10-주어진_환경과_상호작용하며_성장하는_DQN/cartpole_dqn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 카트폴 게임 마스터하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gym\n",
    "import random\n",
    "import math\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torch.nn.functional as F\n",
    "from collections import deque\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 하이퍼파라미터\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 하이퍼파라미터\n",
    "EPISODES = 50    # 애피소드 반복횟수\n",
    "EPS_START = 0.9  # 학습 시작시 에이전트가 무작위로 행동할 확률\n",
    "EPS_END = 0.05   # 학습 막바지에 에이전트가 무작위로 행동할 확률\n",
    "EPS_DECAY = 200  # 학습 진행시 에이전트가 무작위로 행동할 확률을 감소시키는 값\n",
    "GAMMA = 0.8      # 할인계수\n",
    "LR = 0.001       # 학습률\n",
    "BATCH_SIZE = 64  # 배치 크기"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## DQN 에이전트"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DQNAgent:\n",
    "    def __init__(self):\n",
    "        self.model = nn.Sequential(\n",
    "            nn.Linear(4, 256),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(256, 2)\n",
    "        )\n",
    "        self.optimizer = optim.Adam(self.model.parameters(), LR)\n",
    "        self.steps_done = 0\n",
    "        self.memory = deque(maxlen=10000)\n",
    "\n",
    "    def memorize(self, state, action, reward, next_state):\n",
    "        self.memory.append((state,\n",
    "                            action,\n",
    "                            torch.FloatTensor([reward]),\n",
    "                            torch.FloatTensor([next_state])))\n",
    "    \n",
    "    def act(self, state):\n",
    "        eps_threshold = EPS_END + (EPS_START - EPS_END) * math.exp(-1. * self.steps_done / EPS_DECAY)\n",
    "        self.steps_done += 1\n",
    "        if random.random() > eps_threshold:\n",
    "            return self.model(state).data.max(1)[1].view(1, 1)\n",
    "        else:\n",
    "            return torch.LongTensor([[random.randrange(2)]])\n",
    "    \n",
    "    def learn(self):\n",
    "        if len(self.memory) < BATCH_SIZE:\n",
    "            return\n",
    "        batch = random.sample(self.memory, BATCH_SIZE)\n",
    "        states, actions, rewards, next_states = zip(*batch)\n",
    "\n",
    "        states = torch.cat(states)\n",
    "        actions = torch.cat(actions)\n",
    "        rewards = torch.cat(rewards)\n",
    "        next_states = torch.cat(next_states)\n",
    "\n",
    "        current_q = self.model(states).gather(1, actions)\n",
    "        max_next_q = self.model(next_states).detach().max(1)[0]\n",
    "        expected_q = rewards + (GAMMA * max_next_q)\n",
    "        \n",
    "        loss = F.mse_loss(current_q.squeeze(), expected_q)\n",
    "        self.optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        self.optimizer.step()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 학습 준비하기\n",
    "\n",
    "`gym`을 이용하여 `CartPole-v0`환경을 준비하고 앞서 만들어둔 DQNAgent를 agent로 인스턴스화 합니다.\n",
    "\n",
    "자, 이제 `agent` 객체를 이용하여 `CartPole-v0` 환경과 상호작용을 통해 게임을 배우도록 하겠습니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "env = gym.make('CartPole-v0')\n",
    "agent = DQNAgent()\n",
    "score_history = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 학습 시작"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "에피소드:1 점수: 12\n",
      "에피소드:2 점수: 19\n",
      "에피소드:3 점수: 68\n",
      "에피소드:4 점수: 14\n",
      "에피소드:5 점수: 12\n",
      "에피소드:6 점수: 13\n",
      "에피소드:7 점수: 10\n",
      "에피소드:8 점수: 13\n",
      "에피소드:9 점수: 10\n",
      "에피소드:10 점수: 8\n",
      "에피소드:11 점수: 12\n",
      "에피소드:12 점수: 13\n",
      "에피소드:13 점수: 12\n",
      "에피소드:14 점수: 14\n",
      "에피소드:15 점수: 9\n",
      "에피소드:16 점수: 12\n",
      "에피소드:17 점수: 10\n",
      "에피소드:18 점수: 12\n",
      "에피소드:19 점수: 15\n",
      "에피소드:20 점수: 11\n",
      "에피소드:21 점수: 14\n",
      "에피소드:22 점수: 11\n",
      "에피소드:23 점수: 11\n",
      "에피소드:24 점수: 12\n",
      "에피소드:25 점수: 25\n",
      "에피소드:26 점수: 28\n",
      "에피소드:27 점수: 21\n",
      "에피소드:28 점수: 22\n",
      "에피소드:29 점수: 24\n",
      "에피소드:30 점수: 34\n",
      "에피소드:31 점수: 25\n",
      "에피소드:32 점수: 34\n",
      "에피소드:33 점수: 56\n",
      "에피소드:34 점수: 152\n",
      "에피소드:35 점수: 136\n",
      "에피소드:36 점수: 200\n",
      "에피소드:37 점수: 193\n",
      "에피소드:38 점수: 189\n",
      "에피소드:39 점수: 200\n",
      "에피소드:40 점수: 200\n",
      "에피소드:41 점수: 200\n",
      "에피소드:42 점수: 192\n",
      "에피소드:43 점수: 162\n",
      "에피소드:44 점수: 200\n",
      "에피소드:45 점수: 200\n",
      "에피소드:46 점수: 200\n",
      "에피소드:47 점수: 200\n",
      "에피소드:48 점수: 200\n",
      "에피소드:49 점수: 200\n",
      "에피소드:50 점수: 200\n"
     ]
    }
   ],
   "source": [
    "for e in range(1, EPISODES+1):\n",
    "    state = env.reset()\n",
    "    steps = 0\n",
    "    while True:\n",
    "        env.render()\n",
    "        state = torch.FloatTensor([state])\n",
    "        action = agent.act(state)\n",
    "        next_state, reward, done, _ = env.step(action.item())\n",
    "\n",
    "        # 게임이 끝났을 경우 마이너스 보상주기\n",
    "        if done:\n",
    "            reward = -1\n",
    "\n",
    "        agent.memorize(state, action, reward, next_state)\n",
    "        agent.learn()\n",
    "\n",
    "        state = next_state\n",
    "        steps += 1\n",
    "\n",
    "        if done:\n",
    "            print(\"에피소드:{0} 점수: {1}\".format(e, steps))\n",
    "            score_history.append(steps)\n",
    "            break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAD8CAYAAABthzNFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xt4nHWZ//H3nXPStEkPaZum56YFW3oAspxBziIiiMuq7KqorF1XENx1d9X1p+K6ruu66uKuimVlQUFERBSloBUVilIk5dATpUlKS5OmyaSnJM155v79MU/SaTtJpm0mk2Q+r+vKlZnvPM/M/TTTued7NndHRETkaBmpDkBEREYmJQgREYlLCUJEROJSghARkbiUIEREJC4lCBERiUsJQkRE4lKCEBGRuJQgREQkrqxUB3AypkyZ4nPnzk11GCIio8r69eub3L1ksONGdYKYO3culZWVqQ5DRGRUMbOdiRynJiYREYlLCUJEROJSghARkbiUIEREJC4lCBERiStpCcLMZpnZ78xsi5ltNrPbg/JJZrbGzKqC3xODcjOzb5pZtZltMLMzkhWbiIgMLpk1iB7gE+6+GDgHuMXMFgOfAp5y94XAU8F9gLcCC4OflcB3khibiIgMImnzINy9HqgPbreY2atAGXAdcHFw2H3A74FPBuXf9+geqOvMrNjMSoPnEZFR5o81TeRmZXLmnIkn9Tw94QgPvrCLUHPHcZ134aIS/mzupJN6bXfnB+t20tTSeVLPkwyLpo/nmmUzkvoawzJRzszmAqcDzwPTYj709wDTgttlwK6Y02qDsiMShJmtJFrDYPbs2UmLWUROzh2PbWbXvnYe+ptzWDaz+ISf54d/eoPP/XwzAGaJneMOq9Zu52e3nM+p0yec8GvXhFqP+7WHyzXLZoz+BGFmhcAjwMfdvdli/pXd3c3Mj+f53H0VsAqgoqLiuM4VkeHT0NxJe3eYm++r5NGPnsfMiQXH/RwtHd3c+Zsqzp0/mR9++GwswU/pUEsnV39zLR994EUeu/UCCnNP7KOuqqEVgF9+7AJOKys6oecYzZI6isnMsokmhwfc/adBcYOZlQaPlwKNQXkdMCvm9JlBmYiMMp09YQ62d3Pt8hl0dIe5+d5Kmju6j/t5vvv0dvYe6uLTV5+acHIAKBmfyzffczo7mg7xmUc3Em25Pn41oWiCmF8y7oTOH+2SOYrJgO8Br7r712Meegy4Kbh9E/DzmPL3B6OZzgEOqv9BZHRqau0C4LwFk7nrvWdSE2rllgdepDscSfg59hzs4H+f3c61y2ecUBPVuQsm83eXL+LnL+/mwT/tGvyEOKobWykrzqcgZ1QvW3fCklmDOB94H3Cpmb0c/FwN/DtwhZlVAZcH9wFWA9uBauBu4KNJjE1EkigUdOqWjM/l/PIp/Ns7l7K2qonP/mxTwt/mv77mNSIR+Me3nHLCcdxySTkXLpzCHb/YzObdB4/7/OpQKwumFp7w6492yRzF9CzQX53wsjjHO3BLsuIRkeETmyAA3lUxizf2tvE/v6tmzuRx/O3FCwY8f+ueZh5eX8vN589j1qTj77volZFh/Ne7V3D1N9dyywMv8ouPXcD4vOyEzo1EnJrGQ/zZWSc3Emo000xqERlyRycIgL+/YhFvXz6Drzy5lV+8snvA8//9ia2Mz83i1kvLTzqWyYW5/PeNZ7Brfzuf+mni/RH1zR20d4cpT+MahBKEiAy53gQxpfBwgsjIML56wzIq5kzkth+9xL/8YgvtXeFjzv1DdRO/fy3ErZeWU1yQMyTxnDVvEp+4chGPb6jn/nUJbYVAdWO0g7q8RAlCRGTINLZ0MGlcDtmZR37E5GVnct+HzuL958zhnj+8zlV3PsPz2/f2PR6JOP+2+lXKivN5/7lzhzSmj1y0gIsWlfDlJ7Ym1FleEySIdO6DUIIQkSEXaumkJKb2EGtcbhZfuO40HvzwObjDu1et447HNtPW1cNjr+xm8+5m/uEti8jLzhzSmDIyjD8/o4y2rnDf/IaBVIdaKS7IZvK4oanFjEbpOXZLRJIq1Np5RP9DPOcumMyTH7+Q/3jyNe794w6e2tpAd4+zZMYErltelpS4lgaT3TbVHWTxjIFnWFc3tlJeUnhc8y/GGtUgRGTIhVoGTxAABTlZ3HHtEh5aeQ4ZZuxp7uDTb30TGRnJ+VCeO3kchblZbKg7MOix20OtLEjj/gdQDUJEhpi7J5wgep09fzJP3n4RNaHWpC5pkZFhnFY2gY11zQMed6Cti6bWrrQewQSqQYjIEGvp7KGzJ9JvH0R/8nMyh2W9o6VlRbxa3zxgR3V1Xwd1ei6x0UsJQkSGVO8Q16kTji9BDJelM4vp6omwraGl32N612AqLxk/XGGNSEoQIjKk+ibJHWcNYrj0dlRvrO1/6Y3qxlZyszIom5g/XGGNSEoQIjKkGuPMoh5J5kwqYHxeFhvrBk4Q86aMIzNJneWjhRKEiAypeMtsjCQZGcZpM4oGTBA1oUNp30ENShAiMsRCLZ1kZxpF+YktipcKy2YWsbW+ha6eYzuqO7rD7NrfpgSBEoSIDLHeWdQjeYLZaWVFdIXjd1S/3nQId9J+DgQoQYjIEEtkFnWqLZsZdFTHaWbqW6RPNQglCBEZWsc7SS4VZk8qYEJeFhvijGSqbmzFDOZNSe85EJDcLUfvMbNGM9sUU/ZQzO5yO8zs5aB8rpm1xzx2V7LiEpHkiiaIvFSHMSAzY+nMIjbFqUHUhFqZNbFgyBcLHI2SudTGvcD/AN/vLXD3d/feNrOvAbF/nRp3X5HEeEQkycIRZ9+hkV+DgGg/xD3Pvk5nT5jcrMPJoLqxVc1LgaTVINz9GWBfvMcs2nv1LuDBZL2+iAy/va2dRHzkDnGNtaysmO6ws23P4aW/wxFne9MhFpSoeQlS1wdxIdDg7lUxZfPM7CUze9rMLkxRXCJyEhpH+CzqWL0zqmNXdq3b305XT0Q1iECqVnO9kSNrD/XAbHffa2ZnAj8zsyXufsySi2a2ElgJMHv27GEJVkQSE2od2ZPkYs2alE9RfvYR/RDVoeiwVyWIqGGvQZhZFvBO4KHeMnfvdPe9we31QA2wKN757r7K3SvcvaKkpGQ4QhaRBPUt1DcKEoSZsbSs6IiRTH2ruGoOBJCaJqbLga3uXttbYGYlZpYZ3J4PLAS2pyA2ETkJI32ZjaMtnVnEtoYWOrrDANQ0HmJKYQ7FBem7zWisZA5zfRB4DjjFzGrN7ObgofdwbOf0RcCGYNjrT4CPuHvcDm4RGblCLZ2Mz8saNUNEl5YV0R12XtsTbVqq1i5yR0haH4S739hP+QfilD0CPJKsWERkeIyGWdSx+pb+rjvIsplFVDe28rZlpSmOauTQTGoRGTK96zCNFjMn5lNckM3G2oPsPdTFwfZuylWD6KMEISJDZjQssxGrt6N6Y93BmG1GlSB6KUGIyJAZbQkCos1M2xpa2Lw7OqpeQ1wPU4IQkSHR1tVDa2fPqEsQy2YW0RNxHt+wm4KcTEonjOx1pIaTEoSIDImmli5gdMyijnVa0FH94hsHmF8yjow032Y0lhKEiAyJUGsHAFNH2TfwsuJ8Jo2LzntQB/WRlCBEZEiERtE6TLHMrK8Wof6HIylBiMiQGG2zqGMtCxKEJskdSQlCRIZEqKWTDKOvuWY0Ob98CjlZGSwNtiKVqFSt5ioiY0xjSyeTC3PJHIWdvOcumMzGO648YuMgUQ1CRIbIaJtFfTQlh2MpQYjIkBht6zDJ4JQgRGRIjMZZ1DIwJQgROWmRiNPU2jkqNgqSxClBiMhJO9jeTXfYVYMYY5QgROSkjaa9qCVxShAictIam0fnLGoZWDK3HL3HzBrNbFNM2R1mVmdmLwc/V8c89mkzqzaz18zsLcmKS0ROTDji/T7Wuw6TahBjSzJrEPcCV8Up/4a7rwh+VgOY2WKie1UvCc75tplpULLICLGj6RCLP/ck63fG3yp+NC+zIf1LWoJw92eA+O+mY10H/MjdO939daAaOCtZsYnI8Xl97yE6eyLcv+6NuI+HWjrJy86gMFeLM4wlqeiDuNXMNgRNUBODsjJgV8wxtUGZiIwAze3dADyxqZ7mju5jHg+1dDJ1fB5mo2+ZDenfcCeI7wALgBVAPfC1430CM1tpZpVmVhkKhYY6PhGJo7mjB4CO7girN9Qf87hmUY9Nw5og3L3B3cPuHgHu5nAzUh0wK+bQmUFZvOdY5e4V7l5RUlKS3IBFBDhcg5g7uYCH19ce8/hoX4dJ4hvWBGFmpTF3rwd6Rzg9BrzHzHLNbB6wEPjTcMYmIv1r7ugmNyuDG8+azfqd+6kJtR7xuJbZGJuSOcz1QeA54BQzqzWzm4H/MLONZrYBuAT4OwB33wz8GNgCPAnc4u7hZMUmIsenub2HCfnZXH96GZkZxiMxtYiungj727qVIMagpA05cPcb4xR/b4DjvwR8KVnxiMiJa+7oZkJeFlMn5PHmRSX89MU6PnHlKWRmGE2aRT1maSa1iAyqub2bCfnZANxw5kz2NHfwbHUTMHr3opbBKUGIyKCa27uZkBdNEJe9aSrFBdk8XBkdmd6bIKZOUIIYa5QgRGRQzR09fTWI3KxMrls+g19vaeBgW7cW6hvDlCBEZFDRGsThLsu/qJhFV0+Exzbs7qtBTB6nBDHWKEGIyIDcPdpJHdQgAJbMmMCp08fzk8pdhFo6mViQTU6WPk7GGv1FRWRAHd0RusPe1wcBYGbccOZMXqk9yB9rmtS8NEYpQYjIgHrXXpqQf+So+HecXkZWhlETOqQEMUYpQYjIgHqX2YitQQBMKczlklOnAhriOlYpQYjIgA7XILKPeeyGM2cCGsE0VilBiMiAmtujK7kWxUkQl546lYtPKeGChVo4cyzS7h4iMqC+GkTesR8X2ZkZ3PtB7e01VqkGISID6uuDiFODkLFNCUJEBnQwSBDj49QgZGxTghCRATV39JCXnUFuVmaqQ5FhpgQhIgOKXahP0osShIgM6OhlNiR9KEGIyICa23vijmCSsS+ZW47eY2aNZrYppuyrZrbVzDaY2aNmVhyUzzWzdjN7Ofi5K1lxicjxUQ0ifSWzBnEvcNVRZWuA09x9GbAN+HTMYzXuviL4+UgS4xKR46A+iPSVtATh7s8A+44q+7W79wR31wEzk/X6IjI0opsFqYkpHaWyD+JDwBMx9+eZ2Utm9rSZXdjfSWa20swqzawyFAolP0qRNObuNLd3x11mQ8a+lCQIM/sM0AM8EBTVA7Pd/XTg74EfmtmEeOe6+yp3r3D3ipISrf8ikkzt3WF6Iq4mpjQ17AnCzD4AXAP8lbs7gLt3uvve4PZ6oAZYNNyxiciRehfqUyd1ehrWBGFmVwH/BFzr7m0x5SVmlhncng8sBLYPZ2wicqzDC/UpQaSjpPU8mdmDwMXAFDOrBT5PdNRSLrDGzADWBSOWLgL+xcy6gQjwEXffF/eJRWTYHGyPv5ucpIek/dXd/cY4xd/r59hHgEeSFYuInJj+dpOT9KCZ1CLSr4F2k5OxTwlCRPrV10mtpTbSkhKEiPSruW8vCNUg0pEShIj0q7mjm/zsTHKy9FGRjvRXF5F+NbdrmY10lnCCMLMLzOyDwe0SM5uXvLBEZCRo7tBCfeksoQRhZp8HPsnh1VezgfuTFZSIjAzNHVqHKZ0lWoO4HrgWOATg7ruB8ckKSkRGhmgTkxJEuko0QXQF6yY5gJmNS15IIjJSRJuY1AeRrhJNED82s+8CxWb2YeA3wN3JC0tERoKD7dpNLp0l9NXA3f/TzK4AmoFTgM+5+5qkRiYiKdW7F4Q6qdPXoAkiWGX1N+5+CdEtQ0UkDRzqChNxLdSXzgZtYnL3MBAxs6JhiEdERggt1CeJfjVoBTaa2RqCkUwA7n5bUqISkZTTQn2SaIL4afAjImni8EJ9ShDpKtFO6vvMLIfD24C+5u7dyQtLRFKtWZsFpb2E/vJmdjFwH7ADMGCWmd3k7s8kLzQRSSVtNyqJzoP4GnClu7/Z3S8C3gJ8Y7CTzOweM2s0s00xZZPMbI2ZVQW/JwblZmbfNLNqM9tgZmecyAWJyNDorUFoqY30lWiCyHb313rvuPs2ousxDeZe4Kqjyj4FPOXuC4GngvsAbwUWBj8rge8kGJuIJEFzR7QPYrxmUqetRBNEpZn9r5ldHPzcDVQOdlLQBLXvqOLriDZXEfx+R0z59z1qHdFZ26UJxiciQ6y5vZtxOZlkZWpXgHSV6F/+b4EtwG3Bz5ag7ERMc/f64PYeYFpwuwzYFXNcbVB2BDNbaWaVZlYZCoVOMAQRGYyW2ZBE645ZwJ3u/nXom12de7Iv7u5uZn6c56wCVgFUVFQc17kikjjtBSGJ1iCeAvJj7ucTXbDvRDT0Nh0FvxuD8jpgVsxxM4MyEUkB7SYniSaIPHdv7b0T3C44wdd8DLgpuH0T8POY8vcHo5nOAQ7GNEWJyDBTDUISTRCHYoedmlkF0D7YSWb2IPAccIqZ1ZrZzcC/A1eYWRVweXAfYDWwHagmupT4RxO+ChEZcs0d6oNId4nWH28HHjaz3cH9UuDdg53k7jf289BlcY514JYE4xGRJGtu79FmQWku0b/+POB0YDbwTuBsgt3lRGTsiUScFtUg0l6iTUyfdfdmoBi4BPg2msgmMmYd6uqJ7gWhPoi0lmiCCAe/3wbc7e6PAznJCUlEUq13FrVGMaW3RBNEXbAn9buB1WaWexznisgoo3WYBBL/kH8X8CvgLe5+AJgE/GPSohKRlNJucgKJ7wfRRsyGQcH8BM1REBmjDrZrNzlRM5GIxNHXB6EaRFpTghCRY2g3OQElCBGJo3c3ucJcJYh0pgQhIsdobu+hMDdLe0GkOf31ReQY0YX6VHtId0oQInKMZm0WJChBiEgcWupbQAlCROLQZkECShAiEof2ghBQghCROJrb1cQkie8HMWTM7BTgoZii+cDniC4l/mEgFJT/s7uvHubwRNJeJOK0dPaoBiHDnyDc/TVgBYCZZQJ1wKPAB4FvuPt/DndMInJYS2cP7miYq6S8iekyoMbdd6Y4DhEJNGuhPgmkOkG8B3gw5v6tZrbBzO4xs4mpCkoknfUus6E+CElZgjCzHOBa4OGg6DvAAqLNT/XA1/o5b6WZVZpZZSgUineIiJyE5nbtJidRqaxBvBV40d0bANy9wd3D7h4B7gbOineSu69y9wp3rygpKRnGcEXSg2oQ0iuVCeJGYpqXzKw05rHrgU3DHpGIaLtR6ZOSOqSZjQOuAP4mpvg/zGwF4MCOox4TkWGizYKkV0oShLsfAiYfVfa+VMQiIkfqrUEUaphr2kv1KCYRGWGaO7oZn5tFZoalOhRJMSWIBD31agN/rG5KdRgiSRddqE/NS5KiJqbR6MtPbGVSQQ7nlU9JdSgiSaWF+qSXEkQC3J26/e10dIdTHYpI0h1s125yEqUmpgTsO9RFe3eYhuYOIhFPdTgiSaXd5KSXEkQC6g60A9AddpoOdaY4GpHkauno0RBXAZQgElK7v73vdv2BjhRGIpJ80RqEmphECSIhdbEJ4qAShIxd4d69IFSDEJQgElK7v42sYEx4/cH2QY4WGb1ae2dRqw9CUIJISO3+dsqnFpKTlaEahIxphxfqUxOTaJhrQuoOtDNzYgHt3WElCBnTDmqzIImhGsQg3J3a/e3MnJhPaVEe9QfUxCRjl5b6llhKEINobu+htbOHmRPzmVGUrxqEjGm9mwVpqW8BNTENatf+NgDKivPZd6iLhuYOwhHXQmYyJvXVIDTMVVANYlC9k+RmTiygtDifnojT1KrJcjI2NQQ15OKCnBRHIiOBEsQgeifJzZyYT+mEPEBzIWTs+kNNE28qnUBhrmoQogQxqLr97RTkZFJckE1pcZAg1FEtY9Chzh7W79zPRQu1YrFEpexrgpntAFqAMNDj7hVmNgl4CJhLdNvRd7n7/lTFCNFJcjMn5mNmzCjKB2C3ahAyBj3/+l66w86FC0tSHYqMEKmuQVzi7ivcvSK4/yngKXdfCDwV3E+pugPtlBVHE0NxQTa5WRns0WxqGYOe2dZEXnYGFXMnpjoUGSFSnSCOdh1wX3D7PuAdKYwFIJgDUQAQrUUU56sGIWPS2qoQZ8+bTF52ZqpDkREilQnCgV+b2XozWxmUTXP3+uD2HmBaakKLauno5mB7N2UT8/vKpk/IY48ShIwxdQfaqQkd4kL1P0iMVA5VuMDd68xsKrDGzLbGPujubmbH7M4TJJOVALNnz05qgIeHuB5OEKXFeayr2ZvU1xUZbmu3hQC4aJH6H+SwlNUg3L0u+N0IPAqcBTSYWSlA8Lsxznmr3L3C3StKSpL7Zu5d5ru3DwJgRlE+DS2dhLWznIwha6uamDYhl4VTC1MdiowgKUkQZjbOzMb33gauBDYBjwE3BYfdBPw8FfH1OjwHoqCvbHpRHuGIE2rRZDkZG8IR59nqJi5cWIKZVgiQw1LVxDQNeDR4M2YBP3T3J83sBeDHZnYzsBN4V4riA6JNTLlZGUwpPDyrdEYwF2L3wXamF+WlKjSRIbOx7iAH27vVvCTHSEmCcPftwPI45XuBy4Y/ovhq97dRFsyB6DV9QrS5SR3VMlas3RbCDC4oVwe1HGmkDXMdUer2tx/R/wAxNQjNppYx4pmqEKfNKGLSOK2/JEdSghhA7ByIXkX52eRnZ2o9JhkTWjq6efGNAxreKnEpQfSjvSvM3kNdRwxxhehkudIizYWQseG5mr2EI1peQ+JTguhH3YHoPhBHJwiIzoXYreU2ZAxYW9VEQU4mZ87R8hpyLCWIftTGmQPRa/qEfOoPqAYho9/aqhDnzp9MTpY+CuRYelf0I94ciF4zivNobOmgJxwZ7rBEhswbe9vYsbdN/Q/SLyWIftQdaCc705g6PveYx0qL8ok4NGqynIxiz1RFl9e4UPMfpB9KEP2o3d/OjOJ8MuLsPV1apJ3lZPRbWxWirDif+VPGpToUGaGUIPpRu78tbv8DcHhnOXVUyyjVE47wx+q9XLRoipbXkH4pQfSjbn973BFMEG1iAtRRLaPWK7UHaOns0fBWGZASRBwd3WEaWzopKz62gxpgQl4WBTmaLCcjQzji/O/a7eza15bwOU+/FiLD4LwFk5MYmYx2ShBx9H7w91eD6J0spyYmGQl++PxO/vXxV/nbB9bTncDIuj0HO/i/P+7g/PIpFBdoeQ3pnxJEHLX7o9/EyvpJEBBtZlINQlKtqbWTr/7qNWZNymdTXTN3/b5mwOPdnU/9dAM9YeeL1502TFHKaKUEEUfd/mN3kjuaahAyEvz7E1tp7w7zfx84i7cvn8E3f1vFq/XN/R7/cGUtv38txCevOoW5Gr0kg1CCiKN2fzuZGcb0Cf3v91BanE9jS2dCVXqRZKjcsY+frK/l5gvmUz61kH+5dglF+Tl84sevxH1f7j7Qzhd/uYWz503i/efOHf6AZdRRgoij7kA70yfkkZXZ/z9PaVEerslykiI94Qj/72ebmFGUx22XlQMwcVwO/3b9aWypb+Zbv6s+4nh355OPbCDszldvWB53fo/I0ZQg4ujdKGggfZPltC+EDCAccdZsaTiuEUaJ+MG6nWzd08Jnr1lMQc7hfb+uXDKdd6yYwf/8tprNuw/2lf/ohV2srWri0289ldmT44/OEznasCcIM5tlZr8zsy1mttnMbg/K7zCzOjN7Ofi5erhj6zXQHIhefXMh1FEtcfSEIzz6Ui1XfONpPvz9St6zah2NLUPzXmls7uDrv97GRYtKuOq06cc8fse1S5g4LtrU1NUToXZ/G//6yy2ct2Ayf3X2nCGJQdJDKmoQPcAn3H0xcA5wi5ktDh77hruvCH5WpyA2usMR9jR3MLOfWdS9NJta4ukJR/jpi7Vc8Y1n+LuHXiEnM4PPXbOYfYe6+Ov7Kmnr6jnp1/jyE1vp7InwhWuXxJ0FXVyQw5evX8rWPS3892+r+OQjGwD4yp8vU9OSHJdh35Pa3euB+uB2i5m9CpQNdxz92XOwg4jHX8U11oS8bApzs9it2dQS+PnLdXxjzTZ27G3j1Onjueu9Z3Dl4ulkZBizJxWw8geV3P6jl7nrvWeSeYIf1Ou27+XRl+q49ZJy5g0wCunyxdN45xll/Pdvo30RX7r+NGZNUtOSHJ+U9kGY2VzgdOD5oOhWM9tgZveYWdwdTMxspZlVmlllKBQa8ph2JTAHotd07Swngedq9nL7j14mPyeLu957Jqtvu5CrTivt+8Z++eJpfO6axazZ0sCXHn/1hF6juaObz/18E2XF+dxySfmgx3/+miWUFefz5kUl/OVZs0/oNSW9DXsNopeZFQKPAB9392Yz+w7wRcCD318DPnT0ee6+ClgFUFFR4UMdVyJzIHppLoT0uuvpGqYU5vDoR88jLzsz7jEfOH8eO/a2cc8fXmfO5AJuOm/uoM97sL2b32xpYPXGetZWNdEVjnD3+yvIz4n/GrGKCrJZ8/cXkZuVqQX55ISkJEGYWTbR5PCAu/8UwN0bYh6/G/hlKmKr3d+O2eFO6IGUFuXx2p6WYYhKRrItu5t5eluIf3zLKf0mh16fvWYxtfvb+MIvNjNzYj6XvWnaEY9HIk7t/nb+tGNfkBRCdIedGUV5vO/cOVy7fAbLZxUnHFvsCCeR4zXs7x6LfpX5HvCqu389prw06J8AuB7YNNyxQXQOxLTxeQltwVhalE+otZOunoi2bExj332mhnE5mbw3gRFCmRnGne85nXeveo6PPfgSd7x9CaHWTqobW9nW0EJNqJWO7ugkt7LifD5w3lyuXlrKilnFqgXIsEvF14vzgfcBG83s5aDsn4EbzWwF0SamHcDfpCC2hOZA9JpRHJ0s19DcoQ7ANLVrXxu/3FDPB8+bS1FBdkLnjMvN4ns3/RnXf+sP/FMwwmhGUR7l08Zz7vzJLJxWyJIZRSyZMUFJQVIqFaOYngXivetTMqz1aHUH2jljdtz+8WNMD5qh9ihBpK3vPfs6GQY3XzjvuM6bNiGPJz5+Ea83HWJByTjG5yWWXESGk9pFYmyoPcDuAx3MSfDDfkYwm3q3ZlOnpX2HuvjRC29w3Ypxkzg2AAAK6klEQVSyhPqsjlaUn82KWcVKDjJiKUEEave38aF7K5k+IY/3JbiQ2XTtTZ3W7vvjDjq6I3zkzfNTHYpIUmiIA9Hx5R+69wU6e8I8+OGzKRmfm9B54/OyGZ+bpbkQaaitq4f7ntvB5W+aRvnU8akORyQp0r4G0R2O8NH7X2R76BB3vfdMFk47vv/spcV5amJKQz9+YRcH2rpVe5AxLa0ThLvz/x7dxLPVTfzbO5dyfvmU436O6UX57GlWDWK4vPTGfj7wf3/ijC+u4f/+8DqRyJDPlRxUdzjC3Wtfp2LORCrmThr21xcZLmndxPTt39fwUOUubr2knHdVzDqh5ygrzmNdzV7+8eFXuHpZKecvmDKm50Ts2tfG4xvrWb2xnr2tXXz1hmWcdwKJ9Xi9+MZ+7vxNFU9vCzGxIJvyqYV84RdbeGLjHr5yw7J+1yVq7ezh+8/t4AfP7WTelHHcftlCzp4/+aRieXxDPXUH2vnCtUtO6nlERjpzH/5vYEOloqLCKysrT+jcX7yym489+BLXLp/Bne9ZccLjzWtCrXzrt9Ws2dJAS2cPRfnZXLF4Gm9bWsq5Cyaz+0A72xpaqW5sYVtDK1WNrdTtb+OKxdP52KXlA2776O78eksD3/59DT3hCH978QKujlnf50S4Ow3NnVQF8VT3/W4lPzuThdMKWTh1PAunFbJoWiHlU8dzoK2L1Rv3sHpjPRvronsMLJtZRGtnD2/sbeOOa5fw3nOSs4z0+p37+K/fVLG2qolJ43L48IXzed+5cxiXk8kjL9bxhV9spjsc4R+uPIUPnj+vbxG8lo5uvv/cTu5eu50Dbd2cO38yVY2tNLV2cs78Sdx+2SLOXXD8icLdeeudawlHnF99/CKtjiqjkpmtd/eKQY9LxwTx8q4DvOu7z7F8ZhH3//XZ5GYNvq7NYDp7wqzd1sTqjfV9yeJoZcX5LJxWyKSCHFZvqqc77Fy3YgYfu3ThEd+AIxHn11v2cOdT1bxa38y8KePIzDCqG1tZOLWQj122kLctLT2uFUHdnV9uqOeLv9xyxC54EwuyWThtPOVTC+noClPVGE0W7d3hY55j+axi3rZ0Om89rZRZkwpo6ejmtgdf4nevhXj/uXP47DWLyY6zC197V5gHnt/J/et2MqM4n6uXlnLVadOZUhh/MMCufW08samexzfu4ZVdB5g0LoeVF83nfefMYVzukZXePQc7+MyjG3lqayNnzpnIHW9fwjNVob7EcMkpJdx++SJWzCqmvSvMD//0Bnc9XUOopZOz503i45f3nyh6whF27mujKiaRbmtoYeueFr56wzL+4gRrnSKppgQxgIPt3Xx59at88qpTmTguZ8jj6uwJ82xVE6/sOsCsSQUsmjaeBVMLKYz5cGts6WDV09u5//mddPVEeMeKMj56STlVDS3c+VQVW/e0MH/KOD52WTlvXzYDM2P1xnq++VQVVY2tLCgZx22XLeSaZTMGTRShlk4++7NNPLl5D8tnFvHnZ87sqyXE+5CORJy6A+1sa4h+KOZkZfCWJdPiLoEejjhfeXIrq57Zzvnlk/nWX55BcUH037Q3Mdz19HaaWjupmDORfYe62N50iAyDs+dN5uplpVy1ZDod3eG+pqsNtdFaymllE3jHijL+8uzZA64p5O787OU67nhsCwfbuwG47NSp3HbZwrjrFnV0h3nwT2/wnd/X0NjSSVF+NvH+CQ91humK2du5rDifRdMKOXPORP7mzQviJkOR0UAJYpQItXSy6pkafrBuZ98aPPNLxnHbpQt5+/JjP/wjEeeJTXu486ltbGtoZfakAt6+vJSrl5ayuPTIpRncncde2c3nH9tMW1eYv79iEX99wbwB99o+UQ9X7uIzj25iRnEe/33jGTy3vYlVz2ynqbWL88snc/tlizhr3iTcna17Wli9sZ7HN9azPXQIM+h9Gy4tK+LqpaVcvXQ6cyb33/wWT2NzBz95sZYLyqewbObgC9p1dId5eH0tVQ3xF1wsyMmifGq0qW1BSeExtReR0UoJYpRpau3koRd2MWtSQULNR5GI8+TmPdy/bifrtu8l4jB3ckHw4VrK1PG5fOZnm1izpYEVs4r5z79YlvTx+pU79vGR+9fT1NoFwAXlU7j98oX8WT8jfdyd1xpa+NWmBvJzMvqarkQkuZQg0sje1k5+tTm6Z8Bz2/cSjjiZGUZWhvGJKxdx8wXzT3gHs+NVu7+N7z37Om9bWqohoCIjlBJEmtp3qItfbd7Dq/XN3HTeXBaUFKY6JBEZYRJNEGpUHWMmjcvhRm0vKSJDQMMwREQkLiUIERGJa8QlCDO7ysxeM7NqM/tUquMREUlXIypBmFkm8C3grcBiotuQLk5tVCIi6WlEJQjgLKDa3be7exfwI+C6FMckIpKWRlqCKAN2xdyvDcpERGSYjbQEMSgzW2lmlWZWGQqFUh2OiMiYNdISRB0Qu0TmzKCsj7uvcvcKd68oKSkZ1uBERNLJiJpJbWZZwDbgMqKJ4QXgL919cz/Hh4CdJ/GSU4Cmkzh/tNJ1pxddd3pJ5LrnuPug37BH1Exqd+8xs1uBXwGZwD39JYfg+JOqQphZZSLTzccaXXd60XWnl6G87hGVIADcfTWwOtVxiIiku5HWByEiIiNEuieIVakOIEV03elF151ehuy6R1QntYiIjBzpXoMQEZF+pGWCSJcFAc3sHjNrNLNNMWWTzGyNmVUFvyemMsZkMLNZZvY7M9tiZpvN7PagfExfu5nlmdmfzOyV4Lq/EJTPM7Png/f7Q2aWk+pYk8HMMs3sJTP7ZXA/Xa57h5ltNLOXzawyKBuS93raJYg0WxDwXuCqo8o+BTzl7guBp4L7Y00P8Al3XwycA9wS/I3H+rV3Ape6+3JgBXCVmZ0DfAX4hruXA/uBm1MYYzLdDrwacz9drhvgEndfETO8dUje62mXIEijBQHd/Rlg31HF1wH3BbfvA94xrEENA3evd/cXg9stRD80yhjj1+5RrcHd7ODHgUuBnwTlY+66AcxsJvA24H+D+0YaXPcAhuS9no4JIt0XBJzm7vXB7T3AtFQGk2xmNhc4HXieNLj2oJnlZaARWAPUAAfcvSc4ZKy+3/8L+CcgEtyfTHpcN0S/BPzazNab2cqgbEje6yNuopwMH3d3Mxuzw9jMrBB4BPi4uzdHv1RGjdVrd/cwsMLMioFHgVNTHFLSmdk1QKO7rzezi1MdTwpc4O51ZjYVWGNmW2MfPJn3ejrWIAZdEHCMazCzUoDgd2OK40kKM8smmhwecPefBsVpce0A7n4A+B1wLlAcrHMGY/P9fj5wrZntINpkfClwJ2P/ugFw97rgdyPRLwVnMUTv9XRMEC8AC4MRDjnAe4DHUhzTcHoMuCm4fRPw8xTGkhRB+/P3gFfd/esxD43pazezkqDmgJnlA1cQ7X/5HXBDcNiYu253/7S7z3T3uUT/P//W3f+KMX7dAGY2zszG994GrgQ2MUTv9bScKGdmVxNts+xdEPBLKQ4pKczsQeBioqs7NgCfB34G/BiYTXQl3He5+9Ed2aOamV0ArAU2crhN+p+J9kOM2Ws3s2VEOyQziX75+7G7/4uZzSf6zXoS8BLwXnfvTF2kyRM0Mf2Du1+TDtcdXOOjwd0s4Ifu/iUzm8wQvNfTMkGIiMjg0rGJSUREEqAEISIicSlBiIhIXEoQIiISlxKEiIjEpQQhIiJxKUGIiEhcShAiIhLX/wf9r7N3ev43RQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(score_history)\n",
    "plt.ylabel('score')\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}