src/ACP_Tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " $\\Large \\textbf{CONSTRUCTING PREDICTION SETS WITH ACP}$"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook provides an introductory tutorial on Approximate full Conformal Prediction (ACP). The full details can be found in the accompanying paper:\n",
    "\n",
    "J. Abad Martinez, U. Bhatt, A. Weller and G. Cherubin. Approximating Full Conformal Prediction at Scale via Influence Functions. Association for the Advancement of Artificial Intelligence Conference on Artificial Intelligence (AAAI), 2023."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{CONSTRUCTING VALID PREDICTION SETS}$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Consider a training set $Z=(X, Y) \\in(\\mathcal{X} \\times \\mathcal{Y})^{N}$ of $N$ examples. For a test example $x \\in \\mathcal{X}$ and a chosen significance level $\\varepsilon \\in[0,1]$ a Conformal Predictor (CP) returns a set $\\Gamma^{e} \\subseteq \\mathcal{Y}$ that contains $x$’s true label with probability at least $1-\\varepsilon$. This guarantee (validity) holds for any exchangeable distribution on $Z \\cup\\{(x, y)\\}$.\n",
    "\n",
    "Particularly, full CP is a statistically powerful CP algorithm that attains the validity guarantees while also yielding tight prediction sets (i.e., high efficiency). Unfortunately,  full CP is notoriously computationally expensive. \n",
    "\n",
    "ACP consistently approximates full CP by leveraging Influence Functions (IF). Given a sufficiently large training set, ACP inherits the statistical power of full CP. This enables running full CP at a much reduced computational cost, and allows scaling it to large, real-world datasets. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{IMPORTING THE PACKAGES}$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "fbf10862",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import random\n",
    "\n",
    "SEED = 42 #for reproducibility\n",
    "np.random.seed(SEED)\n",
    "torch.manual_seed(SEED)\n",
    "torch.cuda.manual_seed(SEED)\n",
    "random.seed(SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76b7c489",
   "metadata": {},
   "source": [
    "$\\textbf{DATA}$"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "051bbadb",
   "metadata": {},
   "source": [
    "We generate synthetic data for a 5-class classification problem using scikit-learn’s $\\texttt{make_classification}$. We sample points from five Gaussian-distributed clusters, one per class. Each object is defined by 10 features. This is a concrete toy example, but feel free to play around with different synthetic settings!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "535370e0",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import make_classification\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "TRAIN_SIZE = 1000 #number of training points\n",
    "TEST_SIZE = 10 #number of test points\n",
    "N_CLASSES = 5 #number of classes\n",
    "N_FEATURES = 10 #total number of features \n",
    "N_CLUSTERS = 1 #number of clusters per class\n",
    "N_INFORMATIVE = 3 #number of informative features\n",
    "N_REDUNDANT = 2 #number of redundant features\n",
    "SEED = 42 #seed\n",
    "\n",
    "X, Y = make_classification(n_samples = TRAIN_SIZE + TEST_SIZE, n_features = N_FEATURES, n_classes = N_CLASSES,\n",
    "                           n_clusters_per_class = N_CLUSTERS, n_informative = N_INFORMATIVE, random_state = SEED)\n",
    "\n",
    "Xtrain, Xtest, Ytrain, Ytest = train_test_split(X, Y, test_size = TEST_SIZE, random_state = SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "47a04aff",
   "metadata": {},
   "source": [
    "$\\textbf{FITTING A MODEL}$"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "93eaf9cd",
   "metadata": {},
   "source": [
    "It’s time to decide the underlying model on top of which we build ACP. Here we use a 2-layers neural network; however, you can plug in any model as long as it is differentiable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "431be50c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using cuda device\n"
     ]
    }
   ],
   "source": [
    "#Instantiate model\n",
    "from acp.models import NeuralNetwork\n",
    "\n",
    "NUM_NEURONS = [20,10] #layers and neurons for the NN classifier\n",
    "REG = 0.01\n",
    "model = NeuralNetwork(input_size = N_FEATURES, num_neurons = NUM_NEURONS, out_size = N_CLASSES, seed = SEED, l2_reg = REG)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{BUILDING ACP AND GENERATING PREDICTION SETS}$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ACP works as a wrapper for an ML model. We now instantiate ACP and generate prediction sets for a chosen significance level $\\varepsilon$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from acp.wrapper import ACP_D\n",
    "\n",
    "#Building ACP (deleted)\n",
    "ACP = ACP_D(Xtrain, Ytrain, model, seed = SEED, verbose = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r\n",
      "  0%|                                                                                           | 0/10 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing p-values for 10 samples\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:15<00:00,  1.55s/it]\n"
     ]
    }
   ],
   "source": [
    "#Generate prediction sets that are guaranteed to contain the true class label with at least a specified probability 1-epsilon\n",
    "eps = 0.1 # 1-0.1 = 0.9 confidence of including the true label\n",
    "sets = ACP.predict(Xtest, eps, out_file = \"results/test\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us now look at the generated prediction sets. They are guaranteed to contain the true label with the specified probability $1-\\varepsilon$. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction sets: [[0, 1], [3], [1], [4], [2], [0, 2], [3, 4], [2], [2], [1]]\n",
      "\n",
      "True labels: [2 3 1 4 2 2 4 2 2 1]\n"
     ]
    }
   ],
   "source": [
    "print(\"Prediction sets: \" + str(list(sets.values()))) #prediction sets\n",
    "print()\n",
    "print(\"True labels: \" + str(Ytest)) #true labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{VALIDITY AND EFFICIENCY OF ACP}$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ACP inherits the statistical power of full CP, constructing prediction sets with high statistical power (i.e., efficiency).\n",
    "\n",
    "We obtain the accuracy of ACP, that is, the rate at which the true label is included in the prediction set. For $\\varepsilon=0.1$, ACP achieves an accuracy of $90\\%$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ACP's accuracy: 90.0%\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from acp.utils import compute_accum_acc\n",
    "\n",
    "result = pd.read_json(\"results/test\", lines = True) #loading ACP p-values\n",
    "accum_acc = compute_accum_acc(eps, result, Ytest) #computing accuracy for chosen epsilon \n",
    "\n",
    "print(\"ACP's accuracy: {}%\".format(accum_acc[-1]/len(Ytest)*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Naturally, ACP with $\\varepsilon = 0.1$ is more accurate than just predicting the outcomes based on the raw model's outputs:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model's accuracy: 80.0%\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "print(\"Model's accuracy: {}%\".format(accuracy_score(model.predict(Xtest), Ytest)*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now consider a single test point chosen uniformly at random. We can plot how the prediction set size changes w.r.t $\\varepsilon$ for ACP. We include the AUC for the interval $\\varepsilon \\in [0,0.2]$ and use it as a proxy of the efficiency (the lower the AUC, the higher the efficiency). This interval captures settings that are generally of practical interest. \n",
    "\n",
    "Additionally, we report with $\\color{lightcoral}\\bigstar$ the highest $\\varepsilon$ for which the prediction set contains the true label.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from acp.utils import get_average_interval, find_min_epsilon\n",
    "from scipy.integrate import simps\n",
    "\n",
    "cand = random.randint(0, len(Ytest)-1) #Randomly picking an example from the test set\n",
    "epsilons = np.arange(0,1.01,0.01) #Discretized x axis (epsilons)\n",
    "\n",
    "curve = get_average_interval(result, cand, epsilons) #Computing prediction set size w.r.t epsilon\n",
    "min_eps = find_min_epsilon(result, Ytest, cand, epsilons) #minimum epsilon that contains true label\n",
    "AUC = round(simps(curve[:21], epsilons[:21]),3) #AUC for the interval [0, 0.2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt \n",
    "\n",
    "plt.style.use('seaborn-paper')\n",
    "plt.rcParams['font.family'] = 'Times New Roman'\n",
    "plt.figure(figsize=(10, 6))\n",
    "\n",
    "plt.plot(epsilons, curve, label =\"ACP (D) (AUC = {})\".format(AUC), linewidth = 3)\n",
    "plt.scatter(min_eps, curve[np.where(epsilons == min_eps)[0][0]], s=550, marker=(5, 1), color = \"lightcoral\", zorder = 3) # type: ignore\n",
    "\n",
    "plt.legend(fontsize = \"24\", shadow = True, facecolor='white', framealpha=1)\n",
    "plt.xscale(\"linear\")\n",
    "plt.xlim([-0.0001, np.round(min_eps,1) + 0.1])\n",
    "plt.ylim([-0.01, N_CLASSES])\n",
    "plt.xticks(np.arange(0, np.round(min_eps,1) + 0.2, 0.1), fontsize = \"28\")\n",
    "plt.yticks(np.arange(N_CLASSES+1).tolist(), fontsize = \"28\")\n",
    "plt.ylabel(\"Prediction set size\", fontsize = \"36\")\n",
    "plt.xlabel(r\"$\\epsilon$\", fontsize = \"30\")\n",
    "plt.axvline(x=eps, color = \"black\", linewidth = 2, linestyle = \"--\")\n",
    "\n",
    "plt.grid()\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can measure the efficiency on the whole test set by averaging across points. Again, we report the AUC in $\\varepsilon \\in [0,0.2]$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from acp.utils import epsilon_curves\n",
    "\n",
    "curve_all = epsilon_curves(result, epsilons) #average prediction set size w.r.t epsilon\n",
    "AUC_all = round(simps(curve_all[:21], epsilons[:21]),3) #AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(epsilons, curve_all, label =\"ACP (D) (AUC = {})\".format(AUC_all), linewidth = 3)\n",
    "plt.legend(fontsize = \"24\", shadow = True, facecolor='white', framealpha=1)\n",
    "plt.xscale(\"linear\")\n",
    "plt.xlim([-0.0001, 1.0])\n",
    "plt.ylim([-0.01, N_CLASSES])\n",
    "plt.xticks(fontsize = \"28\")\n",
    "plt.yticks(np.arange(N_CLASSES+1).tolist(), fontsize = \"28\")\n",
    "plt.ylabel(\"Prediction set size\", fontsize = \"36\")\n",
    "plt.xlabel(r\"$\\epsilon$\", fontsize = \"30\")\n",
    "\n",
    "plt.grid()\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{EFFICIENCY COMPARISON}$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now consider main-stream CP alternatives, and compare them empirically with ACP on the basis of their predictive power (efficiency). We study split CP (SCP), RAPS and CV+. We keep the same underlying classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1 - Validation Loss: 2.5644\n",
      "Epoch: 10 - Validation Loss: 2.1006\n",
      "Epoch: 20 - Validation Loss: 1.7861\n",
      "Epoch: 30 - Validation Loss: 1.5363\n",
      "Epoch: 40 - Validation Loss: 1.3615\n",
      "Epoch: 50 - Validation Loss: 1.2511\n",
      "Epoch: 60 - Validation Loss: 1.1730\n",
      "Epoch: 70 - Validation Loss: 1.1197\n",
      "Epoch: 80 - Validation Loss: 1.0747\n",
      "Epoch: 90 - Validation Loss: 1.0416\n",
      "Epoch: 100 - Validation Loss: 1.0042\n",
      "Epoch: 110 - Validation Loss: 0.9823\n",
      "Epoch: 120 - Validation Loss: 0.9580\n",
      "Epoch: 130 - Validation Loss: 0.9325\n",
      "Epoch: 140 - Validation Loss: 0.9197\n",
      "Epoch: 150 - Validation Loss: 0.9057\n",
      "Epoch: 160 - Validation Loss: 0.8940\n",
      "Epoch: 170 - Validation Loss: 0.8828\n",
      "Epoch: 180 - Validation Loss: 0.8771\n",
      "Epoch: 190 - Validation Loss: 0.8588\n"
     ]
    }
   ],
   "source": [
    "#SCP\n",
    "\n",
    "from acp.others import split_cp\n",
    "device = torch.device(\"cuda\")\n",
    "\n",
    "model = NeuralNetwork(input_size = N_FEATURES, num_neurons = NUM_NEURONS, out_size = N_CLASSES, \n",
    "                      seed = SEED, l2_reg = REG).to(device)\n",
    "_, _ = split_cp(model, Xtrain, Ytrain, Xtest, labels=np.unique(Ytrain).tolist(), out_file = \"results/test_split_cp\", \n",
    "                validation_split = 0.2)\n",
    "\n",
    "result_split_cp = pd.read_json(\"results/test_split_cp\", lines = True) #load p-values\n",
    "curve_all_split_cp = epsilon_curves(result_split_cp, epsilons) #average prediction set size w.r.t epsilon\n",
    "AUC_all_split_cp = round(simps(curve_all_split_cp[:21], epsilons[:21]),3) #AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1 - Validation Loss: 2.5644\n",
      "Epoch: 10 - Validation Loss: 2.1006\n",
      "Epoch: 20 - Validation Loss: 1.7861\n",
      "Epoch: 30 - Validation Loss: 1.5363\n",
      "Epoch: 40 - Validation Loss: 1.3615\n",
      "Epoch: 50 - Validation Loss: 1.2511\n",
      "Epoch: 60 - Validation Loss: 1.1730\n",
      "Epoch: 70 - Validation Loss: 1.1197\n",
      "Epoch: 80 - Validation Loss: 1.0747\n",
      "Epoch: 90 - Validation Loss: 1.0416\n",
      "Epoch: 100 - Validation Loss: 1.0042\n",
      "Epoch: 110 - Validation Loss: 0.9823\n",
      "Epoch: 120 - Validation Loss: 0.9580\n",
      "Epoch: 130 - Validation Loss: 0.9325\n",
      "Epoch: 140 - Validation Loss: 0.9197\n",
      "Epoch: 150 - Validation Loss: 0.9057\n",
      "Epoch: 160 - Validation Loss: 0.8940\n",
      "Epoch: 170 - Validation Loss: 0.8828\n",
      "Epoch: 180 - Validation Loss: 0.8771\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 110.06it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 190 - Validation Loss: 0.8588\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.45it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2971956729888916\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297276258468628\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 250.04it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971363067626953\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 399.78it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.59it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2969298362731934\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2974075078964233\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once)."
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 487.06it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 476.03it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Optimal T=1.297109842300415\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971150875091553\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 689.68it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.02it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.296987533569336\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972488403320312\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 665.02it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.92it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2974780797958374\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2969958782196045\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 397.79it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 507.54it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2971456050872803\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297317385673523\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 333.41it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297420859336853\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971829175949097\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 507.75it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 504.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2975765466690063\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297413945198059\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 528.92it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 406.68it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971147298812866\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2977653741836548\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 590.58it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2977522611618042\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 687.42it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 673.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970867156982422\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2975715398788452\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.77it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 669.43it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 471.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.297539234161377\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2968237400054932\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 668.95it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 674.54it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2974379062652588\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2977049350738525\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.22it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2973968982696533\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2977416515350342\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 622.35it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 659.84it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2972973585128784\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2976081371307373\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 400.20it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 630.77it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.297223448753357\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2968311309814453\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 399.31it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.01it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2972910404205322\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973709106445312\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 558.01it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.56it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 492.14it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297211766242981\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973929643630981\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2972253561019897\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 504.94it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 528.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973213195800781\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2976253032684326\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 475.17it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.71it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971044778823853\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970727682113647\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.98it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 674.43it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970242500305176\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.29703688621521\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.01it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 504.09it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2975132465362549\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297073483467102\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.71it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 456.57it/s]\n",
      "  0%|                                                                                            | 0/2 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2971477508544922\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2974647283554077\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 634.73it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 547.63it/s]\n",
      "  0%|                                                                                            | 0/2 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2976586818695068\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.29723060131073\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 673.24it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 504.18it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2975367307662964\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2976878881454468\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 402.06it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 676.12it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971805334091187\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2969812154769897\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 516.64it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.56it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973030805587769\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2974345684051514\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.74it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 476.19it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2976269721984863\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2974250316619873\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 495.96it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.62it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2972198724746704\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972548007965088\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.56it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 674.05it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 680.95it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2976083755493164\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971744537353516\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297027826309204\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.27it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 489.62it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972875833511353\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972385883331299\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 501.05it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 399.91it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2975257635116577\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973202466964722\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 472.38it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.10it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2974727153778076\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972954511642456\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 503.49it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.74it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2975354194641113\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971059083938599\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 478.07it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 400.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.297417402267456\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2977871894836426\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.68it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 675.63it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2970141172409058\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2974047660827637\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.29it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.98it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2971621751785278\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972480058670044\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971594333648682\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.62it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.77it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970466613769531\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973941564559937\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 499.86it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.29it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.61it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2972625494003296\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2977524995803833\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 538.53it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.2973322868347168\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2976794242858887\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 425.80it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 487.00it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 669.54it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal T=1.297385573387146\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297310471534729\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970316410064697\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 672.33it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 713.20it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973852157592773\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2975223064422607\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 497.60it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 585.47it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 490.88it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971949577331543\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973690032958984\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973628044128418\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 497.31it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 492.93it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2972244024276733\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971322536468506\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 666.98it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.01it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973501682281494\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2973374128341675\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 422.86it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 674.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2976255416870117\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.297223687171936\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 399.61it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 500.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2975932359695435\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2970114946365356\n",
      "Begin Platt scaling.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 665.92it/s]\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 400.01it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2971203327178955\n",
      "Begin Platt scaling.\n",
      "Computing logits for model (only happens once).\n",
      "Optimal T=1.2969257831573486\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "#RAPS\n",
    "\n",
    "from torch.utils.data import DataLoader\n",
    "from acp.utils import create_set, check_tensor\n",
    "from third_party.RAPS.RAPS_conformal import ConformalModel\n",
    "\n",
    "model = NeuralNetwork(input_size = N_FEATURES, num_neurons = NUM_NEURONS, out_size = N_CLASSES,\n",
    "                      seed = SEED, l2_reg = REG).to(device)\n",
    "Xtrain_prop, Xval, Ytrain_prop, Yval = train_test_split(Xtrain, Ytrain, test_size = 0.2, random_state = SEED)\n",
    "model.fit(Xtrain_prop, Ytrain_prop, seed = SEED) #training on proper training set\n",
    "calib_loader = DataLoader(create_set(Xval, Yval), batch_size=128, shuffle=True, pin_memory=True) #calib data # type: ignore\n",
    "sizes = {}\n",
    "for i, epsilon in enumerate(epsilons): #repeat for each epsilon\n",
    "    cmodel = ConformalModel(model, calib_loader, alpha=epsilon, lamda_criterion='size', allow_zero_sets=True) #building RAPS\n",
    "    \n",
    "    for j in range(len(Ytest)):\n",
    "        _, set = cmodel(check_tensor(np.expand_dims(Xtest[j], axis=0)).to(device)) # type: ignore #constructing prediction sets\n",
    "        \n",
    "        if epsilon not in sizes:\n",
    "            sizes[epsilon] = [len(list(set[0]))]\n",
    "        else:\n",
    "            sizes[epsilon].append(len(list(set[0])))\n",
    "    \n",
    "curve_all_RAPS = [np.mean(sizes[epsilon]) for epsilon in sizes] #average prediction set size w.r.t epsilon\n",
    "AUC_all_RAPS = round(simps(curve_all_RAPS[:21], epsilons[:21]),3) #AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1 - Validation Loss: 2.6547\n",
      "Epoch: 10 - Validation Loss: 2.1409\n",
      "Epoch: 20 - Validation Loss: 1.8083\n",
      "Epoch: 30 - Validation Loss: 1.5507\n",
      "Epoch: 40 - Validation Loss: 1.3418\n",
      "Epoch: 50 - Validation Loss: 1.2044\n",
      "Epoch: 60 - Validation Loss: 1.1104\n",
      "Epoch: 70 - Validation Loss: 1.0377\n",
      "Epoch: 80 - Validation Loss: 0.9775\n",
      "Epoch: 90 - Validation Loss: 0.9327\n",
      "Epoch: 100 - Validation Loss: 0.8972\n",
      "Epoch: 110 - Validation Loss: 0.8656\n",
      "Epoch: 120 - Validation Loss: 0.8377\n",
      "Epoch: 130 - Validation Loss: 0.8122\n",
      "Epoch: 140 - Validation Loss: 0.7959\n",
      "Epoch: 150 - Validation Loss: 0.7790\n",
      "Epoch: 160 - Validation Loss: 0.7645\n",
      "Epoch: 170 - Validation Loss: 0.7536\n",
      "Epoch: 180 - Validation Loss: 0.7410\n",
      "Epoch: 190 - Validation Loss: 0.7327\n",
      "Epoch: 200 - Validation Loss: 0.7241\n",
      "Epoch: 1 - Validation Loss: 2.7529\n",
      "Epoch: 10 - Validation Loss: 2.2280\n",
      "Epoch: 20 - Validation Loss: 1.9128\n",
      "Epoch: 30 - Validation Loss: 1.6499\n",
      "Epoch: 40 - Validation Loss: 1.4725\n",
      "Epoch: 50 - Validation Loss: 1.3511\n",
      "Epoch: 60 - Validation Loss: 1.2545\n",
      "Epoch: 70 - Validation Loss: 1.1903\n",
      "Epoch: 80 - Validation Loss: 1.1370\n",
      "Epoch: 90 - Validation Loss: 1.0936\n",
      "Epoch: 100 - Validation Loss: 1.0589\n",
      "Epoch: 110 - Validation Loss: 1.0321\n",
      "Epoch: 120 - Validation Loss: 1.0102\n",
      "Epoch: 130 - Validation Loss: 0.9880\n",
      "Epoch: 140 - Validation Loss: 0.9727\n",
      "Epoch: 150 - Validation Loss: 0.9559\n",
      "Epoch: 160 - Validation Loss: 0.9429\n",
      "Epoch: 170 - Validation Loss: 0.9315\n",
      "Epoch: 180 - Validation Loss: 0.9229\n",
      "Epoch: 190 - Validation Loss: 0.9112\n",
      "Epoch: 200 - Validation Loss: 0.9066\n",
      "Epoch: 1 - Validation Loss: 2.6575\n",
      "Epoch: 10 - Validation Loss: 2.1074\n",
      "Epoch: 20 - Validation Loss: 1.7595\n",
      "Epoch: 30 - Validation Loss: 1.5126\n",
      "Epoch: 40 - Validation Loss: 1.3245\n",
      "Epoch: 50 - Validation Loss: 1.1957\n",
      "Epoch: 60 - Validation Loss: 1.0980\n",
      "Epoch: 70 - Validation Loss: 1.0295\n",
      "Epoch: 80 - Validation Loss: 0.9753\n",
      "Epoch: 90 - Validation Loss: 0.9251\n",
      "Epoch: 100 - Validation Loss: 0.8853\n",
      "Epoch: 110 - Validation Loss: 0.8548\n",
      "Epoch: 120 - Validation Loss: 0.8298\n",
      "Epoch: 130 - Validation Loss: 0.8072\n",
      "Epoch: 140 - Validation Loss: 0.7886\n",
      "Epoch: 150 - Validation Loss: 0.7714\n",
      "Epoch: 160 - Validation Loss: 0.7584\n",
      "Epoch: 170 - Validation Loss: 0.7453\n",
      "Epoch: 180 - Validation Loss: 0.7367\n",
      "Epoch: 190 - Validation Loss: 0.7287\n",
      "Epoch: 200 - Validation Loss: 0.7197\n",
      "Epoch: 1 - Validation Loss: 2.6277\n",
      "Epoch: 10 - Validation Loss: 2.1306\n",
      "Epoch: 20 - Validation Loss: 1.8491\n",
      "Epoch: 30 - Validation Loss: 1.6320\n",
      "Epoch: 40 - Validation Loss: 1.4573\n",
      "Epoch: 50 - Validation Loss: 1.3324\n",
      "Epoch: 60 - Validation Loss: 1.2417\n",
      "Epoch: 70 - Validation Loss: 1.1665\n",
      "Epoch: 80 - Validation Loss: 1.1059\n",
      "Epoch: 90 - Validation Loss: 1.0657\n",
      "Epoch: 100 - Validation Loss: 1.0299\n",
      "Epoch: 110 - Validation Loss: 1.0008\n",
      "Epoch: 120 - Validation Loss: 0.9775\n",
      "Epoch: 130 - Validation Loss: 0.9627\n",
      "Epoch: 140 - Validation Loss: 0.9434\n",
      "Epoch: 150 - Validation Loss: 0.9249\n",
      "Epoch: 160 - Validation Loss: 0.9120\n",
      "Epoch: 170 - Validation Loss: 0.9023\n",
      "Epoch: 180 - Validation Loss: 0.8934\n",
      "Epoch: 190 - Validation Loss: 0.8841\n",
      "Epoch: 200 - Validation Loss: 0.8780\n",
      "Epoch: 1 - Validation Loss: 2.7481\n",
      "Epoch: 10 - Validation Loss: 2.2296\n",
      "Epoch: 20 - Validation Loss: 1.8714\n",
      "Epoch: 30 - Validation Loss: 1.6096\n",
      "Epoch: 40 - Validation Loss: 1.4231\n",
      "Epoch: 50 - Validation Loss: 1.2777\n",
      "Epoch: 60 - Validation Loss: 1.1803\n",
      "Epoch: 70 - Validation Loss: 1.1111\n",
      "Epoch: 80 - Validation Loss: 1.0460\n",
      "Epoch: 90 - Validation Loss: 1.0085\n",
      "Epoch: 100 - Validation Loss: 0.9683\n",
      "Epoch: 110 - Validation Loss: 0.9389\n",
      "Epoch: 120 - Validation Loss: 0.9146\n",
      "Epoch: 130 - Validation Loss: 0.9005\n",
      "Epoch: 140 - Validation Loss: 0.8764\n",
      "Epoch: 150 - Validation Loss: 0.8619\n",
      "Epoch: 160 - Validation Loss: 0.8508\n",
      "Epoch: 170 - Validation Loss: 0.8402\n",
      "Epoch: 180 - Validation Loss: 0.8257\n",
      "Epoch: 190 - Validation Loss: 0.8166\n",
      "Epoch: 200 - Validation Loss: 0.8072\n"
     ]
    }
   ],
   "source": [
    "#CV+\n",
    "\n",
    "from third_party.ARC.arc import methods\n",
    "\n",
    "model = NeuralNetwork(input_size = N_FEATURES, num_neurons = NUM_NEURONS, out_size = N_CLASSES,\n",
    "                      seed = SEED, l2_reg = REG).to(device)\n",
    "method_cv = methods.CVPlus(Xtrain, Ytrain, model, n_folds=5, seed = SEED, verbose = False) #building CV+\n",
    "sizes = {}\n",
    "\n",
    "for i, epsilon in enumerate(epsilons): #repeat for each epsilon\n",
    "        for j in range(len(Ytest)):\n",
    "            set = method_cv.predict(check_tensor(np.expand_dims(Xtest[j], axis=0)).to(device), alpha = epsilon) # type: ignore #constructing prediction sets\n",
    "        \n",
    "            if epsilon not in sizes:\n",
    "                sizes[epsilon] = [len(list(set[0]))]\n",
    "            else:\n",
    "                sizes[epsilon].append(len(list(set[0])))\n",
    "            \n",
    "curve_all_CV_plus = [np.mean(sizes[epsilon]) for epsilon in sizes] #average prediction set size w.r.t epsilon\n",
    "AUC_all_CV_plus = round(simps(curve_all_CV_plus[:21], epsilons[:21]),3) #AUC"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We measure the efficiency with the  AUC in the interval $\\varepsilon \\in [0,0.2]$ (the lower the AUC; the higher the efficiency). ACP is consistently better than existing alternatives in terms of statistical power: it attains the desired error rate with tighter prediction sets. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.patches as patches\n",
    "fig, ax = plt.subplots(figsize=(10, 6))\n",
    "\n",
    "plt.plot(epsilons, curve_all, label =\"ACP (D) (AUC = {})\".format(AUC_all), linewidth = 3)\n",
    "plt.plot(epsilons, curve_all_split_cp, label =\"SCP (AUC = {})\".format(AUC_all_split_cp), linewidth = 3, color=\"green\")\n",
    "plt.plot(epsilons, curve_all_RAPS, label =\"RAPS (AUC = {})\".format(AUC_all_RAPS), linewidth = 3, color = \"red\")\n",
    "plt.plot(epsilons, curve_all_CV_plus, label =\"CV+ (AUC = {})\".format(AUC_all_CV_plus), linewidth = 3, color = \"m\")\n",
    "\n",
    "rect = patches.Rectangle((0.001,0), 0.2, N_CLASSES - 0.01, ls=\"--\", linewidth=3, edgecolor='black', facecolor='none', zorder = 5)\n",
    "ax.add_patch(rect)\n",
    "\n",
    "plt.legend(fontsize = \"24\", shadow = True, facecolor='white', framealpha=1)\n",
    "plt.xscale(\"linear\")\n",
    "plt.xlim([-0.0001, 1.0])\n",
    "plt.ylim([-0.01, N_CLASSES])\n",
    "plt.xticks(fontsize = \"28\")\n",
    "plt.yticks(np.arange(N_CLASSES+1).tolist(), fontsize = \"28\")\n",
    "plt.ylabel(\"Prediction set size\", fontsize = \"36\")\n",
    "plt.xlabel(r\"$\\epsilon$\", fontsize = \"30\")\n",
    "\n",
    "plt.grid()\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 6))\n",
    "\n",
    "plt.plot(epsilons[:21], curve_all[:21], label =\"ACP (D) (AUC = {})\".format(AUC_all), linewidth = 3)\n",
    "plt.plot(epsilons[:21], curve_all_split_cp[:21], label =\"SCP (AUC = {})\".format(AUC_all_split_cp), linewidth = 3, color=\"green\")\n",
    "plt.plot(epsilons[:21], curve_all_RAPS[:21], label =\"RAPS (AUC = {})\".format(AUC_all_RAPS), linewidth = 3, color = \"red\")\n",
    "plt.plot(epsilons[:21], curve_all_CV_plus[:21], label =\"CV+ (AUC = {})\".format(AUC_all_CV_plus), linewidth = 3, color = \"m\")\n",
    "\n",
    "plt.legend(fontsize = \"24\", shadow = True, facecolor='white', framealpha=1)\n",
    "plt.xscale(\"linear\")\n",
    "plt.xlim([-0.0001, 0.2])\n",
    "plt.ylim([-0.01, N_CLASSES])\n",
    "plt.xticks([0, 0.05, 0.1, 0.15, 0.2], fontsize = \"28\")\n",
    "plt.yticks(np.arange(N_CLASSES+1).tolist(), fontsize = \"28\")\n",
    "plt.ylabel(\"Prediction set size\", fontsize = \"36\")\n",
    "plt.xlabel(r\"$\\epsilon$\", fontsize = \"30\")\n",
    "\n",
    "plt.grid()\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\\textbf{EXPERIMENTS WITH REAL DATA}$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us now see how ACP would perform in CIFAR-10. We load the dataset, apply the pre-processing and optionally reduce the dimensionality of the images with an autoencoder (AE)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import cifar10\n",
    "from acp.utils import normalize_flatten, scalar\n",
    "from acp.models import AE\n",
    "\n",
    "#1-loading cifar-10\n",
    "(Xtrain, Ytrain), (Xtest, Ytest) = cifar10.load_data() \n",
    "\n",
    "#2-Applying pre-processing\n",
    "Xtrain = normalize_flatten(Xtrain) \n",
    "Xtest = normalize_flatten(Xtest)\n",
    "Ytrain = scalar(Ytrain)\n",
    "Ytest = scalar(Ytest)\n",
    "\n",
    "#3-We reduce the dimensionality with an AE. Here we set an embedding of size 8\n",
    "EMBEDDING_SIZE = 8\n",
    "autoencoder = AE(input_shape=Xtrain.shape[1], embedding_size=EMBEDDING_SIZE, seed = SEED).to(device) \n",
    "autoencoder.load_state_dict(torch.load(\"models/AE_CIFAR-10_\"+str(EMBEDDING_SIZE), map_location = 'cuda'))\n",
    "Xtrain_emb = np.array(autoencoder.embed(Xtrain))\n",
    "Xtest_emb = np.array(autoencoder.embed(Xtest))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Similarly to the previous experiment, we build ACP using the CIFAR-10 training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training model on 50000 samples\n",
      "Epoch: 1 - Validation Loss: 2.6859\n",
      "Epoch: 10 - Validation Loss: 2.0244\n",
      "Epoch: 20 - Validation Loss: 1.9764\n",
      "Epoch: 30 - Validation Loss: 1.9665\n",
      "Epoch: 40 - Validation Loss: 1.9663\n",
      "Epoch: 50 - Validation Loss: 1.9619\n",
      "Computing gradients and Hessian inverse\n",
      "Computing scores for 50000 samples\n",
      "Conformal predictor is ready!\n"
     ]
    }
   ],
   "source": [
    "NUM_NEURONS = [20,10] \n",
    "REG = 0.01\n",
    "model = NeuralNetwork(input_size = Xtrain_emb.shape[1], num_neurons = NUM_NEURONS, out_size = len(np.unique(Ytest)), seed = SEED, l2_reg = REG)\n",
    "ACP = ACP_D(Xtrain_emb, Ytrain, model, seed = SEED, verbose = True) # type: ignore"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, we generate prediction sets for an image chosen uniformly at random. We fix a significance $\\varepsilon=0.1$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r\n",
      "  0%|                                                                                            | 0/1 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing p-values for 1 samples\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 1/1 [02:14<00:00, 134.19s/it]\n"
     ]
    }
   ],
   "source": [
    "cand = random.randint(0, len(Ytest)-1)\n",
    "eps = 0.1\n",
    "sets = ACP.predict(Xtest_emb[cand].reshape(1, -1), eps, out_file = \"results/CIFAR-10\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 460.8x316.8 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "True label: truck\n",
      "Prediction set: ['airplane', 'automobile', 'ship', 'truck']\n"
     ]
    }
   ],
   "source": [
    "labels = [\"airplane\", \"automobile\", \"bird\", \"cat\", \"deer\", \"dog\", \"frog\", \"horse\", \"ship\", \"truck\"]\n",
    "\n",
    "image = Xtest[cand].reshape(32,32,3)\n",
    "fig = plt.figure\n",
    "plt.imshow(image)\n",
    "plt.show()\n",
    "\n",
    "print()\n",
    "print(\"True label: \" + labels[Ytest[cand]])\n",
    "print(\"Prediction set: \" + str([labels[label] for label in sets[0]]))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "acp",
   "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": "3.10.10 (main, Mar 21 2023, 13:41:05) [Clang 14.0.6 ]"
  },
  "vscode": {
   "interpreter": {
    "hash": "9a7f6057719d348c6ce2327367d4ca6e817de6050c50fa45956189294cd9b9a9"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}