUXlm0/GNRyMpr5qFGWR+k+6K9ZXZv7N2hC24xWdv2n28fUYcR149huLTF3YxEOlFXrvoU9JRXw2gyrrsviLBHYFVmLAXwIpc/yXiKyV2g+zVA0hbr9wa4Z1QUn9WmDFTuK8fSPGxDg2guaSEQyS1nVPio8XKmX5a0NkQTKGjW1v9rPqKj2Y8NexZSQHYdeAfHEarLaWlRWZ8drKmO+DzY/MsF2ToJd1u89jL2HajSF6OYjDw5VapqCcl/ZLRly5kAlhj/RCy0jfzlByZA+XOHThe78TUVhS4Yb+dvJvSwTSFMJOyGp1wPYA+B7AF8BmKr+X++wEykUji5qt6bbxvdEfuNMTF+tlCqui0kqUhcrIPxYjZqCdoxZ6wrrVItdSwqzQpt3Kqv9uPLV3zDuiZ/AzBjXuxUuHxHaPSpZhHtoY607lU70aN24TuGX0Rh5hFKl1NhPwKpaqZFGqhZttMOv3RO5H7mGtuDIb2yv/4GT7D1UEZSvkOGJ/qD/acwRtS5BnyjsLB1vBNBTbZvZn5n7MXP/qN9KQ+pi/3/kbCUO2VxlU2nWE/sKvS4ai8flwtJtB+HzB4LMWrXN3tW445RgZ6y5V+3rv27Br2rG5p2fLMcni3agfbP4rkrrQkeLMEC3i9CnXWJCYpNBfuPMOoVfRuPOU3rrEVwa0cyUWR43Fm45gB0Hy2JOKJu6XOlbHq9EtNpy5TEFOL5nqyCTaTK1l3hiZ7baBiB8daoIENHNRLSSiFYQ0btElGX6PJOI3ieiDUQ0j4gKanOeeHDtGwswY21hrYXC0V1bBIWXPn/JYABKOn+inZnaQ/n5kp1xzdg1x1aP7d06zJ7Ae79tw46D5Zbd4VIFnz8AZk55G28q8uOtx2HqDceqme5KFJHWizoa2jM2+a1F8LgI90zojdf/YC9X5+KjO+P+0/rUdthx46iC5iit8mGZoXBfKt/rsWBHKGwEMJOI7jRmNUf7EhG1B3ADgKHM3BeAG8AFpt2uAnCAmbsBeBLAo7ENP35ohbzqUubCyMl92yLDrbTbq0t+QDgiWZY0oXDrh0vrVNnUjNn8YkxcGtXd2gRTF+e2U2jNi6r8StOheMXBNyS65jfCke2aIEut8d+jdWNM7N8WVx5TEPW7Rt/OvE1FuHpU15C4fS1ShyjYjPraL5sT0n0sGtX+AL5buQc3qSXygcT0T04EdmarrVD8CRmo6algNy/bAyCbiDwAcgDsNH1+BoDX1dcfARhLyX5C43j2TI8Lhyujh6rVlnC/KWMZbeNrO46wSBh9EuYIpyfOt656kmrT7ZDOzfCf8wcgJ0MpeWHVnlSwT5ZXSQis8gWQl+3F/TYa5NgpTzG2l1Is0U0UsgJfuv1g7QYbR4wVgyP1fkhH7ISk1qqvAjPvIKLHoQiVcgDTmHmaabf2UMxTYGYfERVDyaC2qCWdGOI5iR2q9OHd+Vtt9fCNJy0MiTGXvTJPf31smNW8XRZvVZLRnpw0APd8ugKz1hXi9vE9cd3xoQ1MNFJtEf7xn0YiEGCUV/uxtagspkZKQijaqn/epqKYzHAn9GqFH9fsxQm9rHsHaLH8mWppeKN2caA0+X8zo/N+6fZifHX9sXrZ9nQn7DKJiP6r/v8lEX1h/ol2YCJqBkUT6AKgHYBcIrrEvJvFV0MMI0R0LREtIKIFhYWFFl+JDatiboZz1fn4ZrRCZLFQl1EYHarbimr63V5u0dg8FrTmLBP6tUNplR8b95Xi+Z9+j/idI1KwnpDLRcj2uoP6DAi1I8/QLCmWqqVaouMrVxxl+fnt43vi7auPRoZHqUj66eLtes2qPx+f/JIk/TsE99Lo275JSppKa0Mk3flN9f/HAfzH4ica4wBsYuZCZq6GkuMw0rTPdgAdAUA1MTUBUGTaB8w8hZmHMvPQ/Py61QwpLq/GxGfmBNnDjTjxZ71mVPw1BbvtE43Utd7KONWxbFy1RSvsFckZnUwqfYGIDeYFexgXUbFE2YVJYNZpnpuBY7q1VISCP4DPFtdYnod3De0emGjqcyZ82L8iMy8kIjeAa5j5J/OPjWNvBTCciHJUP8FYAKtN+3wB4HL19bkAfuRowfl1pFg1F1j1MAacMXfkZcWn1aGZSE7xW9XSvBozbxtT5/N1bJ6DeXcF94C4O0EtAuONP8Co9gfCdsoSYicrBv+MucdxOLRy5wfLasqvpUK0WH3RCqyI6FNgZj8R5RNRBjNH7tgd+t15RPQRgEUAfFCyoqcQ0UMAFjDzFwBeBvAmEW2AoiGYo5PiyieLtuOWD5aq47PeJ17RR0ausBGREW+aGvwKt5zYAwUt49N1qnVecA+I0qrQkNeXLhuKq99YELI9lTgiPxdV/gA6plAeRbqTHUPugLlOVji2FpXh71NXBTWwSXYsijaGly8fivbNsutNfoKGHdG+GcDPRHRvLCGpAMDM9zNzL2buy8yXMnMlM9+nCgQwcwUzn8fM3Zh5GDNvrMO1RGX93uglg+P59516g5I0lFvLcg910ZnONqyAo5UdqAvGMgp/UR3O4/qkpsnIyNDOzXGowpdypTjSmViEQixMX703qrkpGYzt3Rq92uTF3Own1bHzROxUf1ywH4qakhhDMo02eePEFk+Z37tNHk4f0A6uWqia0YRTNIFh7F/gRFXGswe3xyeLdujx5ACCGtkP6tQUW/enbk0hr4csm9gLtadVnv1Ogree1APjd0RuYC8kB8dCUlORjKAa8DXbjbH88ZQKLhc5WoEzmuCYe+dYDP/XD3h59kZcdWx8nd1PnD8Ql40oCBIE/To00UtfvHXV0WH9NqmAFmdQn23DieKGE7qhSU5s/QGGdG6OIZ1r3wlQcI6oQoGIZsAiTJSZT3BkRA4S1BjEsL2y2q/UeK/0OeJTSBZtmigrt53Fde82ZYVRIACKv+H7W5Q+x6nQaS0SmizYVVweeUchKrec1NOxY4/q3hKz1yctbalBYufJvc3wOgvAOVAcx2mHuVuURrWfkZ3hVoRCmsiEWPwNqVArJlUxO86F1OJQRfBUM/KI5Iej1nfsmI/MGT4/E5GdkNSUw1jB1DinVvkCusBwNiA2NmqTi2BF2yYy8ZlZsVOxZzeJUuZZSC49WjfCkm1KWYtLhnfCg6f3TfKI6j92zEdGw58LwBAAbcLsntIY69wYJ/8qf0B3QqeKphAvM1a3VkrhMiGYNnmZWIqabmRC6uMPiA8oEdgxHy1ETf8ZH4BNUKqbph1BBeGMQsEX0J3Q9e2Wm67a+IVgtCJrVj0WhNSjTV4WjioQAZ4I7JiPElvNzUGCoo9UqbD/cCV2l5TrlUxTITHGDvEyLTVUUt0RLgQz15RJLzhH2CeDiM6O9EVm/iT+w3EWq8zDIX+frvzfWVmFpIlMAJA+AiwVMVaSFVKXy0YUoE2ce0oLkYm0XDpN/b8VlEJ2P6rvjwcwE0qBu7TCuLo2O5Qz3C60zstEQYv4lIOIB6nk9K5v3HFKr5SotilEpm/7JhH7gwvxJ6xQYOYrAYCIvgLQh5l3qe/bAvi/xAwv/jTN8aLaFwgxvuRmejDvrnFJGZMVogQ4S5bXnfQ+v4KQitipf1CgCQSVPQB6hNs5lWEGBnVsikyvG+ZirJF6LKQiokUIguAEdrxtM4noOwDvQonZuQDADEdH5SCaHb7KH8D5L/yqb99d4kzWr5OIMiEIQryxE330FyI6C8BoddMUZv7U2WE5CwEoKfdh/qaQfj6CIAgNGltxeaoQSGtBANSYXIiA6jCd1wRBEBoyzhXaT1FI/beoNLhnUM8UrIkufgNBEBJNgxIK2hxLBNz16fKgz6ZcNiTxA4qAHX+BRCgJghBvYhIKRNSMiPo7NZhEQKRMuNsPBJdMzpZmK4IgCNGFAhHNJKI8tTDeUgCvEtETzg/NOYiADmpv3p9uHwMAyHSnl1AQ05IgCE5gx9HchJlLiOhqAK8y8/1EtMzpgTmBlptAIBzbrSVa5WWhZaNMAMF1kQRBEBoqdmZCj5rFfD6Ar+wemIh6EtESw08JEd1k2mcMERUb9rkvxvHXAgKR0pfZRdCzWlNRKEQreic+BUEQ4o0dTeEhAN8BmMPMvxFRVwDro32JmdcCGAgAROQGsAPWYa2zmXmi/SHXHYIiFNxEcLsIM24bk3p12lNsOIIgNAzsJK99COBDw/uNUFpyxsJYAL8z85YYvxdXaqKPCH5muFRB0KVl6hTBs4uUzhYEwQkilc5+Bgg/8zDzDTGc5wIoZTKsGEFESwHsBHAbM6+M4bgxo5lc/AFOe/NLvLqzCYIgaEQypC+A0nUtC8BgKCaj9VBMQn67JyCiDACnw6BtGFgEoDMzDwDwDIDPwhzjWiJaQEQLCgsL7Z46BGPEjmY+EgRBEGoIKxSY+XVmfh1AdwDHM/MzzPwMFFPQwBjOcQqARcy8x+IcJcx8WH39NQAvEbW02G8KMw9l5qH5+fkxnDoUgqItBJgtm+6kEhJ2KghCorETctMOgLEGRCN1m10uRBjTERG1IbVsKRENU8ezP4Zj26baH8Dktxaq51WagKeyTIhmGhKBIQiCE9iJPnoEwGIi0splHwfgATsHJ6IcACcC+KNh22QAYObnAZwL4E9E5ANQDuACNjc6iBMB9bArd5bA7SL4A4HUiziKkVQWaoIgpCcRhYK6ip8O4BsAR6ub72Dm3XYOzsxlAFqYtj1veP0sgGdjGXBt0VbeOw6Wo3OLHPjZumezIAhCQyaiUGBmJqLPmHkIgM8TNCZHyfC4QAACavKaIAiCUIMdn8JcIjrK8ZEkiCpfAEQEXyCg5ymkKuHsaAu3FEmWgiAIjmBHKBwP4Fci+p2IlhHR8nStfaRRU+YidYVCuKFV+QI45zmljWjqjl4QhHTFjqP5FMdHkQCMGcAbC0uxsbAU5w7pkMQR1Q5NWAQk/EgQBAewU+ZiCwAQUSsoiWz1hlTWFKLhD4hQEAQh/tjpp3A6Ea0HsAnATwA2Q4lGSnvSWShU+aTHtCAI8ceOT+FhAMMBrGPmLlAymn92dFQOc82oLgAAV+pVyw7GQhnQrEaHKnySqCAIQtyxMy1WM/N+AC4icjHzDMRW5iLlyMlQrGaprClEG9mhiuqEjEMQhIaFHUfzQSJqBGAWgLeJaC8An7PDcpYctR9zKguFaJRUpPWfQBCEFMWOpnAGgDIANwP4FsDvAE5zclBOYAzWqQ9C4WBZFTwpnmchCEL6YUdTmASlO9p6AK87PJ6EoJmP3KnuU4jAvsNVyEjnCxAEISWxIxQKAFxCRAVQeizMhiIkljg3LGfRNIXKFI/gsequpm3bf7gyJftKC4KQ3kSdVZj5PmY+AcCRAOYAuB1K8520ZOZtY5CtCoVtRWVJHk14KIppa86GfSk9fkEQ0pOomgIR3QPgGCh9FBYDuA2KtpB2ZHhcKGiZi72HKgFEn3hTmbIqPzbuK032MARBqGfYMR+dDSXaaCqU5LW5zFzh6KgcxutWhEG1P7XNR4IgCInGjvloMJSEtflQGuYsJ6I5Tg8sEUiPZkEQhGDsmI/6AhgFpePaUADbkKbmI00EtGuaDQC4ZnTX5A2mlhhDa/95Vr/kDUQQhHqJHfPRo1AS154G8Bszp30qbeu8LGx+ZEKyhxGVFTtKIn5+ZLu8BI1EEISGgp0qqROIKBtAp/ogENKFxVsPoLzaH3EfqZMqCEK8sVMl9TQAS6BkM4OIBhLRF04PrKFTUR3eCZ7hcaFZjhcFLXISOCJBEBoCdrKfHgAwDMBBAFCT1gqifYmIehLREsNPCRHdZNqHiOhpItqgdnUbHPsl2CPdetJES1ZefN9JaJqTkZjBCILQYLAjFHzMXBzrgZl5LTMPZOaBAIZAqZ/0qWm3UwB0V3+uBfBcrOeJhXQKNtJyKOZu3J/kkQiC0JCwIxRWENFFANxE1J2IngHwS4znGQvgd62Lm4EzALzBCnMBNCWitjEeu15yxsB2AIALpsxFwNRlLY1kmyAIaYYdoXA9lBIXlQDeAVAC4KaI3wjlAgDvWmxvDyXEVWO7ui0IIrqWiBYQ0YLCwsIYT52edGhW4y9I9RpNgiDUH+wkr5Ux893MfJT6czeA1nZPQEQZAE4H8KHVx1antBjDFGYeysxD8/Pz7Z663hAtCkkQBCFeRBQKRDSCiM4lolbq+/5E9A6Uwnh2OQXAImbeY/HZdgAdDe87ANgZw7EbBBUiFARBSBBhhQIRPQbgFQDnAJhKRPcD+B7APCiOYbtcCGvTEQB8AeAyNQppOIBiZt4Vw7FtY1WGOl0oq6rpspZuUVSCIKQXkZLXJgAYxMwVRNQMygq+v9psxxZElAOlXtIfDdsmAwAzPw/gawCnAtgAJTrpypivIAYoTV20/5i6Gq9eOUx/n05RVIIgpBeRhEK5Vg2VmQ8Q0dpYBIL6vTIALUzbnje8ZgDXxXLMhsTmRyag4I6pmLG2YTjXBUFIPpGEwhGmzOUC43tmPt25YQmCIAjJIJJQOMP0/j9ODkSwRzr7RgRBSH3CCgVm/imRAxGsuWlcd2wrKg/alq6+EUEQUp8G0/k9XaN2mmZ7kZvpTvYwBEFoIDQYoQCkZ9ROOveRFgQh/bAtFIgo18mBCIIgCMnHTj+FkUS0CsBq9f0AIvqf4yMTLElXM5ggCOmBHU3hSQDjAewHAGZeCmC0k4MSgjELArEoCYLgFLbMR8y8zbRJivEkCBEAgiAkkqg9mgFsI6KRAFiteJ4kEVoAAA50SURBVHoDVFNSOiFWF0EQhOjY0RQmQylF0R5KVdOBSNPSFLLoFgRBiExUTYGZ9wG4OAFjEcJgzGIWjUcQBCeJKhSI6GmLzcUAFjDz5/EfkmDESrsRjUcQBKewYz7KgmIyWq/+9AfQHMBVRPRfB8cmCIIgJBg7juZuAE5gZh8AENFzAKZB6ZOw3MGxCSqSmyAIQqKwoym0B2DMZs4F0I6Z/QAqHRmVA3C6zqwSkyoIQgKxoyn8G8ASIpoJxZw9GsA/1bIX0x0cW9ypD3WE0la4CYKQFtiJPnqZiL4GMAyKULiLmXeqH9/u5OAEa+qDcBMEITWxWxCvAsAuAEUAuhGRlLlIIKIbCIKQKOwUxLsawCwA3wF4UP3/ATsHJ6KmRPQREa0hotVENML0+RgiKiaiJerPfbFfQv1GdAJBEBKJHZ/CjQCOAjCXmY8nol5QhIMdngLwLTOfq5bIyLHYZzYzT7R5PEEQBMFB7AiFCmauICIQUSYzryGintG+RER5UJzSVwAAM1cBqKrTaOtAfTHB1JfrEAQhNbHjU9hORE0BfAbgeyL6HMDOKN8BgK4ACgG8SkSLieilMI16RhDRUiL6hoiOtD/02ElXU0xI6ezkDEMQhAZAVKHAzGcx80FmfgDAvQBeBnCmjWN7AAwG8BwzDwJQCuAO0z6LAHRm5gEAnoEieEIgomuJaAERLSgsLLRx6vqDBBoJgpBIIgoFInIR0QrtPTP/xMxfqKagaGwHsJ2Z56nvP4IiJHSYuYSZD6uvvwbgJaKW5gMx8xRmHsrMQ/Pz822cWhAEQagNEYUCMwcALCWiTrEemJl3Q+nFoPkfxgJYZdyHiNqQGnRPRMPU8eyP9VwNCcldEwTBSew4mtsCWElE86GYgAAAzHy6je9eD+BtNfJoI4AriWiy+v3nAZwL4E9E5ANQDuAClpRdC8SpIAhCYrAjFOyGn4bAzEsADDVtft7w+bMAnq3t8WMbSyLOEn9IJIAgCAnETpmLn4ioM4DuzDydiHIAuJ0fmgPI/CoIghAROxnN10BxEr+gbmqPMFFCgjOkq5YjCEL6YSdP4ToAxwAoAQBmXg+glZODEmoICUkVASEIgoPYEQqVxhBUIvJApqakIlYwQRCcwo5Q+ImI7gKQTUQnAvgQwJfODksQBEFIBnaEwh1QylUsB/BHAF8DuMfJQTlCGus24lMQBCFR2AlJPQPAG8z8otODcZp0NLuk45gFQUhf7GgKpwNYR0RvEtEE1acgJAlOZ5VHEISUx05BvCsBdIPiS7gIwO9E9JLTAxPCI+04BUFwClurfmauJqJvoFjms6GYlK52cmBCDaIdCIKQKOwkr51MRK8B2AClVtFLUOohpRXpOrGKUiAIQiKxoylcAeA9AH9k5kpnh+MsYnYRBEGIjJ3aRxcY3xPRMQAuYubrHBuVEBYJTxUEwUls+RSIaCAUJ/P5ADYB+MTJQQnBhLTjFIVHEASHCCsUiKgHgAsAXAil8c37AIiZj0/Q2ARI6WxBEBJLJE1hDYDZAE5j5g0AQEQ3J2RUQgjF5dUoLqtG4yxJExEEwTkizTDnQNEUZhDRt1CczWm7bE1nWzwDuOvT5Zi6bFeyhyIIQj0nbEgqM3/KzJMA9AIwE8DNAFoT0XNEdFKCxhdX0tIWr45ZBIIgCInATkZzKTO/zcwTAXQAsARKkTxBEAShnmGn9pEOMxcx8wvMfIJTAxIEQRCSR0xCIVaIqCkRfUREa4hoNRGNMH1ORPQ0EW0gomVENNjJ8aQrzMDAjk2TPQxBEBoAToeyPAXgW2Y+l4gyAOSYPj8FQHf152gAz6n/CyqaGySN/eSCIKQRjmkKRJQHYDSAlwGAmauY+aBpN61XAzPzXABNiciRukrpPKmu23MIS7cdxKPn9Ev2UARBqOc4aT7qCqVj26tEtJiIXiKiXNM+7QFsM7zfrm5zhHQMPgKA5TuKAQBDOjdL8kgEQajvOCkUPAAGA3iOmQcBKEVo1JLVPB2yqCeia4loAREtKCwsjP9I04SyKn+yhyAIQj3HSaGwHcB2Zp6nvv8IipAw79PR8L4DgJ3mAzHzFGYeysxD8/PzHRlsqmKs7NqvfRMAwKjuLZM1HEEQ6jmOCQVm3g1gGxH1VDeNBbDKtNsXAC5To5CGAyhmZsnSMrBwywH9tZT+FgTBaZyOProewNtq5NFGAFcS0WQAYObnAXwN4FQoDXzKAFzp8HjSjg17DwEAerVpDAB486ph6Nzc7JoRBEGID44KBWZeAmCoafPzhs8ZQEL6MnCaFj+q8gUAAOcO6QAAGNW9YZnPBEFILI4mr6Ua6Wx+KTyc1k3vBEFIExqUUEhHfAFFwzlc4UvySARBaAiIUEhxhndtAQAIpKf1SxCENEOEQopz96m9AQCZHvlTCYLgPDLTpDguF2FYl+Y4tZ8j1T8EQRCCaDC9HXMzPfjbyT2j75iCfPDHEdF3EgRBiAMNRlPI8rox6ahOyR6GIAhCStNghIIgCIIQHREKgiAIgo4IBUEQBEFHhIIgCIKgI0JBEARB0BGhIAiCIOiIUBAEQRB0RCgIgiAIOpRufQaIqBDAllp+vSWAfXEcTjrRUK9drrthIdcdns7MHLUhS9oJhbpARAuY2dz0p0HQUK9drrthIdddd8R8JAiCIOiIUBAEQRB0GppQmJLsASSRhnrtct0NC7nuOtKgfAqCIAhCZBqapiAIgiBEoMEIBSI6mYjWEtEGIroj2eOpK0T0ChHtJaIVhm3Nieh7Ilqv/t9M3U5E9LR67cuIaLDhO5er+68nosuTcS2xQEQdiWgGEa0mopVEdKO6vV5fOxFlEdF8IlqqXveD6vYuRDRPvYb3iShD3Z6pvt+gfl5gONad6va1RDQ+OVcUG0TkJqLFRPSV+r7eXzcRbSai5US0hIgWqNucv8+Zud7/AHAD+B1AVwAZAJYC6JPscdXxmkYDGAxghWHbvwHcob6+A8Cj6utTAXwDgAAMBzBP3d4cwEb1/2bq62bJvrYo190WwGD1dWMA6wD0qe/Xro6/kfraC2Ceej0fALhA3f48gD+pr/8M4Hn19QUA3ldf91Hv/0wAXdTnwp3s67Nx/bcAeAfAV+r7en/dADYDaGna5vh93lA0hWEANjDzRmauAvAegDOSPKY6wcyzABSZNp8B4HX19esAzjRsf4MV5gJoSkRtAYwH8D0zFzHzAQDfAzjZ+dHXHmbexcyL1NeHAKwG0B71/NrV8R9W33rVHwZwAoCP1O3m69Z+Hx8BGEtEpG5/j5krmXkTgA1Qno+UhYg6AJgA4CX1PaEBXHcYHL/PG4pQaA9gm+H9dnVbfaM1M+8ClMkTQCt1e7jrT+vfi2oaGARl1Vzvr101oSwBsBfKw/07gIPM7FN3MV6Dfn3q58UAWiANrxvAfwH8FUBAfd8CDeO6GcA0IlpIRNeq2xy/zz1xGHg6QBbbGlLYVbjrT9vfCxE1AvAxgJuYuURZDFrvarEtLa+dmf0ABhJRUwCfAuhttZv6f724biKaCGAvMy8kojHaZotd69V1qxzDzDuJqBWA74loTYR943bdDUVT2A6go+F9BwA7kzQWJ9mjqoxQ/9+rbg93/Wn5eyEiLxSB8DYzf6JubhDXDgDMfBDATCi246ZEpC3ujNegX5/6eRMo5sZ0u+5jAJxORJuhmH1PgKI51PfrBjPvVP/fC2URMAwJuM8bilD4DUB3NWIhA4oD6oskj8kJvgCgRRdcDuBzw/bL1AiF4QCKVdXzOwAnEVEzNYrhJHVbyqLah18GsJqZnzB8VK+vnYjyVQ0BRJQNYBwUf8oMAOequ5mvW/t9nAvgR1Y8j18AuECN0ukCoDuA+Ym5ithh5juZuQMzF0B5bn9k5otRz6+biHKJqLH2Gsr9uQKJuM+T7WFP1A8U7/w6KHbYu5M9njhcz7sAdgGohrIauAqK7fQHAOvV/5ur+xKA/1OvfTmAoYbj/AGK020DgCuTfV02rvtYKOrvMgBL1J9T6/u1A+gPYLF63SsA3Kdu7wplctsA4EMAmer2LPX9BvXzroZj3a3+PtYCOCXZ1xbD72AMaqKP6vV1q9e3VP1Zqc1ZibjPJaNZEARB0Gko5iNBEATBBiIUBEEQBB0RCoIgCIKOCAVBEARBR4SCIAiCoCNCQWiQEJFfrT6p/USsnEtEk4nosjicdzMRtazrcQTBKSQkVWiQENFhZm6UhPNuhhJDvi/R5xYEO4imIAgG1JX8o6T0LphPRN3U7Q8Q0W3q6xuIaJVat/49dVtzIvpM3TaXiPqr21sQ0TRSegG8AEMtGiK6RD3HEiJ6QS145yai14hoBSm19G9Owq9BaMCIUBAaKtkm89Ekw2clzDwMwLNQ6uyYuQPAIGbuD2Cyuu1BAIvVbXcBeEPdfj+AOcw8CEopgk4AQES9AUyCUvRsIAA/gIsBDATQnpn7MnM/AK/G8ZoFISoNpUqqIJgpVydjK941/P+kxefLALxNRJ8B+EzddiyAcwCAmX9UNYQmUJohna1un0pEB9T9xwIYAuA3tcJrNpTiZl8C6EpEzwCYCmBa7S9REGJHNAVBCIXDvNaYAKXOzBAAC9VqnJFKFFsdgwC8zswD1Z+ezPwAK41QBkCpgnod1MYygpAoRCgIQiiTDP//avyAiFwAOjLzDCiNX5oCaARgFhTzD9S6//uYucS0/RQoLREBpZjZuWqtfM0n0VmNTHIx88cA7oXSclUQEoaYj4SGSrbaxUzjW2bWwlIziWgelEXThabvuQG8pZqGCMCTzHyQiB4A8CoRLQNQhpryxg8CeJeIFgH4CcBWAGDmVUR0D5TOWi4o1W6vA1CuHkdbsN0Zv0sWhOhISKogGJCQUaGhI+YjQRAEQUc0BUEQBEFHNAVBEARBR4SCIAiCoCNCQRAEQdARoSAIgiDoiFAQBEEQdEQoCIIgCDr/D4sbcZ/3p533AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(avg_rewards, linewidth=1)\n",
    "plt.xlabel('Episodes')\n",
    "plt.ylabel('Average Rewards during Episode')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perform Generations of Exploration and Exploitation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generation 0\n",
      "Episode 2500/2500 || Best average reward 8.6248\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 1\n",
      "Episode 2500/2500 || Best average reward 8.7672\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.08\n",
      "\n",
      "Generation 2\n",
      "Episode 2500/2500 || Best average reward 8.5254\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.29\n",
      "\n",
      "Generation 3\n",
      "Episode 2500/2500 || Best average reward 8.8648\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.25\n",
      "\n",
      "Generation 4\n",
      "Episode 2500/2500 || Best average reward 8.6734\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.14\n",
      "\n",
      "Generation 5\n",
      "Episode 2500/2500 || Best average reward 8.2629\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.15\n",
      "\n",
      "Generation 6\n",
      "Episode 2500/2500 || Best average reward 8.5488\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.04\n",
      "\n",
      "Generation 7\n",
      "Episode 2500/2500 || Best average reward 8.3483\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.11\n",
      "\n",
      "Generation 8\n",
      "Episode 2500/2500 || Best average reward 8.7616\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.44\n",
      "\n",
      "Generation 9\n",
      "Episode 2500/2500 || Best average reward 8.7793\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.15\n",
      "\n",
      "Generation 10\n",
      "Episode 2500/2500 || Best average reward 8.6483\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.26\n",
      "\n",
      "Generation 11\n",
      "Episode 2500/2500 || Best average reward 8.5974\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.33\n",
      "\n",
      "Generation 12\n",
      "Episode 2500/2500 || Best average reward 8.6552\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.25\n",
      "\n",
      "Generation 13\n",
      "Episode 2500/2500 || Best average reward 8.6549\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.07\n",
      "\n",
      "Generation 14\n",
      "Episode 2500/2500 || Best average reward 8.5759\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 15\n",
      "Episode 2500/2500 || Best average reward 8.4362\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.44\n",
      "\n",
      "Generation 16\n",
      "Episode 2500/2500 || Best average reward 8.2111\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.21\n",
      "\n",
      "Generation 17\n",
      "Episode 2500/2500 || Best average reward 8.7531\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 18\n",
      "Episode 2500/2500 || Best average reward 8.3893\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.17\n",
      "\n",
      "Generation 19\n",
      "Episode 2500/2500 || Best average reward 8.4474\n",
      "\n",
      "Episode 5000/5000 || Best average reward 8.98\n",
      "\n",
      "Generation 20\n",
      "Episode 2500/2500 || Best average reward 9.0489\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.32\n",
      "\n",
      "Generation 21\n",
      "Episode 2500/2500 || Best average reward 8.7951\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.12\n",
      "\n",
      "Generation 22\n",
      "Episode 2500/2500 || Best average reward 8.9141\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.12\n",
      "\n",
      "Generation 23\n",
      "Episode 2500/2500 || Best average reward 8.6191\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.38\n",
      "\n",
      "Generation 24\n",
      "Episode 2500/2500 || Best average reward 8.7831\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.39\n",
      "\n",
      "Generation 25\n",
      "Episode 2500/2500 || Best average reward 8.7384\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.12\n",
      "\n",
      "Generation 26\n",
      "Episode 2500/2500 || Best average reward 8.8189\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 27\n",
      "Episode 2500/2500 || Best average reward 8.6827\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.25\n",
      "\n",
      "Generation 28\n",
      "Episode 2500/2500 || Best average reward 8.5549\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 29\n",
      "Episode 2500/2500 || Best average reward 8.5636\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.12\n",
      "\n",
      "Generation 30\n",
      "Episode 2500/2500 || Best average reward 8.9772\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.59\n",
      "\n",
      "Generation 31\n",
      "Episode 2500/2500 || Best average reward 8.5227\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.03\n",
      "\n",
      "Generation 32\n",
      "Episode 2500/2500 || Best average reward 8.7468\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.23\n",
      "\n",
      "Generation 33\n",
      "Episode 2500/2500 || Best average reward 8.7565\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.34\n",
      "\n",
      "Generation 34\n",
      "Episode 2500/2500 || Best average reward 8.2726\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.23\n",
      "\n",
      "Generation 35\n",
      "Episode 2500/2500 || Best average reward 8.6635\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 36\n",
      "Episode 2500/2500 || Best average reward 8.6614\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.29\n",
      "\n",
      "Generation 37\n",
      "Episode 2500/2500 || Best average reward 8.7717\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.1\n",
      "\n",
      "Generation 38\n",
      "Episode 2500/2500 || Best average reward 8.7988\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.05\n",
      "\n",
      "Generation 39\n",
      "Episode 2500/2500 || Best average reward 8.6924\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.17\n",
      "\n",
      "Generation 40\n",
      "Episode 2500/2500 || Best average reward 8.6715\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.02\n",
      "\n",
      "Generation 41\n",
      "Episode 2500/2500 || Best average reward 8.6487\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.24\n",
      "\n",
      "Generation 42\n",
      "Episode 2500/2500 || Best average reward 8.9336\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.17\n",
      "\n",
      "Generation 43\n",
      "Episode 2500/2500 || Best average reward 8.5324\n",
      "\n",
      "Episode 5000/5000 || Best average reward 8.96\n",
      "\n",
      "Generation 44\n",
      "Episode 2500/2500 || Best average reward 8.9422\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 45\n",
      "Episode 2500/2500 || Best average reward 8.7977\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 46\n",
      "Episode 2500/2500 || Best average reward 8.6349\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.07\n",
      "\n",
      "Generation 47\n",
      "Episode 2500/2500 || Best average reward 8.5538\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 48\n",
      "Episode 2500/2500 || Best average reward 8.7669\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.35\n",
      "\n",
      "Generation 49\n",
      "Episode 2500/2500 || Best average reward 8.5634\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 50\n",
      "Episode 2500/2500 || Best average reward 8.7221\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.12\n",
      "\n",
      "Generation 51\n",
      "Episode 2500/2500 || Best average reward 8.7741\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.24\n",
      "\n",
      "Generation 52\n",
      "Episode 2500/2500 || Best average reward 8.5659\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 53\n",
      "Episode 2500/2500 || Best average reward 8.5327\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 54\n",
      "Episode 2500/2500 || Best average reward 8.5441\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.15\n",
      "\n",
      "Generation 55\n",
      "Episode 2500/2500 || Best average reward 9.0299\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 56\n",
      "Episode 2500/2500 || Best average reward 8.6991\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 57\n",
      "Episode 2500/2500 || Best average reward 8.5346\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.14\n",
      "\n",
      "Generation 58\n",
      "Episode 2500/2500 || Best average reward 8.8197\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.27\n",
      "\n",
      "Generation 59\n",
      "Episode 2500/2500 || Best average reward 8.3812\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 60\n",
      "Episode 2500/2500 || Best average reward 8.6374\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.04\n",
      "\n",
      "Generation 61\n",
      "Episode 2500/2500 || Best average reward 8.7516\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.11\n",
      "\n",
      "Generation 62\n",
      "Episode 2500/2500 || Best average reward 8.9444\n",
      "\n",
      "Episode 5000/5000 || Best average reward 8.95\n",
      "\n",
      "Generation 63\n",
      "Episode 2500/2500 || Best average reward 8.6161\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 64\n",
      "Episode 2500/2500 || Best average reward 8.4412\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 65\n",
      "Episode 2500/2500 || Best average reward 8.5562\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.27\n",
      "\n",
      "Generation 66\n",
      "Episode 2500/2500 || Best average reward 8.3633\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.21\n",
      "\n",
      "Generation 67\n",
      "Episode 2500/2500 || Best average reward 8.6727\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.05\n",
      "\n",
      "Generation 68\n",
      "Episode 2500/2500 || Best average reward 8.4277\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.26\n",
      "\n",
      "Generation 69\n",
      "Episode 2500/2500 || Best average reward 8.5889\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.31\n",
      "\n",
      "Generation 70\n",
      "Episode 2500/2500 || Best average reward 8.8521\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.26\n",
      "\n",
      "Generation 71\n",
      "Episode 2500/2500 || Best average reward 8.8514\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.07\n",
      "\n",
      "Generation 72\n",
      "Episode 2500/2500 || Best average reward 9.0917\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.11\n",
      "\n",
      "Generation 73\n",
      "Episode 2500/2500 || Best average reward 8.6933\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.18\n",
      "\n",
      "Generation 74\n",
      "Episode 2500/2500 || Best average reward 8.4434\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Episode 5000/5000 || Best average reward 9.39\n",
      "\n",
      "Generation 75\n",
      "Episode 2500/2500 || Best average reward 8.5423\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.2\n",
      "\n",
      "Generation 76\n",
      "Episode 2500/2500 || Best average reward 8.3592\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.25\n",
      "\n",
      "Generation 77\n",
      "Episode 2500/2500 || Best average reward 8.4959\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.48\n",
      "\n",
      "Generation 78\n",
      "Episode 2500/2500 || Best average reward 8.4177\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.29\n",
      "\n",
      "Generation 79\n",
      "Episode 2500/2500 || Best average reward 8.5476\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.25\n",
      "\n",
      "Generation 80\n",
      "Episode 2500/2500 || Best average reward 8.8886\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 81\n",
      "Episode 2500/2500 || Best average reward 8.4684\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.22\n",
      "\n",
      "Generation 82\n",
      "Episode 2500/2500 || Best average reward 8.5635\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.33\n",
      "\n",
      "Generation 83\n",
      "Episode 2500/2500 || Best average reward 8.9847\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 84\n",
      "Episode 2500/2500 || Best average reward 8.5311\n",
      "\n",
      "Episode 5000/5000 || Best average reward 8.96\n",
      "\n",
      "Generation 85\n",
      "Episode 2500/2500 || Best average reward 8.6389\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.19\n",
      "\n",
      "Generation 86\n",
      "Episode 2500/2500 || Best average reward 8.7194\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.07\n",
      "\n",
      "Generation 87\n",
      "Episode 2500/2500 || Best average reward 8.3357\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 88\n",
      "Episode 2500/2500 || Best average reward 8.5949\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.06\n",
      "\n",
      "Generation 89\n",
      "Episode 2500/2500 || Best average reward 8.7655\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 90\n",
      "Episode 2500/2500 || Best average reward 8.7359\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.21\n",
      "\n",
      "Generation 91\n",
      "Episode 2500/2500 || Best average reward 8.7117\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.09\n",
      "\n",
      "Generation 92\n",
      "Episode 2500/2500 || Best average reward 8.8719\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.26\n",
      "\n",
      "Generation 93\n",
      "Episode 2500/2500 || Best average reward 8.6711\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.08\n",
      "\n",
      "Generation 94\n",
      "Episode 2500/2500 || Best average reward 8.7177\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.23\n",
      "\n",
      "Generation 95\n",
      "Episode 2500/2500 || Best average reward 8.7541\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.16\n",
      "\n",
      "Generation 96\n",
      "Episode 2500/2500 || Best average reward 8.5461\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.15\n",
      "\n",
      "Generation 97\n",
      "Episode 2500/2500 || Best average reward 8.6268\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.37\n",
      "\n",
      "Generation 98\n",
      "Episode 2500/2500 || Best average reward 8.3983\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.23\n",
      "\n",
      "Generation 99\n",
      "Episode 2500/2500 || Best average reward 8.4933\n",
      "\n",
      "Episode 5000/5000 || Best average reward 9.07\n",
      "\n"
     ]
    }
   ],
   "source": [
    "num_generations = 100\n",
    "best_avg_reward_gen = -math.inf\n",
    "\n",
    "for i_generation in range(num_generations):\n",
    "    print(\"Generation %d\" % i_generation)\n",
    "    # nomad\n",
    "    agent = Agent()\n",
    "    agent.alpha = 0.1\n",
    "    agent.gamma = 1.0\n",
    "    agent.epsilon = 0.005\n",
    "    num_episodes = 2500\n",
    "    avg_rewards, best_avg_reward = interact(env, agent, num_episodes=num_episodes)\n",
    "    # settler\n",
    "    agent.alpha = 0.001\n",
    "    agent.gamma = 1.0\n",
    "    agent.epsilon = 0.00005\n",
    "    num_episodes = 5000\n",
    "    avg_rewards, best_avg_reward = interact(env, agent, num_episodes=num_episodes)\n",
    "    # get best average reward per generation\n",
    "    if best_avg_reward > best_avg_reward_gen:\n",
    "        best_avg_reward_gen = best_avg_reward"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9.5"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_avg_reward_gen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_avg_reward_gen > 9.1"
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}