notebooks/therapy_evaluation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Experiments with real data (comparison with baselines)\n",
    "\n",
    "This notebook reproduces Figure 6 in the paper. It requires output files generated by executing the scripts `scripts/therapy_compute_cf_mdps.sh` (with the `unobserved_reward=normal` option) and `scripts/therapy_evaluation.sh`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.chdir(\"../\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import glob\n",
    "import json\n",
    "import copy\n",
    "import matplotlib.ticker as mtick\n",
    "from src import utils\n",
    "from src.utils import get_fig_dim\n",
    "from IPython.display import clear_output\n",
    "import matplotlib.pyplot as plt\n",
    "from src.therapy_mdp import Therapy_MDP\n",
    "from scipy.stats import entropy\n",
    "\n",
    "sns.set_theme(context='paper', style='ticks', font_scale=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "width_pt = 397\n",
    "palette = sns.color_palette('husl', 7)\n",
    "type_dict = {\n",
    "    'id' : int,\n",
    "    'k' : int,\n",
    "    'seed' : int,\n",
    "    'counterfactual outcome' : float\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading file 5208/5208\n"
     ]
    }
   ],
   "source": [
    "files = glob.glob(\"./outputs/therapy_eval*.json\")\n",
    "\n",
    "dicts = []\n",
    "for fl_id, fl in enumerate(files):\n",
    "    clear_output(wait=True)\n",
    "    print('Reading file ' + str(fl_id+1)+'/'+str(len(files)))\n",
    "    with open(fl,\"r\") as f:\n",
    "        js = json.load(f)\n",
    "        for seed in js['seeds']:\n",
    "            # new_js = copy.deepcopy(js)\n",
    "            new_js = js.copy()\n",
    "            del new_js['seeds']\n",
    "            new_js['seed'] = seed\n",
    "            new_js['counterfactual outcome'] = js['seeds'][seed]\n",
    "            dicts.append(new_js)\n",
    "\n",
    "input_df = pd.DataFrame(dicts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df = input_df.astype(type_dict) # Fix types\n",
    "data_df = data_df.replace({'random' : 'Random', 'greedy' : 'Greedy', 'randomized' : 'Noisy greedy', 'optimal' : 'Algorithm 2'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "findfont: Font family ['serif'] not found. Falling back to DejaVu Sans.\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 257.086x158.888 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "utils.latexify(font_size=10, legend_font_size=10)\n",
    "\n",
    "fig_height, fig_aspect = get_fig_dim(width_pt, fraction=0.65)\n",
    "fig, ax = plt.subplots(figsize=(fig_height*fig_aspect,fig_height))\n",
    "\n",
    "methods = ['Random', 'Greedy', 'Noisy greedy', 'Algorithm 2']\n",
    "temp_df = data_df[data_df['k']<=10]\n",
    "sns.lineplot(data=temp_df, x='k', y='counterfactual outcome', hue='policy', hue_order=methods, palette=palette[:4], err_style='band', ax=ax)\n",
    "sns.despine()\n",
    "\n",
    "ax.legend(title=None, loc='lower right')\n",
    "ax.set_ylim([50,70])\n",
    "ax.set_xlim(left=0)\n",
    "ax.set_xlabel(r'$k$')\n",
    "ax.set_ylabel('Average counterfactual\\n outcome')\n",
    "\n",
    "fig.tight_layout()\n",
    "fig.savefig('figures/therapy_eval_outcome_vs_k.pdf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "6dea91a18eea4d3d11c1da10287474f71a3ef4eae33b2084b9cce9fec272a02e"
  },
  "kernelspec": {
   "display_name": "Python 3.9.1 64-bit ('env': venv)",
   "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": "3.9.1"
  },
  "orig_nbformat": 2
 },
 "nbformat": 4,
 "nbformat_minor": 2
}