analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import time\n",
    "import json\n",
    "from collections import defaultdict\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import pickle\n",
    "from copy import deepcopy\n",
    "\n",
    "from data_preprocessing import get_all_data\n",
    "from utils import f1, accuracy, split_data\n",
    "\n",
    "RUNS_PATH = \"data_clean/runs\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_results(file_subpath):\n",
    "    with open(os.path.join(RUNS_PATH, file_subpath), \"rb\") as f:\n",
    "        return pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "### load results\n",
    "results = load_results(\"2024-10-28_22-11-30/All_columns__Logistic_Regression__results.pkl\")\n",
    "logreg = results[\"Logistic Regression\"]\n",
    "\n",
    "results = load_results(\"2024-10-28_22-11-30/Selected_columns_+_All_columns_PCA__Logistic_Regression__results.pkl\")\n",
    "svm = results[\"SVM\"]\n",
    "\n",
    "results = load_results(\"2024-10-31_14-44-57/All_columns__Ridge_Regression__results.pkl\")\n",
    "ridge = results[\"Ridge Regression\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "### load results\n",
    "all_results = defaultdict(dict)\n",
    "run_names = [\n",
    "    \"2024-10-31_03-18-13\",\n",
    "    \"2024-10-31_09-55-10\",\n",
    "    \"2024-10-31_12-16-37\",\n",
    "    \"2024-10-31_17-16-37\",\n",
    "]\n",
    "\n",
    "for run_name in run_names:\n",
    "    files = [f for f in os.listdir(os.path.join(RUNS_PATH, run_name)) if f.endswith(\"results.pkl\")]\n",
    "\n",
    "    ### collect test scores\n",
    "    for f in files:\n",
    "        data_name = f.split(\"__\")[0]\n",
    "        results = load_results(run_name + \"/\" + f)\n",
    "        for model_name, res in results.items():\n",
    "            if model_name not in all_results[data_name]:\n",
    "                all_results[data_name][model_name] = []\n",
    "            all_results[data_name][model_name].append(res[\"test_score\"])\n",
    "\n",
    "### print average and std\n",
    "for data_name, models in all_results.items():\n",
    "    print(data_name)\n",
    "    for model_name, scores in models.items():\n",
    "        print(\"  \", model_name, np.round(np.mean(scores),3), np.round(np.std(scores),3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plotting scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "### grid search results\n",
    "def get_2d_hps_grid(hps, val_scores, get_hp_to_plot_1, get_hp_to_plot_2):\n",
    "    uniq_1 = np.unique([get_hp_to_plot_1(hp) for hp in hps])\n",
    "    uniq_2 = np.unique([get_hp_to_plot_2(hp) for hp in hps])\n",
    "    grid_sum = np.zeros((len(uniq_1), len(uniq_2)))\n",
    "    grid_seen_n = np.zeros((len(uniq_1), len(uniq_2)))\n",
    "    for i, (hp, score) in enumerate(zip(hps, val_scores)):\n",
    "        grid_sum[\n",
    "            np.where(uniq_1 == get_hp_to_plot_1(hp))[0][0],\n",
    "            np.where(uniq_2 == get_hp_to_plot_2(hp))[0][0]\n",
    "        ] += score\n",
    "        grid_seen_n[\n",
    "            np.where(uniq_1 == get_hp_to_plot_1(hp))[0][0],\n",
    "            np.where(uniq_2 == get_hp_to_plot_2(hp))[0][0]\n",
    "        ] += 1\n",
    "    assert np.all(grid_seen_n > 0)\n",
    "    grid = grid_sum / grid_seen_n\n",
    "    return grid, uniq_1, uniq_2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "### prepare logreg results\n",
    "hps = logreg[\"hyperparam_search_list\"]\n",
    "val_scores = logreg[\"all_val_scores\"]\n",
    "logreg_get_hp_to_plot_1 = lambda hps_dict: hps_dict[\"class_weights\"][1]\n",
    "logreg_get_hp_to_plot_2 = lambda hps_dict: hps_dict[\"reg_mul\"]\n",
    "logreg[\"grid\"], logreg[\"uniq_1\"], logreg[\"uniq_2\"] = get_2d_hps_grid(hps, val_scores, logreg_get_hp_to_plot_1, logreg_get_hp_to_plot_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "### prepare svm results\n",
    "hps = svm[\"hyperparam_search_list\"]\n",
    "val_scores = svm[\"all_val_scores\"]\n",
    "svm_get_hp_to_plot_1 = lambda hps_dict: hps_dict[\"class_weights\"][1]\n",
    "svm_get_hp_to_plot_2 = lambda hps_dict: hps_dict[\"_lambda\"]\n",
    "svm[\"grid\"], svm[\"uniq_1\"], svm[\"uniq_2\"] = get_2d_hps_grid(hps, val_scores, svm_get_hp_to_plot_1, svm_get_hp_to_plot_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "### prepare ridge reg results\n",
    "hps = ridge[\"hyperparam_search_list\"]\n",
    "val_scores = ridge[\"all_val_scores\"]\n",
    "svm_get_hp_to_plot_1 = lambda hps_dict: hps_dict[\"class_weights\"][1]\n",
    "svm_get_hp_to_plot_2 = lambda hps_dict: hps_dict[\"reg_mul\"]\n",
    "ridge[\"grid\"], ridge[\"uniq_1\"], ridge[\"uniq_2\"] = get_2d_hps_grid(hps, val_scores, svm_get_hp_to_plot_1, svm_get_hp_to_plot_2)\n",
    "\n",
    "### remove last two hyperparam values from ridge\n",
    "ridge[\"grid\"] = ridge[\"grid\"][:, :-2]\n",
    "ridge[\"uniq_2\"] = ridge[\"uniq_2\"][:-2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_13078/3910124194.py:97: UserWarning: This figure includes Axes that are not compatible with tight_layout, so results might be incorrect.\n",
      "  fig.tight_layout(rect=[0, 0, 0.885, 1]) # add space on the right for colorbar\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 850x600 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "### plot heatmap of the validation scores wrt combination of two hyperparameters\n",
    "fontsize = 13.5\n",
    "logreg_xtick_labels = [r\"$10^{{{}}}$\".format(int(np.log10(x))) if x != 0 else \"0\" for x in logreg[\"uniq_2\"]]\n",
    "svm_xtick_labels = [r\"$10^{{{}}}$\".format(int(np.log10(x))) if x != 0 else \"0\" for x in svm[\"uniq_2\"]]\n",
    "ridge_xtick_labels = [r\"$10^{{{}}}$\".format(int(np.log10(x))) if x != 0 else \"0\" for x in ridge[\"uniq_2\"]]\n",
    "\n",
    "fig, axs = plt.subplots(1, 3, figsize=(8.5, 6))\n",
    "\n",
    "### left\n",
    "sns.heatmap(\n",
    "    ridge[\"grid\"],\n",
    "    annot=True,\n",
    "    fmt=\".2f\",\n",
    "    cmap=\"Reds\",\n",
    "    ax=axs[0],\n",
    "    cbar=False,\n",
    "    annot_kws={\"fontsize\": 10},\n",
    "    square=True,\n",
    "    vmin=0,\n",
    "    vmax=0.45\n",
    ")\n",
    "axs[0].set_title(\"Ridge Regression\", fontsize=fontsize, pad=12)\n",
    "axs[0].set_xlabel(\"Regularization\\nmultiplier\", fontsize=fontsize-1, labelpad=10)\n",
    "axs[0].set_ylabel(\"Class weighting (minority/majority)\", fontsize=fontsize, labelpad=16)\n",
    "axs[0].set_xticks(np.arange(len(ridge[\"uniq_2\"])) + 0.5)\n",
    "axs[0].set_xticklabels(ridge_xtick_labels, fontsize=fontsize-1)\n",
    "axs[0].set_yticks(np.arange(len(logreg[\"uniq_1\"])) + 0.5)\n",
    "axs[0].set_yticklabels(logreg[\"uniq_1\"], fontsize=fontsize)\n",
    "for spine in axs[0].spines.values():\n",
    "    spine.set_visible(True)\n",
    "    spine.set_linewidth(0.5)\n",
    "\n",
    "### center\n",
    "sns.heatmap(\n",
    "    logreg[\"grid\"],\n",
    "    annot=True,\n",
    "    fmt=\".2f\",\n",
    "    cmap=\"Reds\",\n",
    "    ax=axs[1],\n",
    "    cbar=False,\n",
    "    annot_kws={\"fontsize\": 10},\n",
    "    square=True,\n",
    "    vmin=0,\n",
    "    vmax=0.45\n",
    ")\n",
    "axs[1].set_title(\"Logistic Regression\", fontsize=fontsize, pad=12)\n",
    "axs[1].set_xlabel(\"Regularization\\nmultiplier\", fontsize=fontsize-1, labelpad=10)\n",
    "axs[1].set_xticks(np.arange(len(logreg[\"uniq_2\"])) + 0.5)\n",
    "axs[1].set_xticklabels(logreg_xtick_labels, fontsize=fontsize-1)\n",
    "axs[1].set_yticks([])\n",
    "axs[1].set_yticklabels([])\n",
    "for spine in axs[1].spines.values():\n",
    "    spine.set_visible(True)\n",
    "    spine.set_linewidth(0.5)\n",
    "\n",
    "### right\n",
    "im = axs[2].imshow(svm[\"grid\"], cmap=\"Reds\", vmin=0, vmax=0.45)\n",
    "sns.heatmap(\n",
    "    svm[\"grid\"],\n",
    "    annot=True,\n",
    "    fmt=\".2f\",\n",
    "    cmap=\"Reds\",\n",
    "    ax=axs[2],\n",
    "    cbar=False,\n",
    "    cbar_kws={\"shrink\": 0.8, \"label\": \"Validation F1 score\", \"pad\": 0.04},\n",
    "    annot_kws={\"fontsize\": 10},\n",
    "    square=True,\n",
    "    vmin=0,\n",
    "    vmax=0.45\n",
    ")\n",
    "axs[2].set_title(\"SVM\", fontsize=fontsize, pad=12)\n",
    "axs[2].set_xlabel(\"Regularization\\nmultiplier\", fontsize=fontsize-1, labelpad=10)\n",
    "axs[2].set_xticks(np.arange(len(svm[\"uniq_2\"])) + 0.5)\n",
    "axs[2].set_xticklabels(svm_xtick_labels, fontsize=fontsize-1)\n",
    "axs[2].set_yticks([])\n",
    "axs[2].set_yticklabels([])\n",
    "for spine in axs[2].spines.values():\n",
    "    spine.set_visible(True)\n",
    "    spine.set_linewidth(0.5)\n",
    "\n",
    "### colorbar\n",
    "cbar = fig.colorbar(im, ax=axs, fraction=0.019, pad=0)  # increased pad to avoid overlap\n",
    "cbar.ax.tick_params(labelsize=fontsize)\n",
    "cbar.locator = ticker.MaxNLocator(nbins=3)\n",
    "# cbar.set_label(\"Validation F1 score\", fontsize=fontsize, labelpad=18, rotation=270)\n",
    "\n",
    "### adjust layout\n",
    "fig.tight_layout(rect=[0, 0, 0.885, 1]) # add space on the right for colorbar\n",
    "fig.subplots_adjust(wspace=0.04)\n",
    "plt.show()\n",
    "\n",
    "# save as pdf\n",
    "# fig.savefig(\"hyperparam_search.pdf\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "metadata": {},
   "outputs": [],
   "source": [
    "cfg = {\n",
    "    \"raw_data_path\": \"data_raw\",\n",
    "    \"clean_data_path\": \"data_clean\",\n",
    "    \"allow_load_clean_data\": False,\n",
    "    \"remap_labels_to_01\": True,\n",
    "    \"seed\": 0,\n",
    "    \"eval_frac\": 0.2,\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/keko24/workspace/epfl/ml/project1/ml-first-project/data_preprocessing.py:303: RuntimeWarning: Mean of empty slice\n",
      "  mean = np.nanmean(arr)\n",
      "/home/keko24/miniforge3/envs/ml_aicrowd/lib/python3.12/site-packages/numpy/lib/_nanfunctions_impl.py:2053: RuntimeWarning: Degrees of freedom <= 0 for slice.\n",
      "  var = nanvar(a, axis=axis, dtype=dtype, out=out, ddof=ddof,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train.shape: (262508, 939), x_test.shape: (65627, 939), y_train.shape: (262508,), y_test.shape: (65627,)\n"
     ]
    }
   ],
   "source": [
    "### get data\n",
    "# x, x_final, y, ids, ids_final, col_idx_map, cleaned_col_idx_map = get_all_data(\n",
    "#     cfg=cfg,\n",
    "#     process_cols=\"all\",\n",
    "#     pca_kwargs=None,\n",
    "#     verbose=False,\n",
    "# )\n",
    "# x_train, x_test, y_train, y_test = split_data(x=x, y=y, split_frac=0.2, seed=cfg[\"seed\"])\n",
    "(x_train, x_test), (y_train, y_test), (ids_train, ids_test), col_idx_map, cleaned_col_idx_map, (x_final, ids_final) = get_all_data(\n",
    "    cfg=cfg,\n",
    "    process_cols=\"all\",\n",
    "    pca_kwargs=None,\n",
    "    standardize_num=True,\n",
    "    onehot_cat=True,\n",
    "    skip_rule_transformations=False,\n",
    "    verbose=False,\n",
    ")\n",
    "sample_weights_train = np.array([logreg[\"hyperparams\"][\"class_weights\"][yi] for yi in y_train])\n",
    "sample_weights_test = np.array([logreg[\"hyperparams\"][\"class_weights\"][yi] for yi in y_test])\n",
    "print(f\"x_train.shape: {x_train.shape}, x_test.shape: {x_test.shape}, y_train.shape: {y_train.shape}, y_test.shape: {y_test.shape}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Setup param update callback\n",
    "def update_callback(w, x, y, epoch, model, **kwargs):\n",
    "    reg_mul_before = model.reg_mul\n",
    "    model.reg_mul = 0\n",
    "\n",
    "    # Time tracking\n",
    "    current_time = time.time()\n",
    "    if hasattr(update_callback, 'genesis_time'):\n",
    "        time_elapsed = current_time - update_callback.genesis_time\n",
    "    else:\n",
    "        time_elapsed = 0  # First call, so no time elapsed\n",
    "        update_callback.genesis_time = current_time\n",
    "\n",
    "    # Save the elapsed time to history\n",
    "    history[\"time_elapsed\"].append(time_elapsed)\n",
    "\n",
    "    ### Loss\n",
    "    history[\"train_loss\"].append(model.log_reg_loss(\n",
    "        x=x_train, y=y_train, w=w, sample_weights=sample_weights_train))\n",
    "    history[\"test_loss\"].append(model.log_reg_loss(\n",
    "        x=x_test, y=y_test, w=w, sample_weights=sample_weights_test))\n",
    "\n",
    "    ### F1\n",
    "    y_hat_train = model.predict(x=x_train, w=w)\n",
    "    history[\"train_f1\"].append(f1(y_train, y_hat_train))\n",
    "    y_hat_test = model.predict(x=x_test, w=w)\n",
    "    history[\"test_f1\"].append(f1(y_test, y_hat_test))\n",
    "\n",
    "    history[\"w\"].append(deepcopy(w))\n",
    "    history[\"epoch\"].append(epoch)\n",
    "\n",
    "    model.reg_mul = reg_mul_before"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'gamma': None,\n",
       " 'use_line_search': True,\n",
       " 'optim_algo': 'gd',\n",
       " 'optim_kwargs': {'epochs': 50},\n",
       " 'class_weights': {0: 1, 1: 4},\n",
       " 'reg_mul': 0,\n",
       " 'verbose': False}"
      ]
     },
     "execution_count": 236,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "### make a new hyperparams dict\n",
    "new_hps = deepcopy(logreg[\"hyperparams\"])\n",
    "new_hps[\"optim_algo\"] = \"gd\"\n",
    "new_hps[\"optim_kwargs\"] = {'epochs': 50}\n",
    "# new_hps[\"optim_algo\"] = \"sgd\"\n",
    "# new_hps[\"optim_kwargs\"] = {'epochs': 3, \"batch_size\": 256}\n",
    "# new_hps[\"use_line_search\"] = False\n",
    "# new_hps[\"gamma\"] = 1e-2\n",
    "new_hps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {},
   "outputs": [],
   "source": [
    "### train\n",
    "history = defaultdict(list) # for tracking the optimization process\n",
    "model = logreg[\"model_cls\"](update_callback=update_callback, **new_hps)\n",
    "model = model.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {},
   "outputs": [],
   "source": [
    "### save for later\n",
    "#history_lbfgs = deepcopy(history)\n",
    "history_gd = deepcopy(history)\n",
    "# history_sgd = deepcopy(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fontsize = 22\n",
    "linewidth = 2\n",
    "\n",
    "# Create a single figure and axis\n",
    "fig, ax1 = plt.subplots(figsize=(14, 7))\n",
    "\n",
    "# Plot Loss on the left y-axis\n",
    "line1, = ax1.plot(history_lbfgs[\"time_elapsed\"], history_lbfgs[\"train_loss\"], color=\"dodgerblue\", linestyle=\"--\", label=\"Train loss (L-BFGS)\", linewidth=linewidth)\n",
    "line2, = ax1.plot(history_lbfgs[\"time_elapsed\"], history_lbfgs[\"test_loss\"], color=\"dodgerblue\", label=\"Test loss (L-BFGS)\", linewidth=linewidth)\n",
    "line3, = ax1.plot(history_gd[\"time_elapsed\"], history_gd[\"train_loss\"], color=\"navy\", linestyle=\"--\", label=\"Train loss (GD)\", linewidth=linewidth)\n",
    "line4, = ax1.plot(history_gd[\"time_elapsed\"], history_gd[\"test_loss\"], color=\"navy\", label=\"Test loss (GD)\", linewidth=linewidth)\n",
    "\n",
    "# Label for the left y-axis (Loss)\n",
    "ax1.set_xlabel(\"Time elapsed in seconds\", fontsize=fontsize + 4, labelpad=20)\n",
    "ax1.set_ylabel(\"Loss\", fontsize=fontsize + 4, labelpad=20)\n",
    "ax1.tick_params(axis=\"x\", labelsize=fontsize + 4)\n",
    "ax1.tick_params(axis=\"y\", labelsize=fontsize + 4)\n",
    "ax1.spines[\"top\"].set_visible(False)\n",
    "ax1.spines[\"right\"].set_visible(False)\n",
    "ax1.set_ylim(0.35, 0.65)\n",
    "ax1.set_xlim(0, min(history_lbfgs[\"time_elapsed\"][-1], history_gd[\"time_elapsed\"][-1]) - 1)\n",
    "\n",
    "# Create a second y-axis for F1 metrics, sharing the same x-axis\n",
    "ax2 = ax1.twinx()\n",
    "line5, = ax2.plot(history_lbfgs[\"time_elapsed\"], history_lbfgs[\"train_f1\"], color=\"tab:red\", linestyle=\"--\", label=\"Train F1 (L-BFGS)\", linewidth=linewidth)\n",
    "line6, = ax2.plot(history_lbfgs[\"time_elapsed\"], history_lbfgs[\"test_f1\"], color=\"tab:red\", label=\"Test F1 (L-BFGS)\", linewidth=linewidth)\n",
    "line7, = ax2.plot(history_gd[\"time_elapsed\"], history_gd[\"train_f1\"], color=\"darkred\", linestyle=\"--\", label=\"Train F1 (GD)\", linewidth=linewidth)\n",
    "line8, = ax2.plot(history_gd[\"time_elapsed\"], history_gd[\"test_f1\"], color=\"darkred\", label=\"Test F1 (GD)\", linewidth=linewidth)\n",
    "\n",
    "# Label for the right y-axis (F1)\n",
    "ax2.set_ylabel(\"F1\", fontsize=fontsize + 4, labelpad=20)\n",
    "ax2.tick_params(axis=\"y\", labelsize=fontsize + 4)\n",
    "ax2.spines[\"top\"].set_visible(False)\n",
    "ax2.set_ylim(0, 0.45)\n",
    "\n",
    "fig.text(0.012, 0.63, \"━\", color=\"dodgerblue\", fontsize=fontsize*2, va=\"center\")\n",
    "fig.text(0.012, 0.65, \"━\", color=\"navy\", fontsize=fontsize*2, va=\"center\")\n",
    "fig.text(0.964, 0.60, \"━\", color=\"tab:red\", fontsize=fontsize*2, va=\"center\")\n",
    "fig.text(0.964, 0.62, \"━\", color=\"darkred\", fontsize=fontsize*2, va=\"center\")\n",
    "\n",
    "# Combine all lines for a single legend at the top\n",
    "lines = [line1, line2, line3, line4, line5, line6, line7, line8]\n",
    "labels = [line.get_label() for line in lines]\n",
    "fig.legend(lines, labels, loc=\"upper center\", fontsize=fontsize + 2, ncol=2, bbox_to_anchor=(0.5, 1.3))\n",
    "\n",
    "fig.tight_layout()  # Adjust layout for clarity\n",
    "fig.savefig(\"optim_algo_combined_loss_f1_comparison_top_legend.pdf\", bbox_inches=\"tight\", dpi=400)\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "epfl",
   "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": 4
}