{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d8d3359a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from utils import *\n",
    "from spike_train_utils import *\n",
    "import numpy as np\n",
    "import scipy.stats as stats\n",
    "from brian2 import *#BrianLogger.suppress_hierarchy('brian2.codegen')\n",
    "import os\n",
    "\n",
    "import joblib\n",
    "import pandas as pd\n",
    "import network_CADEX as cadex"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "32069a34",
   "metadata": {},
   "outputs": [],
   "source": [
    "# for the time being, we select the best network parameters for each network type\n",
    "error_columns = [\"burst_error\"]\n",
    "\n",
    "\n",
    "df_full_networks = pd.read_csv('./output/concat_res_w_mse.csv')\n",
    "network_types = df_full_networks['Network_type'].unique()[::-1]\n",
    "best_networks = {x: {} for x in network_types}\n",
    "for network_type in network_types:\n",
    "    best_network = df_full_networks[df_full_networks['Network_type'] == network_type]\n",
    "    \n",
    "    # calculate the mean error across all error columns\n",
    "    best_network['mean_error'] = best_network[error_columns].mean(axis=1)\n",
    "    best_network = best_network.nsmallest(10, 'mean_error')\n",
    "    #print(f\"Best network for {network_type}: {best_network['path']}\")\n",
    "    #actually get the best 5 per type\n",
    "    \n",
    "    best_networks[network_type] = {}\n",
    "    for i, row in best_network.iterrows():\n",
    "        best_networks[network_type][f\"best_{i}\"] = row.to_dict()\n",
    "        best_networks[network_type][f\"best_{i}\"].update(**best_network.to_dict())\n",
    "        best_networks[network_type][f\"best_{i}\"]['path'] = \"/media/smestern/sgbackup/aoi_paper_2/\" + os.path.basename(os.path.dirname(row['path'])) + \"/\" + os.path.basename(row['path'])\n",
    "        best_networks[network_type][f\"best_{i}\"]['mean_error'] = best_network['mean_error']\n",
    "        best_networks[network_type][f\"best_{i}\"]['args'] = cadex.load_params_from_joblib(best_networks[network_type][f\"best_{i}\"]['path'].replace('.csv', '.joblib'))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "232c346c",
   "metadata": {},
   "outputs": [],
   "source": [
    "#we need to concat the results into a single dataframe\n",
    "_main_dict = {'network_type': [], 'fit_id': [], 'path': [], 'mean_error': []}\n",
    "for network_type in best_networks:\n",
    "    for fit_id in best_networks[network_type]:\n",
    "        _main_dict['network_type'].append(network_type)\n",
    "        _main_dict['fit_id'].append(fit_id)\n",
    "        _main_dict['path'].append(best_networks[network_type][fit_id]['path'])\n",
    "        _main_dict['mean_error'].append(best_networks[network_type][fit_id]['mean_error'])\n",
    "best_networks_df = pd.DataFrame(_main_dict)\n",
    "best_networks_df.to_csv('./output/best_networks_paths.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "modelling",
   "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}