calibration_isotonic.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "f1e54ede-f04a-4726-b23c-cb2efcf450a0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "\n",
    "import argparse\n",
    "import csv\n",
    "import os\n",
    "import sys\n",
    "import tifffile\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "import glob\n",
    "import pickle\n",
    "\n",
    "import datetime\n",
    "import math\n",
    "import json\n",
    "import time\n",
    "\n",
    "import timm\n",
    "from timm.models import create_model\n",
    "from timm.scheduler import CosineLRScheduler\n",
    "from timm.optim import create_optimizer\n",
    "from timm.utils import NativeScaler, get_state_dict\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "from torch.utils.data import Dataset, random_split, DataLoader\n",
    "from PIL import Image\n",
    "import torchvision.models as models\n",
    "import torchvision.transforms as T\n",
    "import torch.backends.cudnn as cudnn\n",
    "import torch.nn.functional as F\n",
    "import torch.nn as nn\n",
    "from torchvision.utils import make_grid\n",
    "\n",
    "import torch.multiprocessing as mp\n",
    "import torch.distributed as dist\n",
    "from torch.nn.parallel import DistributedDataParallel as DDP\n",
    "from torch.utils.data.distributed import DistributedSampler\n",
    "\n",
    "from customloss import CustomLoss\n",
    "from datamgr import *\n",
    "import utils\n",
    "from engine import *\n",
    "from dataset import *\n",
    "from samplers import RASampler\n",
    "\n",
    "import copy\n",
    "from pathlib import Path\n",
    "from tqdm import tqdm\n",
    "\n",
    "def save_pickle(file, data):\n",
    "    with open(file, 'wb') as f:\n",
    "        pickle.dump(data, f)\n",
    "\n",
    "        \n",
    "def load_pickle(file):\n",
    "    with open(file, 'rb') as f:\n",
    "        return pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "2be689ed-33b7-421e-b170-34a07c7e4ef9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "CNN_output_dir = Path('./checkpoints/CBNA/CNN/')\n",
    "MLP_output_dir = Path('./checkpoints/CBNA/MLP/')\n",
    "\n",
    "train_targets = torch.load(CNN_output_dir.joinpath('CNN_train_predictions.pth'))['targets']\n",
    "CNN_train_preds = torch.load(CNN_output_dir.joinpath('CNN_train_predictions.pth'))['predictions']\n",
    "MLP_train_preds = torch.load(MLP_output_dir.joinpath('MLP_train_predictions.pth'))['predictions']\n",
    "\n",
    "test_targets = torch.load(CNN_output_dir.joinpath('CNN_test_predictions.pth'))['targets']\n",
    "CNN_test_preds = torch.load(CNN_output_dir.joinpath('CNN_test_predictions.pth'))['predictions']\n",
    "MLP_test_preds = torch.load(MLP_output_dir.joinpath('MLP_test_predictions.pth'))['predictions']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "137c77ae-7dff-4898-bdb6-882875ae95a9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.isotonic import IsotonicRegression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "f9bc01a0-adf6-4dc5-955b-18a2121f08b0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "CNN_iso = IsotonicRegression(out_of_bounds = 'clip')\n",
    "CNN_calib_test_preds = []\n",
    "\n",
    "for idx in range(train_targets.shape[1]):\n",
    "    CNN_iso.fit(CNN_train_preds[:,idx], train_targets[:,idx])\n",
    "    CNN_calib_test_probs = CNN_iso.predict(CNN_test_preds[:,idx])\n",
    "    CNN_calib_test_preds.append(torch.tensor(CNN_calib_test_probs).unsqueeze(0))\n",
    "\n",
    "CNN_calib_test_preds = torch.cat(CNN_calib_test_preds).t()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "7a7fb7ed-7c12-4915-abed-d44babe5b1f6",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def compute_scores(outputs, targets, threshold=0.45):\n",
    "    \n",
    "    prob = outputs >= threshold\n",
    "    label = targets >= threshold\n",
    "    TP = (prob & label).sum(0).float()\n",
    "    TN = ((~prob) & (~label)).sum(0).float()\n",
    "    FP = (prob & (~label)).sum(0).float()\n",
    "    FN = ((~prob) & label).sum(0).float()\n",
    "    \n",
    "    # macro true skill score\n",
    "    macro_recall = TP / (TP + FN + 1e-8)\n",
    "    macro_specificity = TN / (TN + FP + 1e-8)\n",
    "    macro_tss = torch.mean(macro_recall + macro_specificity - 1)\n",
    "    macro_precision = TP / (TP + FP + 1e-8)\n",
    "    macro_accuracy = (TP + TN) / (TP + FP + TN + FN + 1e-8)\n",
    "    print('macro recall: ', torch.mean(macro_recall).item())\n",
    "    print('macro specificity: ', torch.mean(macro_specificity).item())\n",
    "    print('macro tss: ', torch.mean(macro_tss).item())\n",
    "    print('macro precision: ', torch.mean(macro_precision).item())\n",
    "    print('macro_accuracy: ', torch.mean(macro_accuracy).item())\n",
    "    # micro true skill score\n",
    "    micro_recall = TP.sum() / (TP.sum() + FN.sum())\n",
    "    micro_specificity = TN.sum() / (TN.sum() + FP.sum())\n",
    "    micro_tss = micro_recall + micro_specificity - 1\n",
    "    micro_precision = TP.sum() / (TP.sum() + FP.sum() + 1e-8)\n",
    "    micro_accuracy = (TP.sum() + TN.sum()) / (TP.sum() + FP.sum() + TN.sum() + FN.sum() + 1e-8)\n",
    "    print('micro recall: ', micro_recall.item())\n",
    "    print('micro specificity: ', torch.mean(micro_specificity).item())\n",
    "    print('micro tss: ', torch.mean(micro_tss).item())\n",
    "    print('micro_precision: ', micro_precision.item())\n",
    "    print('micro_accuracy: ', micro_accuracy.item())\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "a035142e-d2df-44a9-aa24-3d80c96fcc53",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "macro recall:  0.6793069839477539\n",
      "macro specificity:  0.8866408467292786\n",
      "macro tss:  0.5659478306770325\n",
      "macro precision:  0.040565989911556244\n",
      "macro_accuracy:  0.8882507681846619\n",
      "micro recall:  0.8912724256515503\n",
      "micro specificity:  0.888227105140686\n",
      "micro tss:  0.7794995307922363\n",
      "micro_precision:  0.058801889419555664\n",
      "micro_accuracy:  0.8882507681846619\n"
     ]
    }
   ],
   "source": [
    "compute_scores(CNN_calib_test_preds, test_targets, threshold=0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "aa52ddaa-a4fe-41b4-bab3-700251571d1c",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fa95625a650>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(test_targets.sum(1), CNN_calib_test_preds.sum(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "fa121a32-3e3a-486a-a4d0-9e96898f18c9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "MLP_iso = IsotonicRegression(out_of_bounds = 'clip')\n",
    "MLP_calib_test_preds = []\n",
    "\n",
    "for idx in range(train_targets.shape[1]):\n",
    "    MLP_iso.fit(MLP_train_preds[:,idx], train_targets[:,idx])\n",
    "    MLP_calib_test_probs = MLP_iso.predict(MLP_test_preds[:,idx])\n",
    "    MLP_calib_test_preds.append(torch.tensor(MLP_calib_test_probs).unsqueeze(0))\n",
    "\n",
    "MLP_calib_test_preds = torch.cat(MLP_calib_test_preds).t()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "2dc98bdd-2966-4afe-ab2a-6c07348110ca",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "macro recall:  0.6846301555633545\n",
      "macro specificity:  0.8679312467575073\n",
      "macro tss:  0.5525612831115723\n",
      "macro precision:  0.037342529743909836\n",
      "macro_accuracy:  0.8700000643730164\n",
      "micro recall:  0.9011030793190002\n",
      "micro specificity:  0.8697562217712402\n",
      "micro tss:  0.7708592414855957\n",
      "micro_precision:  0.05141952261328697\n",
      "micro_accuracy:  0.8699999451637268\n"
     ]
    }
   ],
   "source": [
    "compute_scores(MLP_calib_test_preds, test_targets, threshold=0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "9eee5302-8241-4f8f-a2ae-d64ee2fe79df",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fa9564e4df0>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(test_targets.sum(1), MLP_calib_test_preds.sum(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2d03e031-7d21-4763-b71d-413f154b752a",
   "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": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}