analysis/advantage/3_visualize_statistics.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "df= pd.read_csv(\n",
    "    \"/path/to/ckpt_100000_pt_0_001_vs_Stockfish_1_adv.csv\"\n",
    ")\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i, row in df.iterrows():\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "sampled_moves = json.loads(row['all_sampled_moves'])\n",
    "sampled_moves\n",
    "for ply in sampled_moves:\n",
    "    if ply is None:\n",
    "        continue\n",
    "    for k, v in ply.items():\n",
    "        tuple_everything = [] #original does not have tuples for none\n",
    "        for i in v:\n",
    "            if i is None:\n",
    "                tuple_everything.append((None, None))\n",
    "            else:\n",
    "                tuple_everything.append(i)\n",
    "        \n",
    "        ply[k] = tuple_everything\n",
    "\n",
    "sampled_moves"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "stats = torch.load('./stats_3.pt')\n",
    "\n",
    "\n",
    "# %pip install scipy\n",
    "from scipy.stats import bootstrap\n",
    "\n",
    "test = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])\n",
    "k = bootstrap(test[None], np.mean)\n",
    "# print(k.standard_error * 1.96, np.mean(stats[i]) - k.confidence_interval.low)\n",
    "# call bootstrap function\n",
    "\n",
    "def calc_stat(stat):\n",
    "    mean =  np.mean(np.array(stat))\n",
    "    bootstrap_ci = bootstrap(stat[None], np.mean)\n",
    "    # print(round(mean, 4), bootstrap_ci.confidence_interval)\n",
    "    return round(mean, 4), mean - bootstrap_ci.confidence_interval.low\n",
    "\n",
    "for i in stats:\n",
    "    if not isinstance(stats[i][0], dict):   \n",
    "        stats[i] = np.array(stats[i]) * 100\n",
    "        # print(i, round(np.mean(np.array(stats[i])), 4))\n",
    "        print(i, calc_stat(np.array(stats[i])))\n",
    "    else:\n",
    "        print()\n",
    "        print(i)\n",
    "        for k in stats[i][0]:\n",
    "            print(k, calc_stat(np.array([j[k] for j in stats[i]]) * 100))\n",
    "            # print(k, round(np.mean(np.array([j[k] for j in stats[i]])) * 100, 4))\n",
    "        print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "matplotlib.rcParams.update({\"font.size\": 13})\n",
    "\n",
    "# use tex\n",
    "\n",
    "plt.rcParams.update(\n",
    "    {\n",
    "        \"text.usetex\": True,\n",
    "        \"text.latex.preamble\": r\"\\usepackage{amsfonts}\",\n",
    "    }\n",
    ")\n",
    "\n",
    "\n",
    "# plt.yscale(\"log\", base=2)\n",
    "# plt.xscale(\"symlog\", base=1.0)\n",
    "plt.figure(figsize=(8, 2.15), dpi=300)\n",
    "\n",
    "ax = plt.gca()\n",
    "# ax.set_xlim([-.1, 0.1])\n",
    "ax.set_xlim([-10, 10])\n",
    "\n",
    "COLORS = [\"green\", \"red\"]\n",
    "\n",
    "for i, k in enumerate(['acc_deltas_0.001', 'acc_deltas_0.75']):\n",
    "    # for i, k in enumerate([\"advantage_deltas_0.001\", \"advantage_deltas_0.5\"]):\n",
    "    counts, bins = np.histogram(stats[k], bins=100, density=True, range=(-20, 20))\n",
    "    # counts, bins = np.histogram(stats[k], bins=50, density=True)\n",
    "    plt.hist(\n",
    "        bins[:-1],\n",
    "        bins=bins,\n",
    "        weights=counts,\n",
    "        density=True,\n",
    "        histtype=\"bar\",\n",
    "        label=fr\"$f'=f_{{\\tau={k.split('_')[-1]}}}$\",\n",
    "        color=COLORS[i],\n",
    "        alpha=0.2,\n",
    "    )\n",
    "    mean = np.mean(stats[k])\n",
    "    # plt.axvline(x=mean, color=COLORS[i],                \n",
    "    #             label=rf\"$\\mathbb{{E}}[F]:$ {mean:.2f}\",\n",
    "    #             )\n",
    "\n",
    "#\n",
    "plt.axvline(x=0, color=\"black\", label=rf\"Baseline $\\mathbb{{E}}[F]:$ 0.0\")\n",
    "# plt.text(\n",
    "#     # mean - 0.4175,\n",
    "#     0.07,\n",
    "#     plt.ylim()[1] * (0.75 - 0.2 * 2),\n",
    "#     rf\"Baseline $\\mathbb{{E}}[F]:$ 0.0\",\n",
    "#     color=\"black\",\n",
    "#     verticalalignment=\"top\",\n",
    "#     weight=\"bold\",\n",
    "# )\n",
    "\n",
    "for i, k in enumerate(\n",
    "    [\n",
    "        \"acc_deltas_0.001\",\n",
    "        \"acc_deltas_0.75\",\n",
    "    ]\n",
    "):\n",
    "    mean = np.mean(stats[k])\n",
    "    plt.axvline(\n",
    "        x=mean,\n",
    "        color=COLORS[i],\n",
    "        label=rf\"$\\mathbb{{E}}[F]:$ {mean:.2f}\",\n",
    "    )\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "plt.xlabel(r\"Change in expected reward over base $f_{\\tau=1}$ probability distribution over individual states: $F(f', f)$\")\n",
    "plt.ylabel(r\"$\\mathbb{P}(F)$\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "counts, bins = np.histogram(stats[\"advantage_deltas_0.5\"], bins=50)\n",
    "plt.hist(bins[:-1], bins, weights=counts, density=True)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.11.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}