RL_Summary.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Display current agent performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Import libraries:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\python36\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import gym\n",
    "gym.logger.set_level(40)\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Watch agents play an episode of each game"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's define a list of all the games for which we have a trained agent ready"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "game_name = ['CartPole', 'MountainCar', 'LunarLander', 'Acrobot']\n",
    "version = ['v0', 'v0', 'v2', 'v1']\n",
    "colors = ['b', 'g', 'r', 'c', 'm', 'y', 'k', 'w']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now for each game, load the trained model and run one episode. I'm adding a limit for the maximal number of steps as some of the games can have very long episodes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from models/CartPole/model.ckpt\n",
      "Game ended after 1001 steps\n",
      "Total reward earned in this episode: 1000.00\n",
      "INFO:tensorflow:Restoring parameters from models/MountainCar/model.ckpt\n",
      "Game ended after 609 steps\n",
      "Total reward earned in this episode: -608.00\n",
      "INFO:tensorflow:Restoring parameters from models/LunarLander/model.ckpt\n",
      "Game ended after 345 steps\n",
      "Total reward earned in this episode: 258.06\n",
      "INFO:tensorflow:Restoring parameters from models/Acrobot/model.ckpt\n",
      "Game ended after 162 steps\n",
      "Total reward earned in this episode: -160.00\n"
     ]
    }
   ],
   "source": [
    "for i in range(len(game_name)):\n",
    "    \n",
    "    ep_reward = []\n",
    "    env = gym.make(game_name[i] + '-' + version[i]) # Create an instance of the game environment\n",
    "    env._max_episode_steps = 1000 # Limit maximal number of steps per episode\n",
    "\n",
    "    with tf.Session() as sess:\n",
    "\n",
    "        '''Load model'''\n",
    "        saver = tf.train.import_meta_graph(\"models/\" + game_name[i] + \"/model.ckpt.meta\") # import saved graph\n",
    "        saver.restore(sess, \"models/\" + game_name[i] + \"/model.ckpt\") # load model parameters\n",
    "        graph = tf.get_default_graph() # create a graph instance containing all the tensors\n",
    "        action_prob_dist = graph.get_tensor_by_name(\"Action_PDF/softamx:0\") # get the action PDF output tensor\n",
    "        input_ = graph.get_tensor_by_name(\"input:0\") # get the input tensor\n",
    "\n",
    "        '''Play an episode according to the model'''\n",
    "        obs = env.reset() # Reset env and save observation\n",
    "        t = 0\n",
    "        while True:\n",
    "            env.render() # Render game\n",
    "            # Use our model to create a probability distribution of actions based on observation\n",
    "            apd = np.squeeze(sess.run(action_prob_dist, feed_dict={input_ : obs.reshape((1,len(obs)))}))\n",
    "            # Choose an action out of the PDF and take action\n",
    "            action = np.random.choice(np.arange(len(apd)), p = apd)\n",
    "            obs, reward, done, info = env.step(action)\n",
    "            ep_reward.append(reward) # Append step reward to episode rewards list\n",
    "            t = t+1\n",
    "            if done:\n",
    "                print(\"Game ended after {} steps\".format(t+1))\n",
    "                print(\"Total reward earned in this episode: {:0.2f}\".format(sum(ep_reward)))\n",
    "                break\n",
    "    env.close()\n",
    "    tf.reset_default_graph() # Reset graph for allowing next model to be loaded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate agents performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Game solving thresholds (from [OpenAI wiki](https://github.com/openai/gym/wiki/Leaderboard)):\n",
    "\n",
    "* CartPole-v0 defines \"solving\" as getting average reward of 195.0 over 100 consecutive trials\n",
    "* MountainCar-v0 defines \"solving\" as getting average reward of -110.0 over 100 consecutive trials\n",
    "* LunarLander-v2 defines \"solving\" as getting average reward of 200 over 100 consecutive trials\n",
    "* Acrobot-v1 is an unsolved environment, which means it does not have a specified reward threshold at which it's considered solved\n",
    "\n",
    "\n",
    "In order to get a better sense of the agets performance we need more statisics. We'll let each agent paly 100 episodes, compute the mean and std, and plot a histogram of the results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean episode reward in CartPole is: 1000 ± 0\n",
      "Mean episode reward in MountainCar is: -646 ± 166\n",
      "Mean episode reward in LunarLander is: 226 ± 89\n",
      "Mean episode reward in Acrobot is: -153 ± 35\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEYCAYAAAAJeGK1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAGwVJREFUeJzt3X2YbXVd9/H3J44IclBA5IiAQMkxBZXkaJJpg2ilJWDhnU958Oa6ue4MH/JOpfJKSy3swadS84QKmZ2Doga3Smp4JrU7MBBUHjIQH0AIOIHYID7h9/5jrclxZs/MPrPn4cfM+3Vd+5q911p7re/6zZ792eu31/xWqgpJklrzYytdgCRJgxhQkqQmGVCSpCYZUJKkJhlQkqQmGVCSpCYZUJKkJhlQkqQmGVCrQJIrkoytdB1LKcmDk1ya5L+SvHCl61lqSb6S5IkrXYe0kgyoxg16o0pyUpJPTz6uqsOranxn13M38zJgvKr2rKo3L8YKkzwrycVJJpLcmOT8JD87wvp+pI37x3f2678pybuSrF+M2luT5NS+Lb+T5Mxp8/ZJ8sEkdyT5apJnLeb8ldZ6fXdnBpRGlmTdMmzmYOCKnX3SbLUleQnwRuCPgA3AA4G3Ascv1jZ6T62q9cAjgUcBr9jZ9d9N3AC8BnjngHlvAb5L187PBt6W5PBFnL/SWq/v7quqvDV8A74CPHHatJOATw9aBng58HXgv4AvAscC7wZ+ANwJTAAv65d9CDAOfIPuzf+4Ket8JHBpv573AWcDr5m2zZcDnwe+A6wDTgO+1D/nSuBp05Z/ab/8HcA76P6gz++X/0dg71na4BPAXcC3+/o3zlP7jNqmre8+/XqePke7z7ovs2xj6/Q2nv67A/4U+NB8bT/gd/oA4P3ALcCXgRfuxOvn94C3TXm8N/A9YLcler2+BjhzyuM96N68N06Z9m7g9MWYP0Q99wBe27fn94Dqb59bpP0dqT5vc988glpFkjwYOBV4VFXtCfwC8JWq+nXga/Sf5qvqT5LcA/i/wMeA/YAXAO/pv+vZFfggcCawD92b79MGbPKZwC8Be1XV9+ne0B9HFwB/APxtkv2nLP+rwJPoAuapdOH0u8C+dEfzA79bqqonAJ8CTq3uaOTLs9U+R21THQ3s1u/jbObbl+nbeCbT2njqgkkOAp4CXDpX208vIsmP9ct+DjiA7gPHi5P8wpRl3prkrbPsx8OAy6Y8PhL4YlV9e9DCST6U5Buz3D40yzbmshG4q6r+fcq0zwGHL9L8+byGrs0eB+wFXED3e5/xel7gvo9an+awHF0zGt3fJ5n6Jrsr8NkBy90F3BN4aJJbquorc6zzMcB6uk96PwA+0f8RPpPuiGUd8ObqPhJ+IMlnBqzjzVV13eSDqnrflHlnJ/kd4NHAuf20v6iqmwCSfAq4uaou7R9/kO6NZBhz1f6qQbVNc19gx4Dg+m9D7Mt825g0+bu7HfgwXZfiMPVPehRwv6r6w/7xtUn+GngG8NG+1ufPsf2HAW+Y8vhIujfQgarql+fZn521nm7fp7od2HOR5s8qyZ50H3oePvl7SvJ+4Neq6trpyy9w3xdcn+bnEdTdwwlVtdfkDRj4hlRV1wAvpnuTuznJtiQPmGWdDwCu698gJ32V7lP6A4Cv9+E0adAb8Y9MS/LcJJdNfuoEjqA7Opp005T7dw54POwJBHPVPle9k/4T2Heu746G2Jf5tjFp8nd3cFU9v6ruHLL+SQcDD5j6aZ7uqHPDfBvuj4R/AvjClMmP4EePqJbaBHDvadPuTdd1uhjz5/J44NqqunrKtL2B/xjiucMapT7Nw4BaZarq76rqZ+ne2Ap43eSsaYveABzUdyFNeiDd91c3AgckyZR5Bw3a3OSdJAcDf03XxXjfPkgvBzLgeaOaq/YZtQ3wL3TfZ50waOZO7Mv0bQx7cbVh6p90HfDlqR9QqjuT8SlDbOehdB80vgXQ/z7HmOMIqj+TcWKW2/lD7t9U/w6sS3LYlGmP4IcnvIw6fy73A26bfNDv/9OAgd11C9z3UerTPAyoVaT//ugJSe5J9wZ8J123H3RHKz8+ZfGL6E5WeFmSe6T7P6qnAtvo3sDvAk5Nsi7J8XTdW3PZg+4N+pa+lufRHXUshblqn1dV3Q78PvCWJCckuVe/nicn+RMWvi/T23gx6v8M8M0kL0+ye5JdkhyR5FFDbOdhwH5JfiLJ7sCr6T64fGW2J1TVk/vv0Abdnjzb8/rXyW7ALsAuSXZLsq6q7gA+APxhkj2SPJbuTMl399sbaX6SMzPttPYpLgcemeTIfv//mO73evZi7ft89Wk0BtTqck/gdGAHXTfGfnTdQdD9cb6i7yb67ar6LnAc8OR++bcCz62qf+vn/QpwMt1ZZs+h+9T5ndk2XFVXAn9OF2430b05/vOi72G3rVlr34l1vB54Cd1p37fQHamcCvz9CPvyI228GPVX1V104XUk3ckhO4Az6E7eACDJXyX5qwGbehjd91TnA9f0+3It3Zl9i+0VdB+ITqN7vdzJD0+pfz6wO3Az3Qk3v1FVU48wRpl/ELP8bqrqYroz+D5Ct9/3B55SVd8baU9nmq9+LVB+9GsGabAkFwF/VVXvWulaNJy+W+qMqnr/SteyFPrv2D5HdxLEYoeOGuARlAZK8nNJ7t933WwGHg78w0rXpZ3yMOCqlS5iqVTVd6vqIYbT6uVp5prNg4H30p1Z9yXgxKq6cWVL0rCS7E3XxXv1fMtKrbKLT5LUJLv4JElNaqKLb999961DDjlkpctYNHfccQd77LHHSpfRFNtkMNtlJttkptXWJpdccsmOqrrffMs1EVCHHHIIF1988UqXsWjGx8cZGxtb6TKaYpsMZrvMZJvMtNraJMlXh1nOLj5JUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKT5g2oJO9McnOSy6dM2yfJx5Nc3f/cu5+eJG9Ock2Szyd55FIWL0lavYY5gjoT+MVp004DLqiqw4AL+sfQXT7gsP52CvC2xSlTkrTWzBtQVfVJ4NZpk48Hzurvn8UPr0x6PPA31bkQ2CvJ/otVrCRp7VjoSBIbJke2rqobk+zXTz+A7sJvk67vp80YBTvJKXRHWWzYsIHx8fEFltKeiYmJVbU/i8E2GWzHjgm2bBkfOG/jxuWtpRW+VmZaq22y2EMdZcC0gcOlV9UWYAvApk2bajUN47HahiVZDLbJYFu2jLN169jAedu3L28trfC1MtNabZOFnsV302TXXf/z5n769XSXYJ50IHDDwsuTJK1VCw2o84DN/f3NwLlTpj+3P5vvMcDtXuROkrQQ83bxJdkKjAH7JrkeeCVwOvDeJCcDXwOe3i/+EeApwDXAt4DnLUHNkqQ1YN6AqqpnzjLr2AHLFvCboxYlSZIjSUiSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaNFJAJfmtJFckuTzJ1iS7JTk0yUVJrk5ydpJdF6tYSdLaseCASnIA8EJgU1UdAewCPAN4HfCGqjoMuA04eTEKlSStLaN28a0Ddk+yDrgXcCPwBOCcfv5ZwAkjbkOStAalqhb+5ORFwGuBO4GPAS8CLqyqB/XzDwLO74+wpj/3FOAUgA0bNhy1bdu2BdfRmomJCdavX7/SZTTFNhlsx44Jbr11cLts3LjMxTTC18pMq61NjjnmmEuqatN8y61b6AaS7A0cDxwKfAN4H/DkAYsOTMCq2gJsAdi0aVONjY0ttJTmjI+Ps5r2ZzHYJoNt2TLO1q1jA+dt3768tbTC18pMa7VNRunieyLw5aq6paq+B3wA+Blgr77LD+BA4IYRa5QkrUGjBNTXgMckuVeSAMcCVwLbgRP7ZTYD545WoiRpLVpwQFXVRXQnQ3wW+EK/ri3Ay4GXJLkGuC/wjkWoU5K0xiz4OyiAqnol8Mppk68FHj3KeiVJciQJSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpNGCqgkeyU5J8m/JbkqydFJ9kny8SRX9z/3XqxiJUlrx6hHUG8C/qGqfhJ4BHAVcBpwQVUdBlzQP5YkaacsOKCS3Bt4PPAOgKr6blV9AzgeOKtf7CzghFGLlCStPaMcQf04cAvwriSXJjkjyR7Ahqq6EaD/ud8i1ClJWmNSVQt7YrIJuBB4bFVdlORNwDeBF1TVXlOWu62qZnwPleQU4BSADRs2HLVt27YF1dGiiYkJ1q9fv9JlNMU2GWzHjgluvXVwu2zcuMzFNMLXykyrrU2OOeaYS6pq03zLjRJQ9wcurKpD+sePo/u+6UHAWFXdmGR/YLyqHjzXujZt2lQXX3zxgupo0fj4OGNjYytdRlNsk8G2bBln69axgfO2b1/eWlrha2Wm1dYmSYYKqAV38VXVfwDXJZkMn2OBK4HzgM39tM3AuQvdhiRp7Vo34vNfALwnya7AtcDz6ELvvUlOBr4GPH3EbUiS1qCRAqqqLgMGHaYdO8p6JUlyJAlJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTRg6oJLskuTTJh/rHhya5KMnVSc5OsuvoZUqS1prFOIJ6EXDVlMevA95QVYcBtwEnL8I2JElrzEgBleRA4JeAM/rHAZ4AnNMvchZwwijbkCStTamqhT85OQf4Y2BP4LeBk4ALq+pB/fyDgPOr6ogBzz0FOAVgw4YNR23btm3BdbRmYmKC9evXr3QZTbFNBtuxY4Jbbx3cLhs3LnMxjfC1MtNqa5NjjjnmkqraNN9y6xa6gSS/DNxcVZckGZucPGDRgQlYVVuALQCbNm2qsbGxQYvdLY2Pj7Oa9mcx2CaDbdkyztatYwPnbd++vLW0wtfKTGu1TRYcUMBjgeOSPAXYDbg38EZgryTrqur7wIHADaOXKUlaaxb8HVRV/U5VHVhVhwDPAD5RVc8GtgMn9ottBs4duUpJ0pqzFP8H9XLgJUmuAe4LvGMJtiFJWuVG6eL7b1U1Doz3968FHr0Y65UkrV2OJCFJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlq0oIDKslBSbYnuSrJFUle1E/fJ8nHk1zd/9x78cqVJK0VoxxBfR/4P1X1EOAxwG8meShwGnBBVR0GXNA/liRppyw4oKrqxqr6bH//v4CrgAOA44Gz+sXOAk4YtUhJ0tqTqhp9JckhwCeBI4CvVdVeU+bdVlUzuvmSnAKcArBhw4ajtm3bNnIdrZiYmGD9+vUrXUZTbJPBduyY4NZbB7fLxo3LXEwjfK3MtNra5JhjjrmkqjbNt9y6UTeUZD3wfuDFVfXNJEM9r6q2AFsANm3aVGNjY6OW0ozx8XFW0/4sBttksC1bxtm6dWzgvO3bl7eWVvhamWmttslIZ/EluQddOL2nqj7QT74pyf79/P2Bm0crUZK0Fo1yFl+AdwBXVdXrp8w6D9jc398MnLvw8iRJa9UoXXyPBX4d+EKSy/ppvwucDrw3ycnA14Cnj1aiJGktWnBAVdWngdm+cDp2oeuVJAkcSUKS1CgDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktSkJQmoJL+Y5ItJrkly2lJsQ5K0ui16QCXZBXgL8GTgocAzkzx0sbcjSVrdluII6tHANVV1bVV9F9gGHL8E25EkrWLrlmCdBwDXTXl8PfDT0xdKcgpwSv9wIskXl6CWlbIvsGOli2iMbTLYrO2SLHMl7fC1MtNqa5ODh1loKQJq0J9VzZhQtQXYsgTbX3FJLq6qTStdR0tsk8Fsl5lsk5nWapssRRff9cBBUx4fCNywBNuRJK1iSxFQ/wocluTQJLsCzwDOW4LtSJJWsUXv4quq7yc5FfgosAvwzqq6YrG307hV2XU5IttkMNtlJttkpjXZJqma8fWQJEkrzpEkJElNMqAkSU0yoBZBkn2SfDzJ1f3PvedY9t5Jvp7kL5ezxuU2TJskOTLJvyS5Isnnk/zaStS61OYb+ivJPZOc3c+/KMkhy1/l8huiXV6S5Mr+tXFBkqH+d+bubNhh4pKcmKSSrOpTzw2oxXEacEFVHQZc0D+ezauBf1qWqlbWMG3yLeC5VXU48IvAG5PstYw1Lrkhh/46Gbitqh4EvAF43fJWufyGbJdLgU1V9XDgHOBPlrfK5TXsMHFJ9gReCFy0vBUuPwNqcRwPnNXfPws4YdBCSY4CNgAfW6a6VtK8bVJV/15VV/f3bwBuBu63bBUuj2GG/praVucAxyarfhyJedulqrZX1bf6hxfS/U/lajbsMHGvpgvrby9ncSvBgFocG6rqRoD+537TF0jyY8CfAy9d5tpWyrxtMlWSRwO7Al9ahtqW06Chvw6YbZmq+j5wO3DfZalu5QzTLlOdDJy/pBWtvHnbJMlPAQdV1YeWs7CVshRDHa1KSf4RuP+AWb835CqeD3ykqq5bLR+OF6FNJtezP/BuYHNV/WAxamvIMEN/DTU82Coz9D4neQ6wCfi5Ja1o5c3ZJv2H3DcAJy1XQSvNgBpSVT1xtnlJbkqyf1Xd2L/Z3jxgsaOBxyV5PrAe2DXJRFXdba+XtQhtQpJ7Ax8GXlFVFy5RqStpmKG/Jpe5Psk64D7ArctT3ooZaki0JE+k+8Dzc1X1nWWqbaXM1yZ7AkcA4/2H3PsD5yU5rqouXrYql5FdfIvjPGBzf38zcO70Barq2VX1wKo6BPht4G/uzuE0hHnbpB8K64N0bfG+ZaxtOQ0z9NfUtjoR+ESt/v+gn7dd+u6stwPHVdXADzirzJxtUlW3V9W+VXVI/z5yIV3brMpwAgNqsZwOPCnJ1cCT+sck2ZTkjBWtbOUM0yb/A3g8cFKSy/rbkStT7tLov1OaHPrrKuC9VXVFkj9Mcly/2DuA+ya5BngJc58FuioM2S5/Stfb8L7+tbGqx/Qcsk3WFIc6kiQ1ySMoSVKTDChJUpMMKElSkwwoSVKTDChJ0rySPL0f2PkHwwxSm+TgJJf0Z2BekeR/7+w2DShpgP7U3ln/EXkn1jOxGPUslyTjq32EbM0vyViSM6dNvhz4FeCTQ67mRuBnqupI4KeB05I8YGfqcCQJaYCq+v2VriHJLlV11xKuf13/vzfSvKrqKoDpQ7X1o7CfDowB9wTeUlVv7we8nXRPFnBA5BGU1oQkz0nymb674e39HxVJJpL8eZLP9tccul8//cwkJ/b3T59yXaI/66cd3C8/ea2iB/bTD013jat/TfLqaTW8tJ/++SR/MEudE/3R20XA0UmOSvJPfVfJR5Psn2S/JJf0yz+ivy7Q5Pa/lOReSZ6a7tpSlyb5xyQb+vmvSrIlyceAv0mye5JtfU1nA7svQfNrdTsZuL2qHgU8CvhfSQ4FSHJQks/TDYL7uv6qBUMzoLTqJXkI8GvAY/vuhruAZ/ez9wA+W1WPpLtO1yunPXcf4GnA4f11iV7Tz/pLuiGaHg68B3hzP/1NwNv6P9b/mLKenwcOo7ukwpHAUUkeP6DcPYDLq+qn6a738xfAiVV1FPBO4LX9sD+79eMYPg64mG6cx4OBm/tLVHwaeExV/RTdZRteNmUbRwHHV9WzgN8AvtXvx2v7eVqj+g81lwFnAMdNGeHlF+Z42s8Dz+2fdxHdSPyHAVTVdf1r60HA5skPSsOyi09rwbF0b7z/2ndP7M4PB6/9AXB2f/9vgQ9Me+436a67c0aSDwOTlzk4mq4/HrqR2CcvpvdY4FenTJ+8+ODP97dL+8fr6f6Ip/fn3wW8v7//YLrBQT/e170LXb8+wP/rt/V44I/oLvgY4FP9/AOBs9MN1Lsr8OUp2zivqu7s7z+ePlyr6vP9p12tUf0HI5KMASdV1UlDPC3AC6rqo3Os94YkV9B9oDpn2HoMKK0FAc6qqt8ZYtkfGfurqr6f7lpVx9IN3nkq8IR5njdo/LAAf1xVb59n+9+e8r1TgCuq6ugBy32K7o/9YLqBeF/eb3cyQP8CeH1Vnde/2bxqynPvmKN2aWd9FPiNJJ+oqu8l2Qh8Hdgb+M+qujPJ3nQfqF6/Myu2i09rwQXAiUn2g67bru8Og+5v4MT+/rPousb+W5L1wH2q6iPAi+m656A7gnlGf//ZU573z9OmT/oo8D/79ZHkgMl65vBF4H5Jju6fc48kh/fzPgk8B7i6v4bWrcBT+u1Dd8mOr/f3NzO7T07WmeQI4OHz1KQ1KsnTklxP13vw4SSTR0xnAFcCn01yOd0I9OuAhwAXJfkcXff5n1XVF3Zmmx5BadWrqiuTvAL4WLqLvn0P+E3gq3RHE4f3Jx3cTvdd1VR7Aucm2Y3uiOa3+ukvBN6Z5KXALcDz+ukvAv4uyYv4YVcdVfWx/ruwf+m76yboAmbWy0hU1Xf7EzXenOQ+dH+vb6Q7qvpKv57JLsJPAwdW1W3941fRjQL+dbrLMhw6y2beBryr79q7DPjMbPVo7aiqcWB82rQP0l0eZ/qyPwB+t79N9XFG/MDjaOZa09JdNHL9StchaSa7+CRJTfIISpLUJI+gJElNMqAkSU0yoCRJTTKgJElNMqAkSU36/5SrQBzeeREiAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tf.logging.set_verbosity(tf.logging.ERROR) # Hide loading models INFO messages\n",
    "for i in range(len(game_name)):\n",
    "    sess = tf.Session()\n",
    "    \n",
    "    '''Define environment'''\n",
    "    env = gym.make(game_name[i] + '-' + version[i]) # Create an instance of the game environment\n",
    "    env._max_episode_steps = 1000 # Limit maximal number of steps per episode\n",
    "    \n",
    "    '''Load model'''\n",
    "    saver = tf.train.import_meta_graph(\"models/\" + game_name[i] + \"/model.ckpt.meta\") # import saved graph\n",
    "    saver.restore(sess, \"models/\" + game_name[i] + \"/model.ckpt\") # load model parameters\n",
    "    graph = tf.get_default_graph() # create a graph instance containing all the tensors\n",
    "    action_prob_dist = graph.get_tensor_by_name(\"Action_PDF/softamx:0\") # get the action PDF output tensor\n",
    "    input_ = graph.get_tensor_by_name(\"input:0\") # get the input tensor\n",
    "    \n",
    "    total_reward = []\n",
    "    for j in range(100):\n",
    "        ep_reward = []\n",
    "        '''Play an episode according to the model'''\n",
    "        obs = env.reset() # Reset env and save observation\n",
    "        t = 0\n",
    "        while True:\n",
    "            # Use our model to create a probability distribution of actions based on observation\n",
    "            apd = np.squeeze(sess.run(action_prob_dist, feed_dict={input_ : obs.reshape((1,len(obs)))}))\n",
    "            # Choose an action out of the PDF and take action\n",
    "            action = np.random.choice(np.arange(len(apd)), p = apd)\n",
    "            obs, reward, done, info = env.step(action)\n",
    "            ep_reward.append(reward) # Append step reward to episode rewards list\n",
    "            t = t+1\n",
    "            if done:\n",
    "                total_reward.append(np.sum(ep_reward)) # Append the total reward in this episode \n",
    "                break\n",
    "    sess.close()\n",
    "    mean_reward = int(np.mean(total_reward)) # Compute the mean reward per episode \n",
    "    std_reward = int(np.std(total_reward)) # Compute std of the reward over episodes\n",
    "    print(\"Mean episode reward in \" + game_name[i] + \" is: {:d} \".format(mean_reward) \n",
    "           + u\"\\u00B1\" + \" {:d}\".format(std_reward))\n",
    "    \n",
    "    '''Plot historagms'''\n",
    "    plt.figure()\n",
    "    plt.hist(total_reward, 50, facecolor = colors[i] ,alpha = 0.75)\n",
    "    plt.title(\"Histogram for \" + game_name[i] + \": \" + r\"$\\mu$ =\" + \" {:d}, \".format(mean_reward) \n",
    "               + r\"$\\sigma$ =\" + \" {:d}\".format(std_reward))\n",
    "    plt.xlabel(\"episode reward\")\n",
    "    plt.grid(True)\n",
    "    plt.tight_layout()\n",
    "    \n",
    "    tf.reset_default_graph() # Reset graph for allowing next model to be loaded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Action probability maps"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's try to gain some understanding of the policies our agents have learned. one way to do so is by ploting an action map over observation space, showing the probability of taking some action in each posion in observation space."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Mountain Car"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mountain car has 3 discrete actions (left, right and idle) and a 2D observation space of position and velocity."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can try to visualize the learned policy by ploting a 2D map of the probability to take some action in each point in the observation space, first we have to calculate the PDF based on our model for every point and then save the probability for each action in its own matrix:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos = np.linspace(-1.2, 0.6, 50) # position vector\n",
    "velocity = np.linspace(-0.07, 0.07, 50) # velocity vector\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    '''Load model'''\n",
    "    saver = tf.train.import_meta_graph(\"models/\" + game_name[1] + \"/model.ckpt.meta\") # import saved graph\n",
    "    saver.restore(sess, \"models/\" + game_name[1] + \"/model.ckpt\") # load model parameters\n",
    "    graph = tf.get_default_graph() # create a graph instance containing all the tensors\n",
    "    action_prob_dist = graph.get_tensor_by_name(\"Action_PDF/softamx:0\") # get the action PDF output tensor\n",
    "    input_ = graph.get_tensor_by_name(\"input:0\") # get the input tensor\n",
    "\n",
    "    '''Create action map'''\n",
    "    left_action_map = np.zeros((len(velocity),len(pos)))\n",
    "    right_action_map = np.zeros_like(left_action_map)\n",
    "    idle_action_map = np.zeros_like(left_action_map)\n",
    "    for i in range(len(velocity)):\n",
    "        for j in range(len(pos)):\n",
    "\n",
    "            obs = np.array([pos[j], velocity[i]]) # Create observation for this pixel\n",
    "            # Use model to create a probability distribution of actions based on observation\n",
    "            apd = np.squeeze(sess.run(action_prob_dist, feed_dict={input_ : obs.reshape((1,len(obs)))}))\n",
    "            left_action_map[i,j] = apd[0] # save the probability to move left to the map\n",
    "            idle_action_map[i,j] = apd[1] # save the probability to move idle to the map\n",
    "            right_action_map[i,j] = apd[2] # save the probability to move right to the map\n",
    "\n",
    "tf.reset_default_graph() # Reset graph for allowing next model to be loaded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can visualize the policy, we plot the action probability maps for each action and investigate the policy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.rcParams.update({'font.size': 18})\n",
    "'''Plot action map'''\n",
    "plt.figure(figsize = [15, 4])\n",
    "plt.subplot(1,3,1)\n",
    "plt.pcolormesh(pos, velocity, left_action_map)\n",
    "plt.colorbar()\n",
    "plt.title(\"Left Probability\")\n",
    "plt.xlabel(\"position\")\n",
    "plt.ylabel(\"velocity\")\n",
    "\n",
    "plt.subplot(1,3,2)\n",
    "plt.pcolormesh(pos, velocity, idle_action_map)\n",
    "plt.colorbar()\n",
    "plt.title(\"Idle Probability\")\n",
    "plt.xlabel(\"position\")\n",
    "plt.ylabel(\"velocity\")\n",
    "\n",
    "plt.subplot(1,3,3)\n",
    "plt.pcolormesh(pos, velocity, right_action_map)\n",
    "plt.colorbar()\n",
    "plt.title(\"Right Probability\")\n",
    "plt.xlabel(\"position\")\n",
    "plt.ylabel(\"velocity\")\n",
    "\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can learn from these plots that the agent is on the right path, but didn't quite figure the optimal policy. Our physical intuition tells us that the best way to go would be moving left when the velocity is to the left and moving right when the velocity is to the right in order to gain as much velocity as posible. Our agent got the main idea, but doesn't maximize its potential since it has higher probability to stay idle rather than move left when the car velocity is negative."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cart Pole"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Cart Pole has two discrete actions (left or right) and a 4D observation space (position, velocity, pole angle, pole velocity at the tip). Since we only have two possible actions, we get all the information we need by ploting the right action probability (the left porb. is just 1 - right). We can only plot an action map over a 2D projection of the observation space, so we keep two constant and then scan the other two:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "l = 4\n",
    "a = 5\n",
    "v = np.linspace(-l, l, 50) # velocity vector\n",
    "pv = np.linspace(-l, l, 50) # pole velocity vector\n",
    "pa = np.linspace(-a, a, 50) # pole angle vector\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    '''Load model'''\n",
    "    saver = tf.train.import_meta_graph(\"models/\" + game_name[0] + \"/model.ckpt.meta\") # import saved graph\n",
    "    saver.restore(sess, \"models/\" + game_name[0] + \"/model.ckpt\") # load model parameters\n",
    "    graph = tf.get_default_graph() # create a graph instance containing all the tensors\n",
    "    action_prob_dist = graph.get_tensor_by_name(\"Action_PDF/softamx:0\") # get the action PDF output tensor\n",
    "    input_ = graph.get_tensor_by_name(\"input:0\") # get the input tensor\n",
    "\n",
    "    '''Create action map'''\n",
    "    vpv_action_map = np.zeros((len(pv),len(v)))\n",
    "    papv_action_map = np.zeros_like(vpv_action_map)\n",
    "    for i in range(len(pv)):\n",
    "        for j in range(len(v)):\n",
    "\n",
    "            obs = np.array([0, v[j], 0, pv[i]]) # Create observation for this pixel\n",
    "            # Use model to create a probability distribution of actions based on observation\n",
    "            apd = np.squeeze(sess.run(action_prob_dist, feed_dict={input_ : obs.reshape((1,len(obs)))}))\n",
    "            vpv_action_map[i,j] = apd[1] # save the probability to move right to the map\n",
    "        \n",
    "        for j in range(len(pa)):\n",
    "            obs = np.array([0, 0, pa[j], pv[i]]) # Create observation for this pixel\n",
    "            # Use model to create a probability distribution of actions based on observation\n",
    "            apd = np.squeeze(sess.run(action_prob_dist, feed_dict={input_ : obs.reshape((1,len(obs)))}))\n",
    "            papv_action_map[i,j] = apd[1] # save the probability to move right to the map\n",
    "\n",
    "tf.reset_default_graph() # Reset graph for allowing next model to be loaded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.rcParams.update({'font.size': 16})\n",
    "'''Plot action map'''\n",
    "plt.figure(figsize = [10, 4])\n",
    "plt.subplot(1,2,1)\n",
    "plt.pcolormesh(v, pv, vpv_action_map)\n",
    "plt.colorbar()\n",
    "plt.title(\"Right probability velocities\")\n",
    "plt.xlabel(\"cart velocity\")\n",
    "plt.ylabel(\"pole velocity\")\n",
    "\n",
    "plt.subplot(1,2,2)\n",
    "plt.pcolormesh(pa, pv, papv_action_map)\n",
    "plt.colorbar()\n",
    "plt.title(\"Right probability pole parameters\")\n",
    "plt.xlabel(\"pole angle\")\n",
    "plt.ylabel(\"pole velocity\")\n",
    "\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we see that the agent learned basic physical intuition: when the pole is falling right move right, otherwise move left. If the pole has enough velocity to the right, move right even though the pole is currently tilted to the left and vice versa.\n",
    "\n",
    "Moreover, we see that the probabiliy map is almost a deterministic map - comprises only 1 or 0 probability. "
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}