Approximation/Average Reward Semi-Gradient Sarsa.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Average Reward Linear SARSA\n",
    "\n",
    "This is the implementation of Average Reward Linear SARSA. Instead of applying discount on rewards, we optimize another parameter which is the average rate of rewards. Note that this reward rate is calculated as taking the limit to infinity so this is for continuing tasks. I applied it on CartPole but it is wrong."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import autograd.numpy as np\n",
    "import gym\n",
    "import matplotlib.pyplot as plt\n",
    "from autograd import grad, elementwise_grad\n",
    "import random\n",
    "from IPython import display"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Linear approximation function to expected returns\n",
    "def approx(weights, state, action):\n",
    "    return np.dot(state, weights)[action]\n",
    "\n",
    "def policy(env, weights, state, epsilon):\n",
    "    actions = [0, 1]\n",
    "    if np.random.rand() < epsilon:\n",
    "        return random.choice(actions)\n",
    "    qs = []\n",
    "    for action in actions:\n",
    "        qs.append(approx(weights, state, action))\n",
    "    return np.argmax(qs)\n",
    "\n",
    "dapprox = grad(approx)\n",
    "episodes = 1000\n",
    "beta = 0.01\n",
    "epsilon = 0.1\n",
    "alpha = 0.2\n",
    "avg_reward = 0.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[33mWARN: gym.spaces.Box autodetected dtype as <class 'numpy.float32'>. Please provide explicit dtype.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "env = gym.make('CartPole-v0')\n",
    "episode_rewards = []\n",
    "w = np.zeros((4,2))\n",
    "for e in range(episodes):\n",
    "    state = env.reset()\n",
    "    rewards = []\n",
    "    while True:\n",
    "        action = policy(env, w, state, epsilon)\n",
    "        q_hat = approx(w, state, action)\n",
    "        q_hat_grad = dapprox(w, state, action)\n",
    "        next_state, reward, done, _ = env.step(action)\n",
    "        rewards.append(reward)\n",
    "        if done:\n",
    "            break\n",
    "        next_action = policy(env, w, next_state, epsilon)\n",
    "        q_hat_next = approx(w, next_state, next_action)\n",
    "        del_t = reward - avg_reward + q_hat_next - q_hat\n",
    "        avg_reward += beta * del_t\n",
    "        w += alpha*del_t*q_hat_grad\n",
    "        state = next_state\n",
    "    episode_rewards.append(np.sum(rewards))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i, r in enumerate(episode_rewards):\n",
    "    if i + 100 >= len(episode_rewards):\n",
    "        break\n",
    "    else:\n",
    "        mean = np.mean(episode_rewards[i:i+100])\n",
    "        if mean >= 195:\n",
    "            print(\"Episodes before solve\", i+1)\n",
    "            break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11499a828>]"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10f6dfd30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(episode_rewards)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10f5d5a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAAEICAYAAAC+iFRkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAFiVJREFUeJzt3X+w5XV93/HnK6ygorIL3m43u9sBx63UZAyutxZG41AJCtRx6YxhoE5YzXbWNLTR6EzEOG3jTKeDbSYonQ66ismaEBWJdncIjeKi7dgW9KK4Iki4EnB3C+xFYf2B0aDv/nE+K4fbhXvOvefu2XO/z8fMmfP5fr6fc87nw3d53e/5fH+cVBWSpJXtF8bdAUnS8jPsJakDDHtJ6gDDXpI6wLCXpA4w7CWpAwx7SeoAw14TJckrk/zvJIeSfDfJ/0ryj9u6NyX54lHsy0WtL48l+cIR1p+R5La2/rYkZxytvknzGfaaGEmeB9wA/BfgZGA98B7gx2Pq0neB9wFXzF+R5HhgF/BnwBpgJ7Cr1UtHnWGvSfIPAarqY1X106r6UVV9tqr2JvlHwAeAs5L8IMmjAElOSPKHSb6d5KEkH0jyrLbu7CT7k/x+koeT3JfkjYN2pqo+V1XXAf/3CKvPBlYB76uqH1fVVUCAVy/pv4C0SIa9JslfAz9NsjPJ+UnWHF5RVXcBvwX8n6p6TlWtbquuoPdH4gzghfS+Dfy7vvf8+8DzW/1WYEeSFwEk+RdJ9i6yr78E7K0n349kb6uXjjrDXhOjqr4HvBIo4EPAXJLdSdYeqX2SANuB362q71bV94H/CFw8r+m/bXvf/wP4S+Ci9nl/XlUvWWR3nwMcmld3CHjuIt9PWpJV4+6ANIy2B/8mgCSn05sTfx9wyRGaTwHPBm7r5T7Qm0o5rq/NI1X1w77l+4FfHEFXfwA8b17d84Dvj+C9paG5Z6+JVVXfBP4E+OXDVfOaPAz8CPilqlrdHidV1XP62qxJcmLf8j/gyHPww/oG8JL0/ZUBXtLqpaPOsNfESHJ6knck2dCWN9Lbo7+lNXkI2HD4jJeq+hm96Z4rk/y99pr1SV47763fk+T4JL8KvA745ID9OS7JM+l9Q/6FJM9M8oy2+gvAT4HfaQeJ/3Wrv3n4kUtLZ9hrknwf+CfArUl+SC/k7wDe0dbfTG/P+cEkD7e6dwKzwC1Jvgd8DnhR33s+CDxCb2/+WuC32jcGkrwxydPtif8GvW8OVwO/2sofAqiqnwAXApcCjwK/CVzY6qWjLv54iboqydnAn1XVhnH3RVpu7tlLUgcsS9gnOS/J3Ulmk1y+HJ8hSRrcyKdxkhxH7+KXc4H9wJeBS6rqzpF+kCRpYMuxZ/9yYLaq7m0Hoz4ObFmGz5EkDWg5LqpaD+zrW95P7wyKJ0mynd7VjZx44okvO/3005ehK5I0me677z4efvjhLNxyMGO7graqdgA7AKanp2tmZmZcXZGkY8709PRI3285pnEOABv7lje0OknSmCxH2H8Z2JTktHYl48XA7mX4HEnSgEY+jVNVj7dLwz9D74ZTH6kq7wciSWO0LHP2VXUjcONyvLckaXheQStJHWDYS1IHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHWDYS1IHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBC4Z9ko8kOZjkjr66k5PclOSe9rym1SfJVUlmk+xNsnk5Oy9JGswge/Z/Apw3r+5yYE9VbQL2tGWA84FN7bEduHo03ZQkLcWCYV9V/xP47rzqLcDOVt4JXNhX/9HquQVYnWTdqDorSVqcxc7Zr62qB1r5QWBtK68H9vW129/q/j9JtieZSTIzNze3yG5Ikgax5AO0VVVALeJ1O6pquqqmp6amltoNSdLTWGzYP3R4eqY9H2z1B4CNfe02tDpJ0hgtNux3A1tbeSuwq6/+0nZWzpnAob7pHknSmKxaqEGSjwFnA89Psh/498AVwHVJtgH3Axe15jcCFwCzwGPAm5ehz5KkIS0Y9lV1yVOsOucIbQu4bKmdkiSNllfQSlIHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHWDYS1IHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBC4Z9ko1JPp/kziTfSPLWVn9ykpuS3NOe17T6JLkqyWySvUk2L/cgJElPb5A9+8eBd1TVi4EzgcuSvBi4HNhTVZuAPW0Z4HxgU3tsB64eea8lSUNZMOyr6oGq+korfx+4C1gPbAF2tmY7gQtbeQvw0eq5BVidZN3Iey5JGthQc/ZJTgVeCtwKrK2qB9qqB4G1rbwe2Nf3sv2tbv57bU8yk2Rmbm5uyG5LkoYxcNgneQ7wF8Dbqup7/euqqoAa5oOrakdVTVfV9NTU1DAvlSQNaaCwT/IMekF/bVV9qlU/dHh6pj0fbPUHgI19L9/Q6iRJYzLI2TgBrgHuqqo/6lu1G9jayluBXX31l7azcs4EDvVN90iSxmDVAG1eAfwG8PUkt7e63weuAK5Lsg24H7iorbsRuACYBR4D3jzSHkuShrZg2FfVF4E8xepzjtC+gMuW2C9J0gh5Ba0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHWDYS1IHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHTDID44/M8mXknwtyTeSvKfVn5bk1iSzST6R5PhWf0Jbnm3rT13eIUiSFjLInv2PgVdX1a8AZwDnJTkTeC9wZVW9EHgE2NbabwMeafVXtnaSpDFaMOyr5wdt8RntUcCrgetb/U7gwlbe0pZp689J8lQ/WC5JOgoGmrNPclyS24GDwE3At4BHq+rx1mQ/sL6V1wP7ANr6Q8ApR3jP7UlmkszMzc0tbRSSpKc1UNhX1U+r6gxgA/By4PSlfnBV7aiq6aqanpqaWurbSZKexlBn41TVo8DngbOA1UlWtVUbgAOtfADYCNDWnwR8ZyS9lSQtyiBn40wlWd3KzwLOBe6iF/pvaM22ArtaeXdbpq2/uapqlJ2WJA1n1cJNWAfsTHIcvT8O11XVDUnuBD6e5D8AXwWuae2vAf40ySzwXeDiZei3JGkIC4Z9Ve0FXnqE+nvpzd/Pr/9b4NdH0jtJ0kh4Ba0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHWDYS1IHDHJRlbRi3bbjLT8vv2z7B8fYE2l5GfbqnP6Al7rCaRx1jnvw6iLDXpI6wLCXGqd3tJIZ9pLUAYa9Osl5e3WNYS9JHWDYS1IHGPZSHw/SaqUy7CWpAwYO+yTHJflqkhva8mlJbk0ym+QTSY5v9Se05dm2/tTl6bq0NB6kVZcMs2f/VuCuvuX3AldW1QuBR4BtrX4b8Eirv7K1kySN0UBhn2QD8M+AD7flAK8Grm9NdgIXtvKWtkxbf05rL00E5+21Eg26Z/8+4PeAn7XlU4BHq+rxtrwfWN/K64F9AG39odb+SZJsTzKTZGZubm6R3ZckDWLBsE/yOuBgVd02yg+uqh1VNV1V01NTU6N8a0nSPIPs2b8CeH2S+4CP05u+eT+wOsnhWyRvAA608gFgI0BbfxLwnRH2WRoZD9KqKxYM+6p6V1VtqKpTgYuBm6vqjcDngTe0ZluBXa28uy3T1t9cVTXSXkuShrKU8+zfCbw9ySy9OflrWv01wCmt/u3A5UvronT0eZBWK81Qv1RVVV8AvtDK9wIvP0KbvwV+fQR9kySNiFfQqvOct1cXGPaS1AGGvSR1gGEvSR1g2EtPwTNytJIY9hIepNXKZ9hLUgcY9tLTcCpHK4VhL0kdYNhLUgcY9pLUAYa91HhGjlYyw16SOsCwlxbgGTlaCQx7SeoAw16SOsCwl/p4kFYrlWEvSR0wUNgnuS/J15PcnmSm1Z2c5KYk97TnNa0+Sa5KMptkb5LNyzkA6WjwIK0m3TB79v+0qs6oqum2fDmwp6o2AXt44ofFzwc2tcd24OpRdVaStDhLmcbZAuxs5Z3AhX31H62eW4DVSdYt4XOko8p5e61Eg4Z9AZ9NcluS7a1ubVU90MoPAmtbeT2wr++1+1vdkyTZnmQmyczc3Nwiui5JGtSgYf/KqtpMb4rmsiSv6l9ZVUXvD8LAqmpHVU1X1fTU1NQwL5XGwnl7TbKBwr6qDrTng8CngZcDDx2enmnPB1vzA8DGvpdvaHWSpDFZMOyTnJjkuYfLwGuAO4DdwNbWbCuwq5V3A5e2s3LOBA71TfdIksZgkD37tcAXk3wN+BLwl1X1V8AVwLlJ7gF+rS0D3AjcC8wCHwJ+e+S9lpaZB2m10qxaqEFV3Qv8yhHqvwOcc4T6Ai4bSe8kSSPhFbTSEDxIq0ll2EtSBxj2ktQBhr30FDxIq5XEsJekDjDspSF5kFaTyLCXpA4w7KWn4by9VgrDXpI6wLCXFsF5e00aw16SOsCwl6QOMOylBXiQViuBYa9OSzLQYymvfbr3kI4Ww15apJkPbl+4kXSMWPB+9pKecMMDTwT869btGGNPpOG4Zy8NqD/oj7QsHcsMe2kA029xL16TbaCwT7I6yfVJvpnkriRnJTk5yU1J7mnPa1rbJLkqyWySvUk2L+8QJEkLGXTP/v3AX1XV6fR+j/Yu4HJgT1VtAva0ZYDzgU3tsR24eqQ9lsZk/hy9c/aaJAseoE1yEvAq4E0AVfUT4CdJtgBnt2Y7gS8A7wS2AB9tPzx+S/tWsK6qHhh576Wj7A/+YPqJ8vi6IQ1tkD3704A54I+TfDXJh5OcCKztC/AHgbWtvB7Y1/f6/a1OmmjO22uSDRL2q4DNwNVV9VLghzwxZQNA24uvYT44yfYkM0lm5ubmhnmpJGlIg4T9fmB/Vd3alq+nF/4PJVkH0J4PtvUHgI19r9/Q6p6kqnZU1XRVTU9NTS22/5KkASwY9lX1ILAvyYta1TnAncBuYGur2wrsauXdwKXtrJwzgUPO10vSeA16Be2/Aa5NcjxwL/Bmen8orkuyDbgfuKi1vRG4AJgFHmttJUljNFDYV9XtwPQRVp1zhLYFXLbEfkmSRsgraCWpAwx7SeoAw16SOsBbHKvTeoeYpJXPPXtJ6gDDXpI6wLCXpA4w7CWpAwx7SeoAw16SOsCwl6QOMOwlqQMMe0nqAMNekjrAsJekDjDsJakDDHtJ6gDDXpI6wLCXpA5YMOyTvCjJ7X2P7yV5W5KTk9yU5J72vKa1T5Krkswm2Ztk8/IPQ5L0dBYM+6q6u6rOqKozgJcBjwGfBi4H9lTVJmBPWwY4H9jUHtuBq5ej45KkwQ07jXMO8K2quh/YAuxs9TuBC1t5C/DR6rkFWJ1k3Uh6K0lalGHD/mLgY628tqoeaOUHgbWtvB7Y1/ea/a3uSZJsTzKTZGZubm7IbkiShjFw2Cc5Hng98Mn566r3Q55D/ZhnVe2oqumqmp6amhrmpZKkIQ2zZ38+8JWqeqgtP3R4eqY9H2z1B4CNfa/b0OokSWMyTNhfwhNTOAC7ga2tvBXY1Vd/aTsr50zgUN90jyRpDFYN0ijJicC5wFv6qq8ArkuyDbgfuKjV3whcAMzSO3PnzSPrrSRpUQYK+6r6IXDKvLrv0Ds7Z37bAi4bSe8kSSPhFbSS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdYBhL0kdYNhLUgcY9pLUAYa9JHWAYS9JHWDYS1IHGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdcBAYZ/kd5N8I8kdST6W5JlJTktya5LZJJ9Icnxre0Jbnm3rT13OAUiSFrZg2CdZD/wOMF1VvwwcB1wMvBe4sqpeCDwCbGsv2QY80uqvbO0kSWM06DTOKuBZSVYBzwYeAF4NXN/W7wQubOUtbZm2/pwkGU13JUmLsWqhBlV1IMkfAt8GfgR8FrgNeLSqHm/N9gPrW3k9sK+99vEkh4BTgIf73zfJdmB7W/xxkjuWOJZj1fOZN/YVwnFNnpU6tpU6rheN8s0WDPska+jtrZ8GPAp8EjhvqR9cVTuAHe0zZqpqeqnveSxaqWNzXJNnpY5tJY9rlO83yDTOrwF/U1VzVfV3wKeAVwCr27QOwAbgQCsfADa2zq4CTgK+M8pOS5KGM0jYfxs4M8mz29z7OcCdwOeBN7Q2W4Fdrby7LdPW31xVNbouS5KGtWDYV9Wt9A60fgX4envNDuCdwNuTzNKbk7+mveQa4JRW/3bg8gH6sWP4rk+MlTo2xzV5VurYHNcA4k63JK18XkErSR1g2EtSB4w97JOcl+TudnuFQeb3jxlJNib5fJI72+0k3trqT05yU5J72vOaVp8kV7Wx7k2yebwjeHpJjkvy1SQ3tOUVcYuMJKuTXJ/km0nuSnLWSthmK+m2Jkk+kuRg//U3i9lGSba29vck2XqkzzqanmJc/7n9W9yb5NNJVvete1cb191JXttXP3xuVtXYHvRuvfAt4AXA8cDXgBePs09D9n8dsLmVnwv8NfBi4D8Bl7f6y4H3tvIFwH8HApwJ3DruMSwwvrcDfw7c0JavAy5u5Q8A/6qVfxv4QCtfDHxi3H1fYFw7gX/ZyscDqyd9m9G7mPFvgGf1bas3Teo2A14FbAbu6KsbahsBJwP3tuc1rbzmGBzXa4BVrfzevnG9uGXiCfSuc/pWy8xF5ea4N+hZwGf6lt8FvGvc/9CWMJ5dwLnA3cC6VrcOuLuVPwhc0tf+5+2OtQe9ayf20Lstxg3tf6SH+/5R/nzbAZ8BzmrlVa1dxj2GpxjXSS0UM69+orcZT1y5fnLbBjcAr53kbQacOi8Uh9pGwCXAB/vqn9TuWBnXvHX/HLi2lZ+Uh4e32WJzc9zTOD+/tULTf9uFidK+Br8UuBVYW1UPtFUPAmtbeZLG+z7g94CfteVTGPAWGcDhW2Qci04D5oA/blNUH05yIhO+zarqAHD4tiYP0NsGA9/WhGN7mx027DaaiG03z2/S+5YCIx7XuMN+RUjyHOAvgLdV1ff611XvT+9End+a5HXAwaq6bdx9WQar6H2NvrqqXgr8kHnXgkzoNuu/rckvAicygtuaHKsmcRstJMm7gceBa5fj/ccd9j+/tULTf9uFiZDkGfSC/tqq+lSrfijJurZ+HXCw1U/KeF8BvD7JfcDH6U3lvJ+VcYuM/cD+6l0sCL0LBjcz+dusC7c1GXYbTcq2I8mbgNcBb2x/yGDE4xp32H8Z2NTOGDie3oGi3WPu08CShN4Vw3dV1R/1req/ZcT8W0lc2s4eOBM41Pe19JhRVe+qqg1VdSq9bXJzVb2RFXCLjKp6ENiX5PAdBQ/f/mOitxnduK3JsNvoM8Brkqxp33xe0+qOKUnOozdl+vqqeqxv1W7g4nbm1GnAJuBLLDY3j4GDFRfQO4vlW8C7x92fIfv+SnpfJfcCt7fHBfTmPvcA9wCfA05u7QP81zbWr9P7QZixj2OBMZ7NE2fjvKD9Y5uld/fTE1r9M9vybFv/gnH3e4ExnQHMtO323+idqTHx2wx4D/BN4A7gT+mdxTGR2wz4GL1jD39H79vYtsVsI3pz4LPt8eZjdFyz9ObgD2fIB/rav7uN627g/L76oXPT2yVIUgeMexpHknQUGPaS1AGGvSR1gGEvSR1g2EtSBxj2ktQBhr0kdcD/A0WElEpXWKDwAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10f5d5a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def show_state(env, step=0):\n",
    "    plt.figure(3)\n",
    "    plt.clf()\n",
    "    plt.imshow(env.render(mode='rgb_array'))\n",
    "    plt.title(\"Step: %d\" % (step))\n",
    "    \n",
    "    plt.pause(0.001)  # pause for plots to update\n",
    "    \n",
    "    display.clear_output(wait=True)\n",
    "    display.display(plt.gcf())\n",
    "        \n",
    "done = False\n",
    "state = env.reset()\n",
    "step = 0\n",
    "while not done:\n",
    "    step += 1\n",
    "    show_state(env, step)\n",
    "    action = policy(env, w, state, 0.0)\n",
    "    state, _, done, _ = env.step(action)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}