MAB_failure.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Multi-armed bandits"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import librairies and modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from arm.Bernoulli import Bernoulli\n",
    "\n",
    "from algorithm.UCB import UCB\n",
    "from algorithm.TS import TS\n",
    "from posterior.Beta import Beta as BetaPosterior\n",
    "from algorithm.Greedy import Greedy\n",
    "\n",
    "from environment.MAB import MAB, EvaluationMAB"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Experimental setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "scenario = 0\n",
    "nb_rep = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "if scenario == 0:\n",
    "    horizon = 500\n",
    "    K = 2\n",
    "    env = MAB([Bernoulli(p) for p in [0.9, 0.1]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate various bandit algorithms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "algorithms = [UCB(K, c=0.5),\n",
    "              TS(K, BetaPosterior), \n",
    "              Greedy(K)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start evaluation of UCB\n",
      "0\n",
      "10\n",
      "20\n",
      "30\n",
      "40\n",
      "50\n",
      "60\n",
      "70\n",
      "80\n",
      "90\n",
      "Regret of UCB : 3.760000000003913 +- 0.9226050075736598 \n",
      "\n",
      "Start evaluation of TS\n",
      "0\n",
      "10\n",
      "20\n",
      "30\n",
      "40\n",
      "50\n",
      "60\n",
      "70\n",
      "80\n",
      "90\n",
      "Regret of TS : 2.7040000000037026 +- 1.1014463218877066 \n",
      "\n",
      "Start evaluation of Greedy\n",
      "0\n",
      "10\n",
      "20\n",
      "30\n",
      "40\n",
      "50\n",
      "60\n",
      "70\n",
      "80\n",
      "90\n",
      "Regret of Greedy : 8.880000000003122 +- 55.76091821338613 \n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- 8.392493724822998 seconds ---\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "c = 0.1\n",
    "\n",
    "plt.figure()\n",
    "for algo in algorithms:\n",
    "    print('Start evaluation of', algo.__class__.__name__)\n",
    "    ev = EvaluationMAB(env, algo, nb_rep, horizon)\n",
    "    mean_regret = ev.mean_regret()\n",
    "    std_regret = ev.std_regret()\n",
    "    print('Regret of', algo.__class__.__name__, ':', mean_regret[-1], '+-', std_regret[-1], '\\n')\n",
    "    \n",
    "    plt.plot(np.arange(horizon), mean_regret)\n",
    "    plt.fill_between(np.arange(horizon), mean_regret-c*std_regret, mean_regret+c*std_regret, alpha=.2)\n",
    "    plt.xlabel('Time')\n",
    "    plt.ylabel('Regret')\n",
    "    \n",
    "\n",
    "plt.legend([algo.__class__.__name__ for algo in algorithms], loc=0)\n",
    "plt.show()\n",
    "\n",
    "print(\"--- %s seconds ---\" % (time.time() - start_time))"
   ]
  },
  {
   "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}