Pure Organic vs Bandit - Number of Online Users.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "pycharm": {}
      },
      "source": "## RecoGym -  Pure Organic vs Pure Bandit\n#### Varying the number of online users\n\nThe purpose of this notebook is to show that a \"Pure Organic\" algorithm i.e. an algorithm that orders recommendations according to a next (organic) item prediction method is able to harness relatively plentiful organic data i.e. data the users provide themselves by organically interacting with the website.  Although the data is plentiful, it is only a proxy of the data of interest for making recommendations which is indeed bandit data.\n\nIn contrast a \"Pure Bandit\" algorithm receives feedback from the recommendations however it is only able to discover that poor recommendations are poor by trying them extensively and performing poorly.\n"
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "\n",
        "import gym, recogym\n",
        "import pandas as pd\n",
        "\n",
        "from recogym import env_1_args, Configuration\n",
        "from copy import deepcopy\n",
        "\n",
        "env_1_args[\u0027random_seed\u0027] \u003d 5\n",
        "\n",
        "# initialize the gym for the first time by calling .make() and .init_gym()\n",
        "env \u003d gym.make(\u0027reco-gym-v1\u0027)\n",
        "env_1_args[\u0027num_products\u0027] \u003d 80"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [],
      "source": [
        "env.init_gym(env_1_args)\n",
        "\n",
        "# Create environments that differ only in the amount of noise \"number of flips\".\n",
        "env_dict \u003d {}\n",
        "for f in (int(env.config.num_products * 2 / 8),):  # Number of flips (difference between organic and bandit behaviour).\n",
        "    env_dict[f] \u003d deepcopy(env)\n",
        "    env_dict[f].generate_beta(f)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [],
      "source": [
        "from agents import BanditCount, bandit_count_args\n",
        "from agents import OrganicCount, organic_count_args"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [],
      "source": [
        "d_agent \u003d {\n",
        "    \u0027Pure Organic\u0027: OrganicCount(Configuration({\n",
        "        **organic_count_args,\n",
        "        **env_1_args,\n",
        "    })),\n",
        "    \u0027Pure Bandit\u0027: BanditCount(Configuration({\n",
        "        **bandit_count_args,\n",
        "        **env_1_args,\n",
        "    })),\n",
        "}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "collapsed": true,
        "pycharm": {}
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (129.42773461341858s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (134.23873591423035s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (140.72436571121216s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (158.4089493751526s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (137.20222568511963s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (138.85784220695496s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (147.21830415725708s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (155.47999095916748s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (224.7022602558136s)\n",
            "Start: Agent Training #0\n",
            "Start: Agent Testing #0\n",
            "End: Agent Testing #0 (246.60609769821167s)\n"
          ]
        }
      ],
      "source": [
        "offline_organic \u003d 10000\n",
        "eval_size \u003d 20000\n",
        "l \u003d list()\n",
        "for samples in (1, 10, 100, 1000, 10000):\n",
        "    for name in d_agent.keys():\n",
        "        agent \u003d d_agent[name]\n",
        "        for f in env_dict.keys():\n",
        "            c05, c0025, c0975 \u003d recogym.test_agent(\n",
        "                deepcopy(env_dict[f]),\n",
        "                deepcopy(agent),\n",
        "                samples,\n",
        "                eval_size,\n",
        "                offline_organic\n",
        "            )\n",
        "            df \u003d {}\n",
        "            df[\u0027agent\u0027] \u003d name\n",
        "            df[\u0027training\u0027] \u003d samples\n",
        "            df[\u0027ctr_05\u0027] \u003d [c05]\n",
        "            df[\u0027ctr_0025\u0027] \u003d [c0025]\n",
        "            df[\u0027ctr_0975\u0027] \u003d [c0975]\n",
        "            df[\u0027flips\u0027] \u003d \u0027flips\u003d%03d\u0027 % (f)\n",
        "            df[\u0027f\u0027] \u003d f\n",
        "\n",
        "            l.append(pd.DataFrame().from_dict(df))\n",
        "res \u003d pd.concat(l)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [],
      "source": [
        "# Enables the %%R magic.\n",
        "%load_ext rpy2.ipython\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(\u0027ignore\u0027)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "pycharm": {}
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n"
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "%%R -i res -w 5 -h 5 --units in -r 200\n",
        "# Import df from global environment.\n",
        "# Make default figure size 5 by 5 inches with 200 dpi resolution.\n",
        "\n",
        "library(ggplot2)\n",
        "library(data.table)\n",
        "library(scales)\n",
        "res\u003c-data.table(res)\n",
        "res[agent\u003d\u003d\u0027Pure Bandit\u0027,training:\u003d1.1*training] # Dodge.\n",
        "\n",
        "ggplot(res) + geom_errorbar(\n",
        "    aes(x\u003dtraining, ymin\u003dctr_0025, ymax\u003dctr_0975, colour\u003dagent), width\u003d0) +\n",
        "    facet_grid(flips ~ .) +\n",
        "    theme_bw() +\n",
        "    ylab(\u0027Click through rate\u0027) +\n",
        "    xlab(\u0027Number of Online Users\u0027) +\n",
        "    scale_x_log10() +\n",
        "    scale_y_continuous(labels\u003dpercent) +\n",
        "    scale_colour_manual(values\u003dc(\u0027blue\u0027, \u0027orange\u0027))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "pycharm": {}
      },
      "source": "We see that this plot reproduces the following prediction from our paper (with Pure Bandit and Pure Organic only):\n\n“Pure Organic” behavior is determined by the amount of noise in $\\sigma$ and is unaffected by the number of bandit events (this is not strictly true in the simulation and a small increase is evident on pure organic, as the user also provides an organic signal). If $\\sigma$ is sufficiently small that the actions are correctly ordered it will perform well, in contrast if it is large it will perform poorly. The performance does not change as the number of bandit events increase as these are not used. In contrast the “Pure Bandit” algorithm is unable to make predictions without large numbers of bandit events. It performs poorly when this data is scarce and well when it is plentiful. \n\n\u003cimg src\u003d\"images/section3_n_bandit_events.png\" alt\u003d\"Drawing\" style\u003d\"width: 650px;\"/\u003e"
    }
  ],
  "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": "3.7.1"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}