add rtdp debug notebook

mdp/measure-rtdp.ipynb

@@ -0,0 +1,147 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "22ce969b-6f42-4379-b338-d7555cf06d28",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas\n",
+    "import pickle\n",
+    "import seaborn"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "32dc8e26-cc01-4caa-9438-b345fd9b6ded",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with open(\"measure-rtdp.pkl\", \"rb\") as f:\n",
+    "    results = pickle.load(f)\n",
+    "d = results[\"data\"]\n",
+    "list(d.columns)\n",
+    "d[\"fullmodel\"] = d.apply(lambda x: f\"{x.protocol}-{x.model}-trunc{x.trunc}\", axis=1)\n",
+    "d.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ebad0902-19e2-4787-be2d-fc694d45890f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "log = []\n",
+    "for _, row in d.query('algo == \"rtdp\"').iterrows():\n",
+    "    for entry in row.log:\n",
+    "        params = dict()\n",
+    "        for p in [\"fullmodel\", \"attacker\", \"hyperparams\"]:\n",
+    "            params[p] = row[p]\n",
+    "        log.append(dict(algo=\"rtdp\") | params | entry)\n",
+    "\n",
+    "        ref_row = d.query(\n",
+    "            f'fullmodel == \"{row.fullmodel}\" and attacker  == \"{row.attacker}\" and algo == \"aft20\"'\n",
+    "        )\n",
+    "        assert ref_row.shape[0] == 1\n",
+    "        ref_row = ref_row.iloc[0]\n",
+    "        ref_entry = dict(\n",
+    "            step=entry[\"step\"],\n",
+    "            start_value=ref_row.start_value,\n",
+    "            start_progress=ref_row.start_progress,\n",
+    "            n_states=ref_row.pimc_n_states,\n",
+    "        )\n",
+    "        log.append(dict(algo=\"aft20\") | params | ref_entry)\n",
+    "\n",
+    "log = pandas.DataFrame(log)\n",
+    "log"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ba762257-9446-404d-b354-cd99ca428ad8",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "seaborn.relplot(\n",
+    "    data=log,\n",
+    "    kind=\"line\",\n",
+    "    x=\"step\",\n",
+    "    y=\"start_value\",\n",
+    "    col=\"attacker\",\n",
+    "    hue=\"algo\",\n",
+    "    row=\"fullmodel\",\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d62b4a1d-9a1c-491f-9af1-aa1684bb2620",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "seaborn.relplot(\n",
+    "    data=log.assign(rpp=lambda x: x.start_value / x.start_progress),\n",
+    "    kind=\"line\",\n",
+    "    x=\"step\",\n",
+    "    y=\"rpp\",\n",
+    "    col=\"attacker\",\n",
+    "    hue=\"algo\",\n",
+    "    row=\"fullmodel\",\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0d8e098c-a54d-42d9-9c85-0d22f8206460",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "seaborn.relplot(\n",
+    "    # data = log.query('algo == \"rtdp\"'),\n",
+    "    data=log,\n",
+    "    kind=\"line\",\n",
+    "    x=\"step\",\n",
+    "    y=\"n_states\",\n",
+    "    col=\"attacker\",\n",
+    "    hue=\"algo\",\n",
+    "    row=\"fullmodel\",\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "05232b9e-754b-4b1a-a02c-27ec48a5a760",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "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": "3.9.19"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}