1.Activation_Maximization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# I. Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import torchvision\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "\n",
    "from nn_interpretability.interpretation.am.general_am import ActivationMaximization\n",
    "from nn_interpretability.interpretation.am.am_in_codespace import ActivationMaximizationCodespace\n",
    "from nn_interpretability.model.model_trainer import ModelTrainer\n",
    "from nn_interpretability.model.model_repository import ModelRepository\n",
    "from nn_interpretability.visualization.mnist_visualizer import MnistVisualizer\n",
    "from nn_interpretability.dataset.mnist_data_loader import MnistDataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist_data_loader = MnistDataLoader()\n",
    "\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "MnistVisualizer.show_dataset_examples(mnist_data_loader.trainloader)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# II. General Activation Maximization\n",
    "\n",
    "The goal of ActivNation Maximization is to find an input which produces the highest model output for a given class. This reduces to the following optimization process.\n",
    "\n",
    "![Activation Maximization Formula](./assets/activation_maximization_formula.png)\n",
    "\n",
    "Our implementation is based on the paper \"Methods for Interpreting and Understanding Deep Neural Networks\n",
    "\" by Montavon et al, which can be accessed [here.](https://arxiv.org/pdf/1706.07979.pdf) We consider three starting points:\n",
    "\n",
    "1. Random image from dataset\n",
    "2. Random noise\n",
    "3. Mean class image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'model_cnn.pt'\n",
    "train = False\n",
    "model = ModelRepository.get_general_mnist_cnn(model_name)\n",
    "\n",
    "if train:\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    optimizer = optim.Adam(model.parameters(), lr=0.0005)\n",
    "\n",
    "    model.train()\n",
    "    ModelTrainer.train(model, criterion, optimizer, mnist_data_loader.trainloader, 1)\n",
    "    \n",
    "    ModelRepository.save(model, model_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. AM from random image for each class\n",
    "\n",
    "In the first case, we randomly draw an image from the dataset for each class and start the optimization process from there."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_shape = (28, 28)\n",
    "lr = 0.005\n",
    "reg_term = 1e-3\n",
    "epochs = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAACYCAYAAACfzCG6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAAV+UlEQVR4nO3dfbQdVXnH8d9DXu7N2yUJCZDwogTSQk2RilialqorSCOVhW9YWAqVCtauVV30xdraquBbq6t1VShdxRaUNqUFjcRqeemb1lpiQyyUAPISKMF4Q0KK5D0kJLt/zNzNmefse2bO5Jx7zs39fta6C/aZmT17zrl3nyezn9nbQggCAACQpCN63QAAANA/CAwAAEBEYAAAACICAwAAEBEYAACAiMAAAABEHQkMzOwvzOwjnd63n5jZ1Wa2ouax7zaz77TYfqeZ/XJqXzPbaWaL6pwXwMRCX1x6LH1xBaWBgZk9ZWZ7zGyHmT1vZveY2fvMLB4bQnhfCOETVU7YuK+Zvc7MNlY5zsymmtkjVffPj3m3mR3IP9DtZna/mb2p6vFjJYTwxhDCzaNsmxlCeFKSzOxLZvbJuucxs9PM7N/MbJuZrTezt7jtV+Sv7zSzu8xs4SHUtSz/vHab2TfN7GV12w2g931x/oW8P+8fRn4qfVHSFxd1si/O97/YzL5vZrvM7AkzO6dh23Qz+3Mz25qf79tl7at6x+CCEMIsSS+T9EeSPiTpxorHdsoHJW2pcdzqEMJMSbOVtfk2M5vrdzKzyYfWvP6WX9/XJH1D0lxJ75W0wsx+LN/+WkmflnRhvv1/Jf1dzbrmSfqqpI/k29dKurVb1wZMIL3ui2/NvyBnNn5RVkRfrM72xfn+b5D0GUmXS5ol6eclNX4uX8jrOS3/72+UNjKE0PJH0lOSznWvvUbSQUlL8vKXJH2yYfvvSNokaVjSFZKCpFMa95U0Q9KevJ6d+c/CUdpwkqTvS3qjpI1lbW447t2SvtNQnpG35dWSrpb0FUkrJG3P27lQ0j9Iek7SeklXNhw7sv+tknZI+m9Jr2zY/ruSnsi3PSzpLa4d/ynpOknbJD0iaVnD9m9JumKUNgdJpyj75dkvaV/+Xn1dWbC00l3zdZL+NPFeLMmPs4bX/knSJ/L//2NJ1zdsW5if++Qadb1X0j3ufd8j6dSqnx0//PBT/Ol1X5z3gStqtp2++KXXO9YX59vvkfSeUbb9eP6eDrXzedXKMQghrJG0UdI5fpuZLZf0m5LOzd/E145Sxy5lX/TD4aXoc3iUU14n6cPKfnlryaO0K5R9II/nL1+o7BdstqS/VRaVbVT2Qbxd0qfNbFlDNRdK+rKyqOsWSavMbEq+7Qll78eRkq5RFgEuaDj2p5VFcfMkfUzSV1PR8mhCCF/I2/jZ/L26QNkf0nIzm91wjb8k6W9Sb8Eory1p+H9z29SwvZ26XiHpfxravkvZ+/OKxHEAaupBX3yBmT1nZg+Z2a/VaTN9cef6YjObpCy4mp8PPWw0sz8zs2kN17pB0jX5UMI6M3tb2TUeSvLhsLIPxXuHpC+GEB4KIexW9sHUlo+9TA4h3F6zirPN7HlJz0i6RFn0uC3ftjqEsCqEcFDZL8nPSfpQCGFvCOF+SX8l6dKGur4XQvhKCGG/pM9JGpR0tiSFEL4cQhgOIRwMIdyq7Bf+NQ3HblEWPe7Ptz8q6RdrXpPyc26S9G1JF+UvLZe0NYTwvcTuj+Rt+KCZTTGz85R1FNPz7XdIeoeZnZ7/Un1UWZQ6vUZdM5VF4422KbvNBaCzxqQvlnSbstvR8yVdKemjZnZJG8fTF2c62RcfI2mKsuDpHElnSPopSX+Qbz9eWUCxTVmQ9euSbjaz01pdz6EEBscpu83jLZT0g4byDxL7VGJmMyR9VtL769Yh6bshhNkhhHkhhLNDCP8yStsWSnouhLCj4bUNyq6zaf/8F3gkopWZXZYn1Dyf//IvUfYLPuKHIb+301B3y4SSim6W9K78/9+ldISq/A/ozcr+AJ6R9FvK/tA35tv/VVn0vDJv21PKbsU1JSSV1aXsXwJD7rChvD4AndX1vliSQggP51+4B0II90j6vLIvpKroi9XZvlgv3UW/LoSwKYSwVVmgdH7D9v3Khpf2hRD+XdI3JZ3X6kJqBQZmdpayDyn12McmZVHKiBNaVFW2tONiSS+X9B9m9oyyhLYFZvaMmb28coOrnX9Y0lwza/xX7YmSfthQjteSZwIfL2k4z7j/S2XR2FEhhNmSHlTxdtBxZtZYPjE/Z932jlgl6XQzWyLpTcpucaUPDuGBEMJrQwhHhRB+QdIiSWsatl8fQlgcQjha2S/l5Pw62q3rIUmvHNk3D/BOzl8H0CFj2BePdkzqtngd9MU1+uIQwo+UBQyjfX4PVLyegrYCAzMbyh8x+XtlSSjrErvdJuny/HGM6cpug4xms6SjzOzIUbY/qOwX4Iz854r8mDOUR4xm9i0zu7qd60gJIfxAWRLHH5rZoJmdLuk9Kn64Z5rZW/Pxo6skvSDpu3opkebZvE2Xq3k86GhJH8hvHV2k7JbcHW02c7OyX6DGdu9VNjZ3i6Q1IYSnRzs4vzU1mD++8tuSFihLQFL++hLLnKgsk/Xz+S9eW3VJul3SEjN7m5kNKvsdeCCE8Eib1wsgoQd9sczsQjObk/cRr5H0AWXZ9SPb6Yt70BdL+qKk95vZ0WY2R9n78Y1827clPS3p98xsspn9rKTXSbq71cVVDQy+bmY7lH0Z/76yWxWXp3YMIdwp6VpltyvWS1qdb3ohse8jypJMnsxv+yx0218MITwz8qPsdtnBvHwg3+0EZVmmnXCJsjsUw8q+3D4WQvjnhu1fU5ZQ8iNl411vzcepHpb0J/m1bpb0k4k2/ZeyOyBbJX1K0ttDCP/XZvtulPQT+Xu1quH1m/NzJm9dNbhU2b8itkhaJukNIYSRz2VQ2S/0TmWR62pljxtKkszsw2Z2Z5W6QgjPSnpbfp0/UpYAc3Gb1wqgWU/64tzFeT07JP21pM+E4jP/9MW96Ys/IeleSY8pe3rvvvy6RoYtLlQ2tLBN2d2Uy8r+kWbFoZbOy5McHpQ0EEJ4scN1Hy/pyyGEn+lkveNNHlU+IunYEML2XrcHQP+hL+6+w6Uv7kpgYNmTBP+o7LbOzcr+lf/mjp8II+Nrn1P2nOqv9Lo9APoHffHYOZz64m4tovSrysZ4npB0QFKt513RWp7Ut13SG5RlsQJAI/riMXC49cVdH0oAAADjB8suAwCAiMAAAABEXV3FyswYpzhMhRA6NbEJgC6iHz6sbQ0hzO90pYf18pYjjjiieGOkOOlVWpV9GqVyNfxrBw8ebHkO8j0A9CPfV/my79vq1Dme+792vy+k5uut+X5saPvEFTCUAAAAop7eMUhFWd2IGv0dAx/dTpo0qePnlJqvr+w8+/fv70o7AGBEnX+Zlt0xqNKXT55c/Lrx/XCVdvg6XnyxOE+T7+tT5+mG8Xy3I4U7BgAAICIwAAAAEYEBAACI+u6pBD9G5Mt+TKkKf0yVjNoqY2jttsvX4cfLUjkIZU82lOVPVGnH4TY+BqCosZ/wf/8HDhzwuzfx/crUqVPbrsPvU6ffKWvHvn372q6zX/RTP8wdAwAAEBEYAACAiMAAAABEBAYAACDqafJhKtnCJ+D5ZBOfsFeFr6NsauIUv0+dJEifKOgnNEpNzuFf8+9PlURK39Z+SnIB0H2Nf/N1/v47MQWyVzZZUcr06dML5Z07dxbKvZrgaP784nIFzz77bNM+AwMDhbK/3ioJnGOFOwYAACAiMAAAABGBAQAAiLqeY5Aa82m1rd2Fh1LazQdIbfdt68TkHL7OssmLUvz7UbZ0pyRNmTKl5T5VzlsnpwJA/+lEjkGVftr3mTNmzCiUfZ+S6mN8HsLu3btbtiOVg9aJSY9822fOnFkoz5kzp1A+6qijmurweQc+P6LOolLdwh0DAAAQERgAAICIwAAAAERdzzFo9xnSducpSI3D+PkB6ujGs6915hPw7Vi6dGmhfMsttxTKy5cvb6pj/fr1Lc9bVgYwvjWOxXcix8DXkXoGf2hoqOU+fk4CvyCS1JxTUNYv18kn8PkAqdy3E088sVD2OQf+2nbt2tVUx7Rp0wrlvXv3Fsr+e4scAwAA0BcIDAAAQERgAAAAop6ulZAaL/LjKq961asK5blz5xbKd911V1Md/pnafnkGvxNjRmeddVahvHbt2rbPW5aDUTa/BDkIwPhyqH+zvk/1Y/l+rhRJ2rNnT6Hscw58TkGqX/LzBfi+3M8FkDJv3rxC+cwzzyyUBwcHW7ZTku67776Wxzz//POFcirHwLe1G+tPdAp3DAAAQERgAAAAIgIDAAAQERgAAIBoTBdRqjKRjk9Aef3rX18oL168uFC+++67m+roRrKhn3ipVwmNixYtKpT9xBt+QhCp+T31iUSpyUk8kg+B8avdv9myRZOq8BMHlU0KlOqHtmzZ0vZ5veOOO65QnjVrVqG8YMGCQnlgYKCpDp9suG7dupbnrJIUWbYQVZV+uVu4YwAAACICAwAAEBEYAACAqOs5Bo3jKFUmcPDjLpdddlmhvHr16kK5E5NC+LEdqXl8JzXpz1jwk3NceeWVhfKKFSsK5aeeeqq0zrKxq36aaANAd6XyB9rNSUgtfuf7zO3btxfKCxcuLJTrLIBUxaZNmwplP/nQ/fffXyincgz8Mb6tfkGklLIJjfz71YnPpS7uGAAAgIjAAAAARAQGAAAgGtMcgyr8WFUnxvZ9GzoxTuPb1a1x+RtvvLHl9scff7wr5wUwMaT6w3YXovMLJknNfaRfAG/btm2FcurZ/yOPPLLlMVX4uRA6MTdCHWXfO/47pM7cEZ3CHQMAABARGAAAgIjAAAAARF3NMTCztsdJ/FoAxxxzzCG3w+ct+LUDUnkMqbkNWm1P1dHutaf292NsXmqtCAA4FN2Yp9/X6ecLeOGFF5qOSb02UfRyTRruGAAAgIjAAAAARAQGAAAgIjAAAABRV5MPQwhNiX5lzj///EJ52rRpLfefMmVK02v+nD7ppc6ER36fOnX4Y3z56KOPbjrmpJNOalnn8PBw2+ftZVILgP5Xljjt+5DUwkN+UqTdu3cXynPmzCmUZ8yY0VTHRE4+7CXuGAAAgIjAAAAARAQGAAAg6voiSu069dRTW25/8MEHC+XUWFgq76DsGM+PoZUtIlLH1KlTC+VrrrmmaR8/wdNjjz1WKO/YsaP0POQUAGhHu31GakIk32f6fAE/KVwqx8DnIXibN29ueU5J2rt3b8s6qhirRfP6BXcMAABARGAAAAAiAgMAABD1XY5BmXvvvbdQrjN+XmfOgTr8s73Lly8vlC+99NJC+bzzziut8+Mf/3ih7MfUUvkVY5EvAWDiqtOn+HH7k08+uWmfU045pVD2uQzr1q0rbcfGjRsL5a1btxbKQ0NDhfL27dub6vA5BYODg4WyzxdL1TGecMcAAABEBAYAACAiMAAAANG4yzGYO3duoVwnF2DRokWFsh/rkqRzzz23UD7++OMLZT+m9M53vrOpDl+vnyt8zZo1hXJqXvDJk4sf0dq1awvlOtfv6yTnAMBYmzdvXqGc6suOPfbYQtn3kf774Lnnnmuqw+d6zZo1q1BO9f9lyuZGmDRpUtNr/jztriM0lrhjAAAAIgIDAAAQERgAAICIwAAAAER9l3y4Z8+eQtlPznPDDTcUyk8//XTb5zj99NML5VTSi08M8e16+OGHC+WbbrqpqQ4/GZNPHNyyZUuh/OijjzbV4SfSGB4ebtqnjL8+X/aTIqUWRGn8HFiUCcCh8hMNpRL6fDKhTz70EwmlFpXzydWzZ89u2a5UErjv/719+/YVyuN9kSXuGAAAgIjAAAAARAQGAAAg6rscg6uuuqpQ3rBhQ6G8dOnSQz6Hz0tYtWpV0z4PPfRQoexzCurwY/N+UqT58+c3HfPkk08Wyn7syk+a0a3xf/IKABwKP+mPzzHwZak5H8r3Q77OVH6Un4zO54/5PtVPAJd6zdfp2+W3S825Drt27Wrap19wxwAAAEQEBgAAICIwAAAA0ZjmGPjxoiqLV1x77bWF8vXXX18oVxn79mM7dcbL/bP+VRYvKjvPsmXLSutYuXJly/P6c6TG2MryEFLHAEAn1elnyvrQKnX6OQZ8joHvU/3+UvN3iM9tmD59emkdPu/At923K3VtZf1/p3DHAAAARAQGAAAgIjAAAABR13MM/FhMo9Q4vR8z6cRz+mXHpHIdWrU7VWfqWqrkIZS5/fbbC2U/DuVzH1JzdI/3ebsBoFN83162Pk9Kaj2FRn6NmxT//VAlX2Ks5pPhjgEAAIgIDAAAQERgAAAAIgIDAAAQdT35sDFZomwBIKk86a+Ksgks/Dk6kSTozyk1J4r4RME6fNtT5wWA0fh+N9Xn+kQ435eNp0XVyhZa8hMP7d27t+1z+Pc0lfBd9h6O1eRFVXDHAAAARAQGAAAgIjAAAABR13MMWk2ukxpDKZvAxx/jF7eocp4qx3RD2XlTuQ6LFy8ulNetW1co+7Etcg4AtFJlMbtp06YVymV5WalFg3y9fmzfl1OTBpW11ecH7N69u6kOv8DRwMBAobxr167SdpQp+95Ktc1/N/ZT3gZ3DAAAQERgAAAAIgIDAAAQdT3HoJUqYyrjeczcj49Nntz67U69H35MrexZ19Q5epVTAaD/pfID/Ni973eGhoYK5VTf5cfZfd/kz1GlbZs3by6Uff+Yyg/w3yH+vD7HoAp/Lb4dqbyNWbNmFco+56BObkO3cMcAAABEBAYAACAiMAAAANGY5hjUmQvaj+X063h5lXUQ6qzJsHTp0kJ55cqVbddZZR5vABODnz8gZceOHS23+3H71Jj64OBgoez78hkzZhTKfr4BqXndghNOOKFQ3rZtW6Gc+k7xfeSePXua9mmXv7ZUnobnv7uqfA69wh0DAAAQERgAAICIwAAAAEQEBgAAIBrT5MM6i0T0a7KhT2hJtdMn25Rdf53kxCrvqV8AxZd9MmIqObGfFvgA0Ft+cp5U3+X7DN+v+MTCnTt3NtXhj/F9l0/69kmBqfP4Oqrwi0r56/d9fZVr6eckcO4YAACAiMAAAABEBAYAACDq6SJK45kfP0tN8OHzDvwxd9xxR6F80UUXtX3eTqiT2wBg4vD9W5X+z4+z+37GLxqUGvv39fqJl3xOQaov8/2w32dgYKBQTk08VJbr4M/Rz/kDVXDHAAAARAQGAAAgIjAAAACRdfP5dDObMA+/V3mOtw7/fKw/j1/MxO+f4uvw42FVFvcIIZCYAIwD3eiHfR+S6v/GYpzd93epPAXfDj8ngc91SOnjBZC+F0J4dacr5Y4BAACICAwAAEBEYAAAACLmMeiQbuVqtLtWRGr/smduU88gA8BofH/Xq7VU/Fh/lf7Pr3PgcxBS1zLR+siJdbUAAKAlAgMAABARGAAAgIjAAAAARCQfjsJP2NGr5JpOKJuMY7wv+AFgYqrSL3diMqKJ1kdyxwAAAEQEBgAAICIwAAAAUbdzDLZK2tDlc3TFeM4pGAMv63UDAFQ2bvvhftHH3wdd6Yu7uroiAAAYXxhKAAAAEYEBAACICAwAAEBEYAAAACICAwAAEBEYAACAiMAAAABEBAYAACAiMAAAANH/AxU+26W8G774AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.get_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "    mean_img = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "    \n",
    "    interpretor = ActivationMaximization(model, [], None, img_shape, img, i,  lr, reg_term, mean_img, epochs, threshold=0.999)\n",
    "    end_point = interpretor.interpret()\n",
    "    \n",
    "    scores = model(end_point.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "    \n",
    "    images.append(end_point.reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.1. AM from random noise with class mean image as regularization\n",
    "\n",
    "In the second case, we start the optimization process from random noise with class mean image as regularization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_shape = (28, 28)\n",
    "lr = 5e-3\n",
    "reg_term = 1e-2\n",
    "epochs = 10000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAACeCAYAAABJlcKnAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAAWmElEQVR4nO3de7BdZ1nH8d/TNMlJTnKS5tqkTeNQqlJrKaPVOmMBByoFQQrDRQYQqgg4A44DAo43UBCFGS5anUEEoVCRS9FCgXpBQEYoLWAhlBLapm3MpUmb28mlzYX28Y+1zuvez3pOzjo7++xzTs73M9Np1trvXutde+/z7ues9znPa+4uAAAASTpjujsAAABmDgIDAABQEBgAAICCwAAAABQEBgAAoCAwAAAARV8CAzN7n5n9cb/bziRm9hYzu67H577czP77JI/fZGYvy9qa2WEze0wv5wUwtzAWT/hcxuIWJgwMzOw+M3vYzA6Z2QEz+7qZvdrMynPd/dXu/tY2J+xsa2ZPNrPtE5zfzOwdZra3/u+dZmZtzlW/sY/Ub+hBM/uOmT2zzXMHyd2f7u7XjvPYEne/R5LM7MNm9rZez2NmjzOzL5nZqJndbWbPCY+/ot5/2Mz+1czWn+RYXzGzo3Xbw2b2w47HLjSzb5nZ/vq/L5rZhb32G8CMGIt/ycy+XI8f902y74zFHfo5Ftftf83MfmBmR8xsi5ld3vHYC+rHDpnZHWZ21UT9a3vH4FnuvlTSRkl/KelNkj7Y8rmn6pWSrpL0eEkXS3qmpFdN4vk3u/sSSctV9fmTZrYiNjKzM0+5pzNYfX2fkfQ5SStUva7XmdmP148/SdLbJT27fvxeSf80wWFfU/+wLHH3n+jYv1PS8+rjrJL0WUkf7+PlAHPVdI7FRyT9g6Q39Ph8xmL1fyw2syskvUPS1ZKWSnqipLEA5hxJ10l6naQRVe/dx8xszcn6OKmpBHcfdffPSnqhpJeZ2UX1ybuiJzN7o5ndb2Y768jHzeyxnW3NbFjSTZLWd/zWmUVFL5P0Lnff7u47JL1L0ssn0++674+q+lAvkvQYq25HXW9m15nZQUkvN7P1ZvZZM9tXR2u/FQ4zZGafqCOv/zGzx3dc8+/XkdpYVPac8Fwzs2vqCHGzmT2l44GvmNkrsn6PvXZm9kpJL5b0xvq1utHM3mBmnw7trzGz9yaH+klJ6yW9x90fcfcvSfqapJfWjz9L0qfc/fvuflzSWyU90czOP9nrmnH3A+5+n1dlNU3SI5IeO9njAMhNx1js7re6+0dVf+mcQt8Zi/s7Fv+ppD9z92+4+6PuvqP+rpSkcyUdcPebvPJ5VQHeScf1nnIM3P1WSdslXR4fM7MrVUUnT1X1ZfCkcY5xRNLTJe3s+K1zZ9L0pyR9t2P7u/W+SamjtFdIOizprnr3syVdryqC/UdVUdl2VW/a8yS9vfNDU7f/lKoo7mOSbjCz+fVjW1S9HstUvVHXmdm6juf+vKofqFWS3izpn7NoeTzu/v66j++sX6tnqYoErzSz5R3X+EJJH81egnH2XdTxbwuPqePxzF+Y2R4z+5qZPblxcLMDko5KukZVBAygjwY8FvcFY3H/xmIzmyfpZyWtrgOo7Wb2N2a2qG7yLUk/MLNfNbN5Vk0jHJO06WTXeCrJhztVvSnRCyR9qI52HlL1xpyKJZJGO7ZHJS0xa5dnIOmy+gtql6QXSXqOu48d72Z3v6GOYFdJ+kVJb3L3o+7+HUkf0P9HcZL0bXe/3t1PSHq3pCFJl0mSu3/K3XfWEdsnVH3gf67juQ9Ieq+7n6gf/6GkX2l5DSl3v1/SVyU9v951paQ97v7tpPnmug9vMLP5ZvbLqgaKxfXjX5D0AjO7uP5Q/Ykk73g8epOkx0g6R9L7Jd0YI1p3X67qh/M1km7r6SIBTGRQY/GpYiyu9HMsXitpvqrg6XJJl0h6gqQ/qvv1iKSPqAqejtX/f1UdDI7rVAKDcyTtS/avl7StY3tb0mYyDquaGxkzIumwt1/96RvuvtzdV7n7Ze7+xXH6tl7SPnc/1LFvq6rrbLSvP8BjEa3M7NetSqg5UH/4L1L1AR+zI/R569hzT9G1kl5S//slyiNU1T9AV6n6Adgl6fWSPllfg9z9P1VFz5+u+3afpENjjyfHu8XdD7n7sTpZ52uSnpG0OyLpfZI+YhPMawHoyaDG4lPFWKy+j8UP1/+/xt3vd/c9qgKlZ0iSmT1V0jslPVnSAlUByAfM7JKTXUhPgYGZXarqTcr+7ON+VfMaYzac5FBtvty/ryrxcMzj63390Hn+nZJWmNnSjn3nSdrRsV2uxapM4HMl7TSzjZL+XtVvxivr35RvV/ftoHPCXY7z6nP22t8xN0i6uJ5jfKaqW1z5k903ufuT3H2luz9N1W/8t3Y8/rfufoG7r1H1oTyzvo62fRvvLs4ZqqLdc8Z5HEAPBjwWTyXG4h7GYnffrypgGO/9u0TSV939W/UdlG9KukXV9NK4JhUYmNmIVX9i8nFJ17n795Jmn5R0tVV/jrFY1W2Q8eyWtNLMlp2kzUckvc7MzqkTYl4v6cMdffqKmb1lMteRcfdtkr6uat58yMwulvSb6n5zf8bMnlvPH/2uqlsz35A0rOqNebDu09VqzgetkfQ79a2j50t6nKpbRpOxW9UHqLPfR1XNzX1M0q3u/r/jPbm+NTVkZovN7PckrVP9Wtb7L7LKeaqmB/6q/uDF4yw3s6fVzznTzF6sKhP23+rHrzCzJ9RzWiOqItj9kn4wyesFkJiOsdjMzjCzIVW3rq3++V/Q8Thj8YDH4tqHJL3WzNaY2VmqXo/P1Y99U9LlY3cIzOwJqqYc+pJjcKOZHVJ1++YPVQ30V2cN3f0mSX8t6cuS7pZ0c/3QsaTtZlVJJvfUt32y2zl/J+lGSd9TFTF9vt43ZoOq29j98CJJP6YqevwXSW929//oePwzqhJK9qua73puPU91h6q/lrhZ1Qfmp5M+3SLpAkl7JP25pOe5+95J9u+Dki6sX6sbOvZfW58zvXXV4aWqfot4QNJTJF3h7mPvy5CqD/RhVZHrzZJK8RMz+wMzu6nenC/pbap++PZIeq2kq9x9rJbBclXv66iqRKDHSrqy/sEB0LvpHIufqOrW9RdU/Zb9sKR/73icsXjwY7FU/dXCNyXdqeqXr9vq65K7/5ekt0i6vv7cfFrS2929831rsPZT9b0xs8ep+kJf6O4/6vOxz1X1Zx2/0M/jzjZ1VLlZ0tnufnC6+wNg5mEsnnqny1g8JYGBVX83+nlVt3WulfSou1/V9xNhbH7t3ZJG3P03prs/AGYOxuLBOZ3G4qlaROlVqm4zb1FV3Oa3p+g8c5pVhUkOSrpCVRYrAHRiLB6A020snvKpBAAAMHuw7DIAACj6uliFmXH7YQ5y97ZVKAEMAGPx3NSvsfi0XsWqF2ec0byJ8uijj/bl2BaqOA9yGieeO8O0EgC0d7qOq0wlAACAYs7fMWgT8bUxb968SR83uztx/PjxSZ87O9eZZ3a/tdldj7gvO85sjHYBzF3z58/v2n7kkUcabdrcBV64cGHXdjYWthlDY5vs3DNtnOWOAQAAKAgMAABAQWAAAAAKAgMAAFDMyOTDQf4JSDxOv46bJbzEZMNe/wwyJjq2SYppc66ZlgADAJM1NDTUtX3ixIlGm6NHJ17o9Uc/6l5navHixY02MVk8G2fj99lsSPLmjgEAACgIDAAAQEFgAAAAihmZYzDI+ZZ+zfvHgkJt9HquLH8BACajzdx3mwI+0XTPl2eF43qxYMGCCY87PDzctd0mn6FfJfanEncMAABAQWAAAAAKAgMAAFAQGAAAgGJGJh9OlSx5JBYLyopYxEIXmSVLlnRtt0nkyYpsxOSVNok02blIUAQwJhsj4tjSpk0bbcbLfonJf5K0evXqru2DBw822rQpcBS/G2IyotT8vhgdHW20ias9ZgmKMw13DAAAQEFgAAAACgIDAABQzKkcg6zwxqJFi7q24wIc2b6smFHMH4jzU9n5ly5d2mizbt26ru02BTOOHDkyYZtBzvsBmFnaLNyTjVnxeW0Wh+uXbCyOfd6wYUOjzcKFC0/6HEl64IEHJjz/yMhI13ZWmKhNTlgvBZ/avF+xTT8LS3HHAAAAFAQGAACgIDAAAAAFgQEAACjmVPJhVqAiJpTEYhSZWMxIaiYkHjt2rNEmJjpmxZRiwktWiOP48eNd2zt27Gi0iUU9Dh8+3Gjz8MMPN/YBOP1kiXMxeS2OK221ST6MbWKCoNRMnmszXmfJ2cuWLevazhIm47VfcMEFjTbx/NlYHBM29+3b12jTpphSG1OZbBhxxwAAABQEBgAAoCAwAAAAxWmTY5AVHYpzMNlcWJxHygpLxFyAAwcONNosX768azsWKpKklStXTtifmL+Q5Qa0KV4U5wsfeuihRps2Yh+zuUoAs0+/5qjjmJDlBsR9WTGlKFv8KPZ51apVjTZxvM4KycXxMSsSF8e+7LriuJoVZeplwabsvZnKnIKIOwYAAKAgMAAAAAWBAQAAKAgMAABAMaXJh1kiX7ZvIrEwkNRM+siSWWJhi6xNLJCRJdfFYhxZYaCYmJIV3ti/f3/XdpageOjQoa7ttWvXNtrElcGygkvDw8Nd21kiTxskGwKYjKxQUkyQ7nVciYmFWbJfHK9Xr17daBP7mCWCx6TBXbt2NdrERPQsibGNrAhT1Mt3Z6+4YwAAAAoCAwAAUBAYAACAYuA5BlGbQhe9LqYRC0JkBSJigYw4Ny81C2RkOQaxPw8++GCjTZx/2r59e6NNm6IacV4tmx+L1xXzGyRpdHS0sQ8A+q1fuUoxtywrbBcXUcpysGKbLVu2NNps3bq1azu7hpiH0GsRorh4X5vvs6nEHQMAAFAQGAAAgILAAAAAFAQGAACgmNLkwzYJJ23aZAUzYmJjVgQpJoZkxYLWrFnTtZ0l8sU+Zol8sehQtuJh7HNWdCgW59i0aVOjTUymWbFiRaPNvn37urZjcgsATJcs6Twm12XfDXEMzZLFY4LiZZdd1mgTV63NxuJeixX1Iks2nE7cMQAAAAWBAQAAKAgMAABAMfACR3EeqU2bNnM9sQiRJI2MjHRtr1+/vtFmw4YNXdvbtm1rtLn77ru7trPiRVMlFiqSpJUrV3Ztn3322Y02cQ7vlltu6W/HAJxW2uRXZWIRtrjQUXbsLMdg7969E54rFiaK25J06aWXdm1nBelijkGWE9aLNov5zQbcMQAAAAWBAQAAKAgMAABAQWAAAACKgScfxn1Z4Z1Y5CcWrJCaySLxOVIzSW/jxo2NNjFx5sCBA402u3btauybKnG1sCxxJrbJVg+LKydmiUUA5oZsLI6JctlKrseOHevazsbrWFwuK0gXx+dsvI7njwXqpOb4nCX7xWvNrj0mkPdrjJ+NiYYZvi0AAEBBYAAAAAoCAwAAUEz7IkrZ4hFxXisTCxpdcskljTZx7j3LVYhzVvfee++E555KbQptxAWSsuJO2bUCmJti0Tipt+JF2fiULSoXxTE95kBJ7ebn41h33nnnNdqce+65Xdv33HNPo00c99sU2ptLuGMAAAAKAgMAAFAQGAAAgILAAAAAFFOafJieMBTn6XVVq7iC11lnndVos3bt2q7t7du3N9rs2bOna7vNSo6DlBU4Ghoa6trO+hyLjvQqFkZqk7AEYOaLyXXZWByT8npNyIvHzpL92ojJkNnqinGM2rRpU6PN7bff3tP55wruGAAAgILAAAAAFAQGAACgGHiOQZtFlOIcUbbgxrp167q2s0JJcX4+zs1L0tatW8fv7DhGRkYa+w4ePDjp47SR5U4sX768a3t4eLjRZseOHV3b2UJLq1ev7trOriF7XQHMfjFfIFv8aFDnbisbD6O9e/f25VxzGXcMAABAQWAAAAAKAgMAAFAQGAAAgGLgyYdRm+S2LNkvFsxos8JgVtSizcpg0VQWQYqJhdl1xSShu+66q9Emrh526NChRpu4wln2XsSCVBQ4AjBdYgJ5VgDu8OHDXdtbtmyZ0j5NJCbcx6JxmTYrTU4l7hgAAICCwAAAABQEBgAAoBh4jkGcO8kKHLUpvBHnvletWtVoE/MHsvnxuNjQvHnzGm1irkK/5n/WrFkz4b6sOEcsypQVJorXnuUYtClo0usiVwCmV+dc9mzMDcrm4mOxu5UrVzbaxByD3bt397djJ5H1Ob722QJSbdpEU1m4iTsGAACgIDAAAAAFgQEAACgIDAAAQDHw5MOYZNEmKSYm/0nNwha33XZbo01MrssS6WIhoKkUE1Oywk0xuSZbWTL2ed++fY02sQhTlqgSEzizpEpWJgNmp4nG1phoPd1FdaKs/0ePHu3a3rx5c6NN/L7IEtyjOBZKvSVet/k+y9rEcTZLPhzkWMwdAwAAUBAYAACAgsAAAAAU076IUhsrVqxo7Itz79l8UJzTX7x4caPNhRde2LV9xx139NLFVuKCSFmf4zxflj+wd+/eru1jx4412sS5OABzVzbPvmDBgq7tbA47jqGxeNBUWrZsWWNfvI44FkrN4m7ZWBgLyWV5bFN1rb0WmyLHAAAATAsCAwAAUBAYAACAgsAAAAAU1s+EBjObkuyIbMXDjRs3dm2vW7eu0Wb16tVd21kSSkzAyRL5tm3bNuFxRkdHu7azZJ94HVlRjVi4KetPXClxuldOc/eJlwIDMDBxLM4KpcWxP2sTC+1kRZCy8Xmic2VjXxzrhoeHG21i0uCSJUsabWIfsyTCoaGhru1spdmYkJgdp814PcgVavs1FnPHAAAAFAQGAACgIDAAAADFrChwlM3Xx3n2bNGJOPcV8xKkZm5CNq8V54h2797daHPnnXd2bR88eLDRJu6LCx1l+7I2ADAZ2Tx3nOePxYyk5hiaLfwW97WZU8/m4mNhoix3Ko6HbYrEZeJYnOU8xAJLWZs2OWoxVyHmJcxE3DEAAAAFgQEAACgIDAAAQEFgAAAAilmRfJgVFIpJMSdOnJiwTVaIIyaCnH/++Y02sfhFlhAYE3CyBJO4L1vRK0teAYB+i8l92dgTExKzgngxKS8rlBTH3izRMSYfZueKfcyOExMSs2TEeO1ZEmPsT5vvmDYrVGbJ9PH8g1xJMcMdAwAAUBAYAACAgsAAAAAUfV9EKSs0dDLZHFGbAhVtxIUyVqxY0WizaNGikz5HahY9ynIeYh5CXFRJauYmxCJN0vTPLfWCRZSAmWWQC9rFfXFMlaSlS5d2bWfFgrKcqyiOz1leRByfs0JJsU2WP9CL7PsvXmvWps13Xps2LKIEAAD6jsAAAAAUBAYAAKAgMAAAAMXAkw/j41kSSptEkHicLMEktsn6FotNZP2Jx87axMSQrMDRbEwsbIPkQ2Bmmarkw3HO1bUdVxzM2mRjYZtVGWPxpOy7YqYVC4qyhPs2r0/cN04bkg8BAEB/ERgAAICCwAAAABR9zzGYqE02vxLFPmW5AW36PdPmlgYpvs5ZDka/kGMAzCyDzDFIzt3YF/Oyei0oNMhxbTYixwAAAPQdgQEAACgIDAAAQEFgAAAAimalnj7KklB6SRY5XZMIe02qbON0fc0A9Eebojr9EosOtRn7sjaTXb23rTbnarMScJvrmg1jM3cMAABAQWAAAAAKAgMAAFAMvMDRTNevef82c2G9FHtq259BzmNR4AiYWU6Hsbhfes1nmDdvXl/OH/MQMn3MLaPAEQAA6C8CAwAAUBAYAACAgsAAAAAU/S5wtEfS1j4fc6AGWWCoTVLKLLBxujsAoGHWj8X90muydizKNAv0bSzu618lAACA2Y2pBAAAUBAYAACAgsAAAAAUBAYAAKAgMAAAAAWBAQAAKAgMAABAQWAAAAAKAgMAAFD8H8mqRj3g9b6FAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    image_noise = torch.rand((1, 1, 28, 28)).to(device)\n",
    "    mean = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "    \n",
    "    interpretor = ActivationMaximization(model, [], None, img_shape, image_noise, i,  lr, reg_term, mean, epochs, False, 0.999)\n",
    "    endpoint = interpretor.interpret()\n",
    "    \n",
    "    scores = model(endpoint.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "    \n",
    "    images.append(endpoint.reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.2. AM from random noise with L2 regularization\n",
    "\n",
    "In the second case, we start the optimization process from random noise with L2 regularization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_shape = (28, 28)\n",
    "lr = 5e-3\n",
    "reg_term = 1e-1\n",
    "epochs = 10000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAf4AAACYCAYAAAAFtYvWAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAATt0lEQVR4nO3dfYwd11nH8d+TxF7b6/WuY5sE12ubKq2UNqQVlBKgqEgpIhGNWvcFVFqShgbaIkQBCVoBIQFKqCpRISpU3kITkra0KUkgglaAaFVKU4oCoSFVCI6bxMn6ff2ya3vXb4c/Zja589yzd+bOzn3ZPd+PtJLPvTNnXnZ9zsw8z5xjIQQBAIA0XDToHQAAAP1Dxw8AQELo+AEASAgdPwAACaHjBwAgIXT8AAAkpJGO38z+xMxubXrZYWJmt5vZvTXXfbeZfbXD918ws5tiy5rZrJm9tM52AaSFtrh0XdpiVej4zexpMzttZjNmdszMvmZm7zOzF9YNIbwvhPC7VTbYuqyZ/YiZPVdhH77HzL6Sn/gDZvaBKtvKf3Hn8/VOmNmjZvbGKuv2Uwjh+hDC3Yt8tz6EsEeSzOwuM/tw3e2Y2aVm9oCZnTSzZ8zsp0qW/2Uz229mx83sL81spG5dAJZm0G1x3inOtvycMbPHKu47bXGLptpiMxsxszvzOmbM7L/M7Pqy7Ve9478hhDAmaYekj0j6oKQ7K667JGa2WdIXJf2ppE2SrpD0j11U8XAIYb2kCWX7/DkzuzSynUuWvrdD748lnZF0maR3SvqEmb0ytqCZ/ZikD0m6VtJOSS+V9Nt16gLQmIG1xXmnuH7hR9LXJN3XRRW0xS9qqi2+RNJeSa+XNC7pVmXndWfHrYcQOv5IelrSG9xnr5V0QdJVefkuSR9u+f7XJO2TNCXpFklB0hWty0oalXQ6r2c2/9ka2f4dku4p289F9v3dkr7aUh7N9+U1km6X9HlJ90o6ke/nVkl/J2la0m5JP9uy7sLyn5U0I+k/Jb2q5fsPSXoq/+5bkna5/fg3SR+XdFzSE5Kubfn+y5JuWWSfg7KLnZ+TdFbZH8uspIck/aqkv3HH/HFJfxg5F6P5ui9v+eweSR9Z5Nx9WtIdLeVrJe2vUxc//PCz9J9Bt8VuuzslnZf0XRX3nba4eOyNtMWLLP9NSW/t9PuoFeMPIXxD0nOSfth/Z2bXSfoVSW/IT9LrF6njpKTrJU2FF68ipyKLXiNpOn+sddDMHjKz7d3uc34VeYuyX9T/5R+/Sdkf0ISkT0n6TH5cWyW9TdIdZnZtSzVvUnaFe6myX8aDZrYq/+4pZedjXNnV2L1m9p0t636/pD2SNku6TdL9savdxYQQ/izfx4/m5+oGZf9RrjOziZZj/Ellf0TeyyWdDyE82fLZf0ta7C79lfn3rcteZmabatQFoAf63Ba3ulHSv4YQvt3tPtMWN9oWF5jZZXn9j3c6hqUk900pO+neT0j6ZAjh8RDCKRUfD9exTdJNkj4gabukbyv7o6jqGjM7Jmm/pHcou/o7nn/3cAjhwRDCBWV/BK+T9MEQwlwI4VFJfyHpp1vqeiSE8PkQwllJH5O0RtmFiUII94UQpkIIF0IIn1X2B/3alnUPKrv6O5t//7+SfryL42gTQtgn6SuS3p5/dJ2kwyGERyKLr1d2hdvquKSxRar3yy/8e6xGXQB6p19tcasblT0x6AZtcabJtvgF+YXPpyTdHUJ4otP+LqXjf4myxzDeVmUxhwV7I8t047SkB0II/xFCmFP2x/uDZjZecf2vhxAmQgibQwjXhBD+eZF92yppOoQw0/LZM8qOs235/A904YpUZnZjnrByLP/jvkrZH/CC50P+HKal7q0Vj6GTuyW9K//3uxS/wpSyq+sN7rMNyh6HVVl+4d8zNeoC0Dv9aoslSWb2OkmXK7tD7wZtcabJtliSlCd43qMshPALZTtaq+M3s+9T9kuIvRaxT9ld+oLJDlVVmRrwm265hX9bhXXLtNY7JelSM2u9itou6fmW8gvHkp/obZKmzGyHpD9XdsI3hRAmJP2P28eXmFlreXu+zbr7u+BBSVeb2VWS3qjsii/mSUmXmNnLWj57lRZ/JPR4/n3rsgdCCEdq1AWgB/rcFi+4SdL9IYTZLtYpQ1tcry1Wfix3KksUfGv+FKSjrjp+M9uQv4Lx15LuDSHEXuX4nKSbzexKM1sn6bc6VHlA0qaSu/dPStplZq/OH2Xcqizh4li+T182s9u7OY6YEMJeZVmqv29ma8zsaknvUfGX971m9pY8fvNLkuYlfV0vJqocyvfpZmVXma2+Q9IvmtkqM3u7pCsl/UOXu3lAWUZn637PKbvy/rSkb4QQnl3k+E5Kul/S75jZqJn9kLI42WJXpX8l6T1m9goz2yjpN5U/2qtRF4AGDagtlpmtVfY4+67Id7TFfW6Lc5/Ij+GGEMLpKjtfteN/yMxmlD1e+Q1lMZWbYwuGEL4g6Y8kfUlZNubD+VfzkWWfUBav35M/lml73BJC+BdJvy7p75XFZq6Q1PrO46SyLM0mvENZtuqUpAck3RZC+KeW7/9WWcLGUWXxprfkcaJvSfoDZcd6QNJ3R/bp3yW9TNJhSb8n6W0LV2xduFPSK/Jz9WDL53fn2yzreH9e0lpl5/Ezkt4fQnhcksxsu2Xv2G6XpBDCFyV9VNnv8Zn857YqdQHomYG1xbk3K4sxfynyHW1xn9vi/AnHeyW9WtJ+e3GMhXd22rgVQx3NM7MrlT1qGQkhnGu47m2S7gsh/ECT9S43+R/IE5IuDyGcGPT+ABg+tMW9t1za4p50/Ga2S9kd+qiyK6ALIYQ3N74hLMS3PiZpQwjhZwa9PwCGB21x/yyntrhXk/S8V1mM5Sllgzy8v0fbSZqZjSob8OJHVXwMDwASbXFfLLe2uOeP+gEAwPBgWl4AABJCxw8AQEIanQXJzIgbrFAhhCYGTALQY7TDK9rhEMKWpVaSwvSHfXPRRcUHKBcuXCiUi4NFVUMOBoBe821T6u2Ob8u9OuenoXP6TBOV8KgfAICEJHnHH7vzbuJqzNfht+OvImP7ce5cd+NqVHmKkPrVO4AX1Xny2MR2llM75J/WXnzxxR3LMefPny+Uy/qHfp4f7vgBAEgIHT8AAAmh4wcAICHLLsZflm0ptcdnqqzTRHzFx318DGfVqlWF8qlTp7qu08eNYrEmfyx+HQBoVdZG+jYktrxvd5ZTTL+MPxZ//LHz4fuhsjr7iTt+AAASQscPAEBC6PgBAEgIHT8AAAnpa3JfE0l2se99vT6Jzq9TlnRRxZo1a9o+84l2Y2NjhfLs7GyhvG7durY6fMKf39fVq1cXyrHBOPxnvo6VlHQDoLOygcRiy5S1kb4dktoTAM+ePVt1F4eePx/+WGMDrw1zUjV3/AAAJISOHwCAhNDxAwCQkJ7G+OtMQuCXGRkZKZTPnDnT9X74Oubn59uW6TbuH4vprF+/vlD2eQCnT58ulKvEwPz58HkEsVwDf46W82QZAJamysA6ZXlSVSYYu+SSYnfi26q5ubnynV0mfPvfr4mPmsIdPwAACaHjBwAgIXT8AAAkpKcx/iqx5LLYUSweX2bt2rUd6/Sx9jpiMX5/LD7+7mP6VY7Nvy/r66jyHr+P1/l9H+b3TQE0K9Yul7XVvk2JtTujo6OFso/5nzhxolCOjQUwMzNTKPu2qokxWHohdv788ft9H+SxcMcPAEBC6PgBAEgIHT8AAAnp63v8sTGifWzEL1MnDuK324sYth8bQJImJiYK5SNHjhTKdcYg8DF9f75iuQb+/dmy74nxA+mo0g77/CO/jo9fS+25Uz7Hyc9dEpurZHJyslD27/77+U6OHTvWVodfJ9ZGLpXPZ4gpm//An+N+tsPc8QMAkBA6fgAAEkLHDwBAQuj4AQBISF+T+6okL/ikuSqJGX4dn3jiB7DxCSJ1xJL7/HZ9Ml+dJJM6E+qUnWeS+YB01fn/7xPTYonK/jM/GM+GDRtKt+OX2bx5c6Hskwz9xGiSdPz48ULZJ0j7djh2LD7x0Ccujo+PF8pVBiPySYeDbIe54wcAICF0/AAAJISOHwCAhPQ0xt+LSQhiA0f4mI2PC/nYyuHDh5e8Hz5vQJKOHj1aKPtYEwCkwucn+fYw1j7u37+/UPa5VD4HwE/IJpVP7LNjx45C+eTJk211lE2w5vc9NpCQP/4mJodrCnf8AAAkhI4fAICE0PEDAJCQnsb46/AxHR8n8bEWqf09TB9vKZu0pgr/vmjsPf5Y3B8AUI2Px/u4eJ04uc/58nkBW7ZsaVtnenq6Y52+T4nlCcT6qm75nLamJhzijh8AgITQ8QMAkBA6fgAAEtLXsfpj4877MZF9nNzHSfw7+TFl72DW4Y9l+/btbcvs3r17ydsBgCb58e1j7XCdOUGWiyNHjhTKPo9gbGysbR3fZxw8eLBjHb0Ys6aX9XLHDwBAQuj4AQBICB0/AAAJoeMHACAhPU3u84McxAYf8Ml858+fL5R9Ul2MH+RgcnKyUH7sscdK6/Bikz+U8Uk0ADBsYgOa+eQ+36b6drmpgWS82CRsS92uPzZ/LCdOnGhbx3/m1+mFKn1dU+ipAABICB0/AAAJoeMHACAhA5+kZ35+vlCuMilPmSeffLJQrhM7GR8fL5T9pDzPP/982zobN24slDds2FAo+/jVqVOn2urwsSU/gIM/ltjkEL2KvwFYfqoMzlM2wZifpMxPjCa1D8bm28yyfC6pfbAdv+91Bs7x2/Xt9OHDh9vW6UdM3+vnwErc8QMAkBA6fgAAEkLHDwBAQnoa468Sa/YxmtHR0UL56NGjXW/Hx9J97CkWv/Gx80OHDhXK/h19H7+XpKuvvrpQ3rp1a6Hs42SxPAEf95+amiqUn3322ULZx9Gk8vNeZfIkACuD//8dax98++bbSN+G+rLU3t7t2LGj43Zjbbvf1+np6UK5zqQ1ftI2f2xlYwesRNzxAwCQEDp+AAASQscPAEBCBh7c8O/px95L75aPA/nY+qZNm9rWee655wplH/epMnazfwfVx7R8ORav8ufDx978ONux9299noTfV7+NQbyzCmAwYuOaxN7LbzUzM1Mo+zZVas9P8jH8ubm5Qnn37t1tdZw+fbpQbiL/yLdvfj98DoDUfo58O+zLdcabGSTu+AEASAgdPwAACaHjBwAgIXT8AAAkZODJfT55wyev1eETM3wiyrFjx0rr8ANHeLE6Hn300ULZJ5H4Y9u8eXNbHatXry6UffKeTyKJDSTkk2x8UqE/P7HEHhL+gJWpTsKcb0N8sp/U3q745D6fEFelHa4zwZrfjq+jTjvsy7Ozs4VybKKfYZ4sjTt+AAASQscPAEBC6PgBAEjIwGP83vz8/JLr8APYlA2kI7UPHFFWZ2ygIR/T93zc/MCBA23LTE5OFsplAwf5WJPUfix+oCAf46oTRwOQDp8XEMsT8G1RWX5Sr/j2zpf9BEOxSXr8Mn6QH59HFavDn6Nhypvijh8AgITQ8QMAkBA6fgAAEjJ0Mf46/DuWe/bsKZR37dpVKE9MTLTVcejQoY7bGBsbK5RjMZ3jx493rGNkZKS0Dj/Rj38n1ceWYvH5sliSP1/DFHsCMPxibYZvq7yyiYBimpikp2xytP3797et48cg8GPB+LZ+mN/Zj+GOHwCAhNDxAwCQEDp+AAASYk3EUF6ozGzJlfl3zOvEn7ds2VIo+3fwfYxHkvbt21co13nn1Mfsffzdx/hjMTG/jn83thfx+FieQOS9XV72B5aBJtrhmtstlP2772VjpcQ00R/4+Uz8fh08eLDrOgfokRDCa5ZaCXf8AAAkhI4fAICE0PEDAJAQOn4AABIydAP41Ene8IPr+OSNmZmZQnnt2rVtdVx++eWF8vj4eKHsE0D85DlS+yAPPkHQJ6rEkl18Ul0sEbFpTSZ4Alh+fGJenTahykQ+3Woimdn3B3UGElppuOMHACAhdPwAACSEjh8AgIQ0HuNvjRX1K3bsY/i+7GM827Zta6tj586dhfLevXsL5Y0bNxbK69ata6vDT9xQNmlF7Pz4+BPxdwC91kQ74wcoayI/qYkBfPw6TErGHT8AAEmh4wcAICF0/AAAJGRFxPjL+Bj/008/3bbM0aNHC2U/sYP/3r+zL0knT54slH2M69y5c6X7CgDDzsfepfb2rs5EZ6tWrSqU/fgCVeLzfrI0X/YTn6WIO34AABJCxw8AQELo+AEASEjjMf5+jC1fxseF/Lj6sX308XkfB/J1HjlypK0OH9MalhwHAOnybVdMt21VLNbexPvxvo46bajP6fJi+QllmpjLYJhwxw8AQELo+AEASAgdPwAACaHjBwAgIY0n9w2jKgmHfnCd6enpjsvHJuBZ7gkfANLk2zPfZvrktljCYBOJ3b4NrdOm+sRsX8epU6dK6/CD/qy0iX244wcAICF0/AAAJISOHwCAhKzIGH8/Yu2xeJaPk/my368qcaOy2BsAdOLbnVh+klc26E+v2the1HvmzJkl19GL/RrkoEDc8QMAkBA6fgAAEkLHDwBAQlZkjH9QfMxmZGSkUPaTQ8RiOn6iHx+fWmmTRQDor9TzhHy7HDsf/Xhvf5BtN3f8AAAkhI4fAICE0PEDAJCQoYvxD0sM24/V7PfLj+0vtceK5ubmOm6j7F3ZWJ0AgOp83tSw9DGDxB0/AAAJoeMHACAhdPwAACSEjh8AgIQMXXLfoBItfMJH2QQ7PvlPak/E68cgEACA6lJM5vO44wcAICF0/AAAJISOHwCAhDQd4z8s6ZmG6+wLH/fxk+MkbsegdwBAZcu2HUapRtpiI9EBAIB08KgfAICE0PEDAJAQOn4AABJCxw8AQELo+AEASAgdPwAACaHjBwAgIXT8AAAkhI4fAICE/D9k0/59qbLx8wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    image_noise = torch.rand((1, 1, 28, 28)).to(device)\n",
    "    mean = torch.zeros((1, 1, 28, 28)).to(device)\n",
    "    \n",
    "    interpretor = ActivationMaximization(model, [], None, img_shape, image_noise, i,  lr, reg_term, mean, epochs, False, 0.999)\n",
    "    endpoint = interpretor.interpret()\n",
    "    \n",
    "    scores = model(end_point.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "    \n",
    "    images.append(endpoint.reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. AM from mean image of class\n",
    "\n",
    "In the third and final case, we start from the mean image class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_shape = (28, 28)\n",
    "lr = 6e-6\n",
    "reg_term = 8e-2\n",
    "epochs = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAACYCAYAAACfzCG6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAAXV0lEQVR4nO3deawdZ3nH8d+D43iJ9y3O9RLbcZxAXCANZVGLQAq0QUDZaWmBAk2BSgihtiktFSUVlFaopQsCUQqCtGkoeygtVBQKRZSwVoQQiBMcx05ysX2927lekvjtHzMezvzm9Zlzbu65i+/3I1nJc+bMOzPnXD/39bzPvG+klAQAACBJj5rsEwAAAFMHHQMAAFChYwAAACp0DAAAQIWOAQAAqNAxAAAAlXHpGETE+yPireP93qkkIq6PiBvHuO+rIuLrXbZ/ISJ+K/feiDgWEZvGclwAMwu5uHVfcnEPWjsGEXFPRByPiKMRcSgivhERr4+Iat+U0utTSm/v5YCd742Ip0fEfS3Hvy4iflgef0dEXNfLccp9XxURD5df6JGI+H5EPKfX/SdKSulZKaUbzrJtQUrpbkmKiI9ExDvGepyIeHRE/HdEHI6In0TEC2z7teXrxyLiPyNi6BG0dXVE3BERoxHxlYi4eKznDWBK5OI3RcTdZS4djoi/iYjzejx3cnGHCc7FL42IH5c/Nz+KiOe3nV+vdwyem1JaKOliSX8p6c2SPtTjvo9USHqlpKWSrpH0hoj49T72vyWltEDSEhXn/PGIWNY4SI8/4NNVeX2flfTvkpZJeq2kGyNiS7n9aZLeKel55fYdkj46xrZWSPq0pLeW278r6WODujZgBpnMXPw5ST+fUlokaaukx0l6Yx/7k4s14bl4jaQbJf2epEWSrpN0U0Ss6nqSKaWufyTdI+kZ9toTJZ2WtLWMPyLpHR3b/1DSTyUNS7pWUpK0ufO9ki6QdLxs51j5Z6iH8/l7Se9pe1/53ldJ+npHfEF5Lk+QdL2kT5Yf2pHyPIck/ZukA5J+Iul3OvY98/6PSToq6f8kPa5j+x9J2l5u+5GkF9h5/K+k90g6LOkOSVd3bP+qpGvPcs5J0ubyC39Q0qnys/pc+SV/yq75PZL+NvNZbC33i47Xvijp7eX//5Wk93ZsGyqPfckY2nqtpG/Y535c0uW9fG/84Q9/mn+mUi6WtFzSlyS9r8dzJxf/7PWJzMVPkrTX9hmR9JRu39eYagxSSt+WdJ+kp/q2iLhGRe/kGeWH+LSztPGApGdJGk7FLZoFKaXhbseNiCiPeXu/51z2rK5V8SHeVb78PBU/YEsk/YuKXtl9Kr6IF0t6Z0Rc3dHM8yR9QkXP7CZJN0fE7HLb9vLcFkv6MxW9tos69n2SpLslrZD0NkmfzvWWzyal9IHyHN9VflbPVfEX6ZqIWNJxjb8m6Z9zH8FZXtva8f9h29SxvZ+2rpB0a8e5P6Di87kisx+AMZroXBwRvxERRyTtU3HH4B/6PWdy8YTm4u9K+nFE/GpEzCqHEU5K+kH+6gqPpPhwWMWX4l4q6cMppdtTSqMqvpjxcr2Kc/5wH/s8OSIOSdot6WUqeo+Hy223pJRuTimdVvFD8kuS3pxSOpFS+r6kD0p6RUdb30spfTKl9KCkd0uaK+nJkpRS+kRKaTildDql9DEVP/BP7Nh3r4re44Pl9m2Snt3HdTSklH4q6WuSXlK+dI2kfSml72Xefkd5DtdFxOyI+GUViWJ+uf3zkl4aEY+NiHmS/lRFL3X+GNpaoKI33umwpIVjuEwA3U1YLk4p3ZSKoYQtkt4vaU8fu5OLCxOWi1NKD0v6JxWdp5Plf19XdgbP6pF0DNaouM3jhiTd2xHfm3lP3yLiDSpqDZ6dUjrZx67fTCktSSmtSCk9OaX0pbOc25CkAymlox2v7VRxnY33lz/AZ3q0iohXlgU1h8of/q0qfsDPuD+V93E62j5rQUkfbpD08vL/X658D1XlX6Dnq/gLsFvS70v6eHkNSil9WUXv+VPlud2j4lZcoyCprS0V/xJYZLstKtsDML4mNBdLUkrpLhV3bt/Xx27kYk1sLo6IZ0h6l6SnSzpfRafhgxHx+G4XMqaOQUT8goovKffYx08lre2I13VpqqelHSPiNSrGja5OKXWtnO1T5/GHJS2LiM5/1a6XdH9HXF1LWQm8VtJwWXH/j5LeIGl5SmmJpB+qfptnTTkU0tl216GTlvM942ZJj42IrZKeo+IWV37nlH6QUnpaSml5SulXJG2S9O2O7e9NKV2aUlql4ofyvPI6+m3rdhW3GSVJEXGBpEs0hiEgAGc30bnYnKfi7/V4IBcPJhc/XtLXUkrfLe+gfEfSt1QML51VXx2DiFhUPmLyr5JuTCndlnnbxyW9unyEYr6K2yBns0fS8ohY3OWYv6miQvOZqXxUxLZ/NSKu7+c6clJK90r6hqS/iIi5EfFYSb+t+pd7VUS8sBw/epOKWzPf1M8KaUbKc3q1muNBqyS9sbzd8xJJj1Zxy6gfe1R86Z3nfULF2NxNkr6dUtp1tp3LW1NzI2J+RPyBpItUFCCpfH1rFNZL+oCkv0spHey3LUmfkbQ1Il4UEXNV/Az8IKV0R5/XCyBjknLxtWeq2SPiMZL+WNKXO7aTi6deLv6OpKeeuUMQEVeqqL8YlxqDz0XEURW3b/5ExZjOq3NvTCl9QcWTA19RUU16S7mpcfu//EXxUUl3l7d9crdz3qGiAvY7UTzTeSwi3t+xfZ2KKtPx8DJJG1T0Hj8j6W0ppf/q2P5ZFQUlB1WMd72wHKf6kaS/VnGteyT9XOacviXpUhVFO38u6cUppf19nt+HJD2m/Kxu7nj9hvKY2VtXHV6h4l8ReyVdraKzdeZ7maviB/qYit7mLSoeN5QkRcRbIuILvbSVUhqR9KLyOg+qKPbp5xFTAHmTmYt/UdJtEfGAil+kn5f0lo7t5OKpl4v/R+VTHOXPzackvTOl9MVuJxf1oZbxFxGPVnELZE5K6aFxbnutpE+klJ4ynu1ON2Wv8g5Jq1NKRyb7fABMPeTiwTtXcvFAOgZRzLz0Hypu69wg6XRK6fnjfiCcGV97t6RFKaXXTPb5AJg6yMUT51zKxYNaROl1KsZ4tkt6WNLvDug4M1pZ1HdE0jNVVLECQCdy8QQ413LxwIcSAADA9MGyywAAoELHAAAAVAa6ilVEME5xjkop5eboBjDFkIfPaftSSivHu9GBL29Zn2Cqrpf6hm7799ruWNqYKvqtAZmIa6UuBQCmhJ2DaJShBAAAUBn4HYNOY/mXZtu//ntpczz+hTse/xIfjzsZbdc/lmudzndUAADjizsGAACgQscAAABU6BgAAIDKwGsMuo1558a2J6Li3Y+bO49e3tPPdql5badPn+66fVDHBQDgbLhjAAAAKnQMAABAhY4BAACo0DEAAACVCZ0SeTwm4+nW/hmPelS9vzN79uxaPHfu3Fo8b968RhsXXHBBLZ4/f37Xfc47r/lR+vWdPHmyFh87dqwWHz16tNHGAw88UItPnDhRix988MFa7AWNOW0Fi2MpggSAfnhO8byde62tKLyXgva2OJdD+21juuOOAQAAqNAxAAAAFToGAACgMqGLKI1F2xhSbmzfx/8XL15ci1etWlWLh4aGGm2sX7++Fq9bt65rG16TIDXHqo4cOVKLd+3aVYu3b9/eaGPnzvqqmsPDw7X4wIEDtXh0dLTRxkMPPVSLz7XxMAATy/PwrFmzGu/x2i7Py54zPU9L0pIlS2rxggULarHXi+VqDDz/eW3XwYMHa7HnaalZ/+W1X6dOnarFDz/8cKON6ZR3uWMAAAAqdAwAAECFjgEAAKhM+RoD52NZPr+AJC1fvrwWr127thZv2rSpFl966aWNNvy1iy++uBavXr26FvvYl9Qc72qrMbjtttsabSxatKgW+7idyz2D63UHbc/gMmcBgE4+n8CcOXNqca7GyusDPGdu2LChFm/cuLHRhr/H68GWLVtWi3O1DsePH6/Fe/furcVe23XnnXc22vD33H///bV4ZGSkFnsdgzS9ar24YwAAACp0DAAAQIWOAQAAqEz5GgMf7/Yx9tzY1ooVK2qxj0v5nARr1qxpbcOfwfXnVH0Ng9y5+rjcypUra3FujM3rEjz252tzY1t+bj7W5Z8xayUAM5v/fT///PNrsdd2eV2X1MyzmzdvrsWXXXZZLb7kkksabXh9mNcULFy4sOt55niO9N8PuWvx3zOe2/33ga9hk3sPNQYAAGBaoGMAAAAqdAwAAECFjgEAAKhMu+JDXzTJiwJzr3lBihd9+IIYkrRnz55a7Att+IQfuYk1/Lg+CZIXsOQKZ5YuXVqLvfjGJ0DyRUVy59qGQkNgZvOc4XnX80yuCNxf8/zmxXieY6VmEd99991Xi9sWapKaOdKLwP3afGImqblonv9+8NgXt5OmV17ljgEAAKjQMQAAABU6BgAAoDKhNQbjMcbSy3i5T+DjNQS+iMbJkycbbfgiGb44kR8jV2PgY1sXXXRR19jHuqTmmFlb/USuDf/cp9NYF4DJ5znD8+GpU6ca+xw6dKgWe+72ydpy9VHOj+t5d/HixY19fAI7n3jJayFyEw95XvXjnms5ljsGAACgQscAAABU6BgAAIDKpM5jMJbFenyMKbdYhS8s1Lbdn2vN8TE0rzHw52klafXq1bXYx6V8jgKf50BqH9tyg1qYYyov+AFgsDzfeV3W4cOHG/t4bvZ5Cjxn9jIu7/nPc6YvuiQ186z/DvHj+vwKkjQ6OlqLfSGm48eP12L/vKTplUO5YwAAACp0DAAAQIWOAQAAqExojUEvYyz+Ho997MbHfqTmGJLPY+DP/ufmRmibt8D5nAWStHDhwq5t9rLegvPz8HG86T62BWDyteXdXsbUPTd7fvM4t1aMz9uyfPnyWuxrGGzevLnRxpYtW2rx0NBQLfb6Ma8fkJpz34yMjNRin5MhN6/DdMrD3DEAAAAVOgYAAKBCxwAAAFToGAAAgMqkLqI0lmLEXooPvfCj3wUwejk3nxQpV3w4f/78WuyTcXgbuWN6caFfr8e9FL20fQ/TqUgGwPjzHOCT/rRtl5q5yyc08sWLvFhbkjZs2FCLL7/88lq8devWrtul5mJ1fu47duyoxV5oKEnDw8O1eN++fbXYCxZzE+9Np7zKHQMAAFChYwAAACp0DAAAQGVSF1HqRb9jXbn3eOwTC+UmOPK6BJ98w2sKLrzwwkYbPra1bNmyrm3mxqV84gyPBzG21baYyXQaKwPQv7a/4z5ZW47XcnlNlefMK664otHGVVddVYuvvPLKWnzZZZd1bVNq5nevIfCc6RPiSc0JnXwRqV5+L00n3DEAAAAVOgYAAKBCxwAAAFQmtcagl/kDxjL3gY8peb2Aj3XNnTu30YbPObBkyZJa7PUDmzZtarSxcePGWrxixYqu53Xo0KFGGwcOHOga+3hYbjET/wzH8pkCmLnGkiO8xsDnKfAc6osdSc26A3/PypUra3GuXszrA3yuF59fwWvBJGnp0qW12OdgOHz4cNdjSL3VZUwV3DEAAAAVOgYAAKBCxwAAAFQmtcZgLONWPobk9QJScyzL5xzw8aLly5c32vCxK38+du3atbV4zZo1jTZ8H187wecgOHr0aKON/fv312KvKfDnZ3N1G21rQ7BWAoB+tNUtSc280zY3TC7v+Lwt99xzTy3es2dP67l63ZUfx88zl8u9hqyt1quXGoOpXHPAHQMAAFChYwAAACp0DAAAQIWOAQAAqExo8WEvE+t4caHv4wV8uckoVq9eXYt9Ig0vHBwaGmq04QUoHnubixcvbrThRS1eXOiTYuSKD0dHR2uxFxv65Bzz5s1rPQ9fNMS/h9wkSQBmrrZiw1zxoRfXnThxohZ74eC2bdsabXiO9ImF/Li5CY58AjufaM5/X3jxuiStX7++Fu/evbsWj4yM1GIvLJeaeZfiQwAAMC3QMQAAABU6BgAAoDKhNQZtCyTl+ARGXlPgCxVJzckoNmzYUIt9vMjrBaT2OgVfVMkn75Cai3f4GJuP5Xv9QK5dH//yNr3mQGrWKfh5+feSOw8mPQJmDh+r99jrlnJj+54zPA/5OH2uxuquu+7q67x80iSpOaGd/37wa8nVnPkkef57yGvMvOZAal/wbirlWO4YAACACh0DAABQoWMAAAAqA68x6Bw38fGgXuYxWLBgQS32sX8fL5KkLVu21GKvKVi1alUt9nqB3Gv+LKyPS+WeSfXnVnMLa3TKzUHgCzydPHmyFvuYmi/uITUXIvHPuO28pKk1/gVg/PSyAJLXLnntU67GoC3fe01B7tn/tnP18/Q8LTXrsDy3+9i/522pfQEoP26u1svPdSrjjgEAAKjQMQAAABU6BgAAoDLwGoPOMaFexqnbntv3+oDcM6f+mo/T+zOpuXUOcs/DdvJx+dw4fdtzq36M3LoP3q6PXfnc4bk6Bd/H2zx06FAtzn1P/X6PAKYvz8OeV3oZU3deh+V5KLdGi8+p4m34uH0ub/u5ed2aX1uuFsCP63Fb7YM0vfImdwwAAECFjgEAAKjQMQAAABU6BgAAoDKhiyj1om3CCi+2yxUOehHfhRdeWItXrFhRi70YRWovPvRJM3ILgPjEQv6etmJEqXktbQtP+QRIUntxYW5yJgAzQ25yorbiQ58kyPNyrg0vJPQc6nHuNc9VnjM9t0vNSfA89t8PuUJKPw9fiM7zbm6SJL/+qVyMyB0DAABQoWMAAAAqdAwAAEBlQhdRcrnx8rYx9F7GaXyMyMfDfBwqN7bvx/F6gZGRkVq8d+/eRhsHDx6sxW3jZblr97Eqj31sa3R0tNGGT7Tk++QmFgEwc3ndwZw5c2qx13blxvb9Pd6G5zLPU1K+7qCT14etXbu28Z7NmzfXYq8x8En0vCZLkvbv39819lyfuxb/nTKVcccAAABU6BgAAIAKHQMAAFCZ1HkMcvUBPg7jYzU+lrN79+5GG6tXr67FS5curcX+fG3uOV6fc2DPnj21eHh4uOt2STp8+HAt9rF8P27uPHwczmsd/Li5z2Pfvn21+NixY13PC8DMkcvDXv/kednroXxhOklat25dLfa8PH/+/K5t5s7Dc7cfN1fr4IvoeRueH++9995GG9u3b6/FO3furMVec5Cr9aLGAAAATEt0DAAAQIWOAQAAqExojYGPIeXGtnxM3cfpd+3aVYtz4/L+7KvXA/Sy/oCPw/s4lM9bkHv21efTbpu3oJcaA782rznwzyv3mtdtUGMAzFy95OG2tWE8X+b4vAYbN26sxStXrmzs42sw+No5Xi+QW/fFz9XrA26//fZafOuttzba8Pd4HYLXvp06darRxlReG8FxxwAAAFToGAAAgAodAwAAUKFjAAAAKgMvPuy2KFJum08C4YUjXtSRK/rzySh8oQ0vYMkV/flxPPZinFyxSS8LPrVt99e8UNDj3Hn4axQbAjgjl3c8R3hx4axZs2qxFwHm3uOFgR7n8lLbgne+IFxuMbsdO3bU4m3bttXiO++8sxZ7caLUnEjOi779PKbTZEY53DEAAAAVOgYAAKBCxwAAAFRikJMuRESyuJd9um738829319ra3M8zitnLDUFbe8ZjzqFsei8/pSSUkr9fyAAJpzn4R73qcVeLzB79uxa7AsiSdLChQtr8ZIlS2qxL27k26XmBEdeD+Zj+z72LzUXODpw4EAtbpsALnccr4eYxMmLvpdSesJ4N8odAwAAUKFjAAAAKnQMAABAZVJrDHo5dts+gxj7n+l6qfOgxgCYHsZSY9Cv3FwwbXUKvk8vbbTJLaLUNn9CWyxN6d8Z1BgAAIDBomMAAAAqdAwAAEBlyq2V0O9z+zltdQi9jFu1tTFZ9RFtcufV77n30gYAnJEbl3fTff2AmYQ7BgAAoELHAAAAVOgYAACACh0DAABQGXjxYTfjMWnEeBQn5ozHwksTUWzoBrFQFQBg5uCOAQAAqNAxAAAAFToGAACgMugag30ppZ0DPkarQdUhTIU2J4qd+8WTdR4A+rZP0qTnYQzEQHLxQFdXBAAA0wtDCQAAoELHAAAAVOgYAACACh0DAABQoWMAAAAqdAwAAECFjgEAAKjQMQAAABU6BgAAoPL/glm9TiGOIwgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAACYCAYAAACfzCG6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAAY/klEQVR4nO3deaxcZ3nH8d+TxGsSx7vjPYtpmspJU1GWVqUgsaukLGVVC4E2LJVaVLUsFVWBlhZapLJFSIhCITSlJQkQoBDUFkqBEnYhwpISnNixfb1d29dLFju2n/5xzj2Z85z3zntmPHPvXPv7kazknXPmPWfm3nnmved5zvuauwsAAECSzpnpEwAAAKODgQEAAKgwMAAAABUGBgAAoMLAAAAAVBgYAACAykAGBmb2ATP7y0HvO0rM7K1mdlOfz325mX29y/bbzey61L5mdtTMLuvnuADOLsTi7HOJxS1kBwZmttXMHjSzI2Y2YWbfMLPXmFn1XHd/jbu/rc0BO/c1syeZ2Y42zzOzuWZ2V9v9y+e83MxOlj/Qw2b2AzN7VtvnTxd3f6a73zjFtgvc/R5JMrOPmtnf9HscM7vSzL5sZofM7Odm9tyw/fry8aNm9kUzW9Oiz0eZ2UOdH1Qzu8TMvOxn8t+sC0DAKJnpWFx+IT8cPtetviiJxXUtYvELzeyn5c/6J2b2nC59HQ3/TprZDR3be47rba8YXOvuF0raKOnvJL1R0odbPndQXi9pbx/Pu8PdL5C0WMU532xmS+NOZnbe6Z3eaCtf32ck/bukpZJeJekmM/uFcvsTJb1d0rPL7fdK+tcWXb9f0nem2La4/DBd0DZYAehqpmPxJzo+09UXZUvEYrWKxWsl3STpTyUtUvHd93EzW5nqr/PnIWmVpAcl3VL21Vdc7ymV4O6H3P2zkl4k6Toz21wevDZ6MrM3mNkuMxsrRytuZps69zWz8yXdLmlNx0gnOZIxs0sl/Z6kd/RyvuHcT0n6J0kLJF1Wjn5vNbObzOywpJeb2Roz+6yZHShHWK8M3cw3s0+Uo7jvm9kvd5zjn5vZlo4R3nPDc83MbihHiHeZ2ZM7NnzFzK6f4rW7mW0ys1dJ+l1Jbyjfq8+Z2evN7JNh/xvM7D2Jrn5R0hpJ73b3k+7+ZUn/K+ml5fZrJd3i7j929+OS3ibpN83s8qneUzN7saQJSV+aah8AgzdTsXhA504s7h6L10macPfbvfB5SfdLmjIWd3i+ij+gv1a2e47rUp81Bu7+bUk7JD0hbjOzZ6gY6TxF0iZJT5yij/slPVPSWMeIZ2yKQ94g6U0qRkJ9KUdp10s6Kunu8uFnS7pVxQj2X1SMpHao+KE9X9LbO39pyv1vUTHy+rik28xsTrlti4r34yJJf6ViBLi647mPk3SPpOWS3iLpU6nR8lTc/YPlOb6zfK+uVTGqfIaZLe54jS+S9M+pt2CKxzZ3/L+FberYXn+i2SJJfy3pz7qc9jYz22FmHzGz5V32A9CHGYjF15Zf1j82sz/s55yJxdlY/F1JPzWz3zazc61IIxyT9MMWp3edpI/5I2sd9BTXJ51O8eGYih9K9EJJHylHKA+o+MH0rRztnefun+6zi8eb2YSk3ZJeIum57n6o3HaHu99WjmCXS/oNSW9094fc/QeSPqRHRnGS9D13v9XdH5b0LknzJT1ektz9Fncfc/dT7v4JFb/wj+147l5J73H3h8vt/yfpt/p8TSqPuUvSVyW9oHzoGZLG3f17id3vKs/h9WY2x8yepiJQLCy3f0HSC83sajNbIOnNkrxje/Q2SR929+2JbeOSHqPicuejJV2o4oMEYPCmJRZLulnSlZJWSHqlpDeb2Ut6eD6xuNA1Frv7SUkfUzHgOVb+99XlAG5KZrah7KezRqLXuC7p9AYGayUdSDy+RlLnl0Xqi6OV8hLXOyX9cb99SPqmuy929+Xu/nh3/68pzm2NpAPufqTjsW0qXmdj//IXeHJEKzN7mRUFNRPlL/9mFb/gk3Z2jOIm+x7E5bobVaRZVP43NUJV+QF6jooPwG4Vf+nfXL4GufuXVIyeP1me21ZJRya3dzKza1T8FfLuKY511N2/6+4n3H2PpD+S9LTyKgOAwRp6LJYkd/9J+YV70t2/Iem9Kv6ab4tYrHwsNrOnqPjee5KkuSq+7D9Uxt1uXibp6+5+b8exWsf1Tn0NDMzsMSp+SKnbPnapyJFMWt+lq9zSjo+SdImkr5nZbkmfkrTazHab2SWtT7jd8cckLTWzCzse2yBpZ0e7ei1WVAKvkzRmZhsl/aOKL8Bl7r5Y0o9Uv4Sz1sw62xvKY/Z7vpNuk3R1mWN8lrr8Ze7uP3T3J7r7Mnd/uqTLJH27Y/v73f1R7r5SxS/SeeXriJ6k4udyX/lzeZ2k3zGz72fOO3UJDUCfpjEWT/WcQX2micWPxOJrJH21/OPqlLt/R9K3VPwx1s3LVL9aMHmstnG90tPAwMwWWXGLyb9Jusnd70zsdrOkV1hxO8ZCFZcuprJH0jIzu2iK7T9S8QtwTfnv+vI516gcMZbFIm/t5XWklJfEvyHpHWY238yulvQHqv9wH21mzyvzR3+i4jLPNyWdr+IXZV95Tq9QM4ezUtJry0tHL1BxSe4LPZ7mHhW/QJ3n/ZCK3NzHJX3b3e+b6snl5aT5ZrbQzF4nabWkj5bb5pvZZitskPRBSe9194OJrj6oohDmmvLfByR9XtLTy74eZ2ZXmNk5ZrZM0vskfaXjsiGA0zADsVhm9mwzW1LGiMdKeq2K6vrJ7cTiAcRiFXd5PWHyCoGZ/YqKmokpawzM7NdVDBBvCY/3EtcrbQcGnzOzIyq+jP9CRU7nFakd3f12FV8E/y3p55LuKDcdS+x7l4oik3vKyz5rwvYT7r578p+Ky2WnyvbJcrf1Kio6B+ElKv4SHpP0aUlvcff/7Nj+GRUFJQdV5LueV+apfiLpH8rXukfSVYlz+paKKyDjkv5W0vPdfX+P5/dhSb9Uvle3dTx+Y3nM5KWrDi9V8VfEXklPlvRUd5/8ucxX8Qt9VMXI9Q5J1dwDZvYmM7tdktz9gfBzOSrpIXffV+5+maQvqrhk9SMVP/tecpEA0mYkFpdeXPZzREUO/O+9fs8/sXgAsdjd/0fSWyXdWv6sPynp7e7+H1I9Fne4TtKnQvpFysT1qVg91TJ4Znalii+Hee5+YsB9r1NxK8avDbLf2aYcCd4l6WJ3PzzT5wNg9BCLh+9MicVDGRhYcSfB51Vc1rlRxV/5zxn4gTCZX3uXpEXu/vszfT4ARgexePqcSbF4WIsovVpFjmeLpJOS+rrfFd2Vd20clvRUFZWnANCJWDwNzrRYPPRUAgAAmD1YdhkAAFQYGAAAgMpQV7EyM/IUZyh3Z7IiYBYgDp/Rxt19xaA7HfrylvUJpgbf36lTp067j36Oc8459YstbWo14j799JGTem29vke594e6FGB26faZTm2Ln/HYHkQMbePcc8/teh5RP8cYhBjLB6Hld8q2gR9YpBIAAECHoV8xOF1x1DmIv1Zzf7mnxJFrlBpBnzx5MrHnI+LoNtVHr69/pkbMAM4cubjTTwzNHaNNH7krFbk43UYubqfkrihLw3kPh2V0zgQAAMw4BgYAAKDCwAAAAFRmtMagTTVsP30M4zk5w6h9SD2Wy/2l8lS5uoNhvB8ARlM/dxjkpGJMr/n+VPybidiUOu9+6g5yhlE/NyhcMQAAABUGBgAAoMLAAAAAVBgYAACAyrQWHw6ikGQYxXbDOK82/cZzH9Z59PocihEB9KKfONzPhD6xzzZTJuemno99pM6712ni20x4F89rGAWO/eKKAQAAqDAwAAAAFQYGAACgMq01Bm0Wmsjl/9tMmpHro01uP7dPm1x+fL1z5szpeozU+5E7j3iMVJ7q+PHjtXY89xMnTnQ9Ruo5AGaPzs90mxqiXj/vbfbPxbIYH1OP5drz5s3L9nHeed2/9lKvJcbIhx9+uNaOcTfGXEk6duxY13ZukaWpHhsGrhgAAIAKAwMAAFBhYAAAACpDrzHolhNJ3S+au7c1Vy8gNXNK8Tlz586ttVM5p1w9QC7HlDpuzH8tXLiw1p4/f36jj3iu8T174IEHau1UbuvIkSO19v33319rx/xZP/fxAhhdvX5+c3MQtOk/1oPl4t+iRYsafSxbtqzWXrp0aa29ePHiWvuiiy5q9LFgwYJaO/f9kIrlDz30UK0dY2iMsQcPHmz0sX///lp73759XbfHGoTpxBUDAABQYWAAAAAqDAwAAEBlWucx6EeupiB172vMy8f2hRdeWGvHXFeq39x9qjHXn3L++efX2jF/FvNlUjNPF3NdMbc1MTHR6COee8xdpeoSAJwd2twbn5sbpk0cjnF3xYoVtfb69esbfWzcuLHW3rBhQ629evXqWnvJkiXZ88itnZDK7R89erTWjjUE4+PjtfbevXsbfezcubPWjrE9xuFY+zXVY8PAFQMAAFBhYAAAACoMDAAAQIWBAQAAqIxc8WGcWCMWaMR2anKiOFFQnDgjFr2kJsWIxTRxQqNDhw7V2qmJlmKhSCy+WblyZdfzkpqvLxYXxsKZNkWQuWKj1PbO18eCSsDsFT+/bSYzis+JcTgV/2IcjsXVa9asqbU3bdrU6OOKK66otdetW1drL1++vNZOLaIU5QrJ28Ty+NpiAXssNJearz/G8jgRU/yOkfpbzK8fXDEAAAAVBgYAAKDCwAAAAFRGrsYgijmUmP9J5YNinilOenHxxRfX2nFhDqmZQ4u5+5iXixNgSM1zj3mpeNw4WUeqj5jrOnDgQNftUnPCjphji8dILYhCXQEwe3V+fvv5LMd42GZ7rsYg1lil4l+M1bGP+FoOHz7c6CPG7th+8MEHu26XmosmxXjfZhGluE+sW4uvJfVzosYAAABMOwYGAACgwsAAAABUhl5j0JkTaZMPibmqXN4lNY9BvIc0V2MQ74VNHScuihG3x3yR1Mzlx5xbnLdg1apVjT5i/mvfvn21dqwfSOXH4sJLqXPt1Ca3BWB2yi2IJOXnOsjNLyM1a73ifDEx/sWaAym9KFKnmMvfs2dPY5+4wFGcHyDWD6RiaIzDscYgxtjYTvUbzyNXxyZNXxzmigEAAKgwMAAAABUGBgAAoDL0GoPOXFWb/EgulxXnLZg7d26jj7gmQczdr127ttaO98ZKzfth4/wAcZ7r1P2zce7reF6x1iG1VkLMmcWagnhvbGo+hficNvfLAjg7pD7/ublNYhxuU+sV42yctyDGZak510uMd3Eelx07djT6GBsbq7X3799fa8dcfyqGxvx/rCGI6y2k6rjiY/E5sY4h9X3ZZl2LQeCKAQAAqDAwAAAAFQYGAACgwsAAAABUpnWCo1hM0WYCh9iOxYixwE9qTooRi1pie86cOY0+4sQZcZKM2I6FI1JzQo9YBLlu3bpaOxYnStLOnTtr7VjkGAtnUueRWzQpYjIj4MzSrQg8tRBdjBlR7CNVfBhjcywkjPFv48aNjT5iDM2dVz9Ff3GCo1hYLuUnI4rHSJ1nbtKo+BwmOAIAACOBgQEAAKgwMAAAAJWh1xh0GsTkDDGXdcEFFzT2iRMHbdiwodZes2ZNrR3zR1Izlx8nxYh5qlStQ6wp2LRpU60dc2yphTdyi3fE8xjEe8wiSsCZK8aIVI1BbjG7NhMcxZiYW8xu/fr12T7iZG1xUrjU5ETxtcRJ8VI1ZlGMszFWx/qAVK1DfN/7mWhuuiaj44oBAACoMDAAAAAVBgYAAKAyrfMY9JMfyS2aFPNWUjN3H2sMli1bVmun7v2POfU4x0DMh8UFQyTp6quvrrWvuuqqWjvm2O67775GH3HxpqhNri8+lruflkWVgDNLt890qi4p5uVz88mk4k6MTfE58+bNy/YRxZqDWGOQmj8gPifWpcW5EhYuXNjoIxdD28xBMJviLFcMAABAhYEBAACoMDAAAACVaZ3HIOac2txzH58T81KxXkCSVq5cWWvHPFTMKaVqDOKcAzEvFXP/ixYtavRx5ZVX1tqbN2+utWMua/fu3Y0+4uuNtQzxvOIc3qlzbZMPA3D2ijEh1hjE/HiqFiquQRDngok1Vancfox/Mb4dOXKk1k7Fsvnz59faMf7H75jUnC1xXoI4j0F8/an6gdx7mlsHYjpxxQAAAFQYGAAAgAoDAwAAUGFgAAAAKtNafNhGLMiI7ViMEttSczKKXHFdapKkWCgYixGjOAGSJK1du7bWjkWRuYU4pOa55yYWiRNApfaJxTb9FIUCOHPlYkIstkstADcxMVFrb9mypdaOixdt27at0UeMZ/E8cgV9UjP+xRgaixNT3wex6PHgwYO1dlyILxZeSunFqjq1KT7MFYEOClcMAABAhYEBAACoMDAAAACVaa0x6GdCo1xOJZXLGRsbq7XjJEAHDhyotWOuK3WcKOa+UrUOsd9jx47V2nv37q21d+7c2egjvpaY24qTM8VjSL1PnJHK043ygh8A+hdz7lIzZuTqklIxI+blt2/fXmvHyYlinJaa9QGxnZu8SGrWDCxevLjWjpPGxT6l5uRL8Vzjc1LfKfE7I8bq1M8hmq76L64YAACACgMDAABQYWAAAAAqMzqPQSqnEnPZMacS81I7duzI9pGrOWhz73/MGS1durTWvvTSSxt9xIU34uvdunVrrX3nnXc2+oj3/sbXOz4+Xmunai5iLmuUFusAMHpirIrxL+bcU4vIxTqEONdBatG4XB/xPOL8MatWrcr2EWsK2uT243dEPI9YYxa/P1LnkarLGBVcMQAAABUGBgAAoMLAAAAAVIZeY9Dt/vfUtph3ifnwo0eP1tqpe//jHN3xHtPc3NlSM2cU74+95JJLau1Uvuj48eO1dny99957b6199913N/rIzWMQ7xWOx5Sa72E8j9ge5dwXgNPTJqceY0CMmbFOK64DIzVjaJxzJdZDpdZbSMWzTm3WSoj1AfH7IJ5nrA1LHSeKx02tizCb1qDhigEAAKgwMAAAABUGBgAAoMLAAAAAVKZ1gqN+JniIBRuxMOTw4cON58RJkHILgKQWvIgTZ8RzjUV/sSgwJRbS7Nq1q9aOizuljhP7iIWFqQKXXLFhbn+JgkTgbBZj5oIFC2rt1atXN54TJ4GLsTsWice4LTXjXSyCjBMrpYogly9fXmvHRZRiMWKq+PDEiRO1doy7ue1SM67mnjOTMZcrBgAAoMLAAAAAVBgYAACAyrTWGORy2220yeXkxDxVajKKNnUInVKLF8V8f1zM6NChQ7V2aoKP3OREgxBzWaljDOO4AKZfjCltJjzKTc4T8/SStG7dulo7ThIXc/mpGBprrOK5x4mWli1b1ugj1h3E58Q+U3VrMTbHc42TN6XqFHqN5TNZ68UVAwAAUGFgAAAAKgwMAABAZVoXUUrl8rvtL+Xz321y3/G4MacWF9mQpIULF9baMYcWjxsXd5Ka9+DGGoNc/iwl9x6mclC9voe5PBb1BsCZIxV3YozM1QOk5iCIcWTVqlW1dqwHiLVfqXOL7Vj7lap1iPE9xuV9+/bV2qk5acbHx2vt/fv319q5mgOp9zjLPAYAAGAkMDAAAAAVBgYAAKAyrfMYRG3u08zlZVL318Z9Yl4+5rLmzZvX6CPmpWKfsV4glQ+KuazYjvfGxjka2si9Vqn5HsX8YXyP27ynAM5cMSbE3H6sj9q+fXujjyVLltTacV2DuJbCihUrGn3ENWty88mk5g+I9Q+xPmDbtm219j333NPoI+4T6xLinDTx+0HKzwURMY8BAAAYCQwMAABAhYEBAACoMDAAAACVkSs+jNoUG+ae02s7dZzUAkedUsUmvU4SkuojFiTmJidKvT+9TqyRKmBkUiPg7BHjSPz8xwl8du3a1egjF/9iwd7ll1/e6GP16tW1dlwAKZ5n7FOSxsbGau0tW7bU2j/72c9q7a1btzb62Lt3b9fjtCkk7zWGMsERAAAYCQwMAABAhYEBAACozGiNQT+57F7z5VIzDxXzP6kFLyYmJmrtmEOKkySlJt6I5xaPG/tMnUecSCTuE/tsU4MR92HyIuDskZtErs0+cbK2GC+lZk3BgQMHau04kVBcVEmSFi9eXGvHRZJiLDt8+HCjj3jcuCBSPPdY+yXlawhyNRlS8/suThoVt89kXRdXDAAAQIWBAQAAqDAwAAAAFRtmHsPMvDNXNV05k9wiQTGXE/eX8osT5dopMQ8Vc0yxnXpOm3kLojY5xV64u9ydwgRgFjAz7xafUjEk7p+rQ2rTR4yzsU4rFYdTj3XKLfaUEusD2tRp5eaPaSP3nrbpM3Ee33P3X+35ZDK4YgAAACoMDAAAQIWBAQAAqAx9HoOZuBczl4dvk6fvJ6eWO49R0c+8BaP6WgCcnlweX8rPH5OKDzHfn5tPJtVHm3PrdgwpH+/a1CUMQu47Y5Tmk+GKAQAAqDAwAAAAFQYGAACgwsAAAABUhl58OBMTHA3iOL1OaDGIwpFhvT+DnuAIwNklF99S23Nxp00cyhWOz6ZYlns/RilOc8UAAABUGBgAAIAKAwMAAFAZdo3BuLtvG/IxGnK5mWHkqUY519VmMqYebRx0hwCGZvzUqVMjF4enq49RkXstfcbpocTioa6uCAAAZhdSCQAAoMLAAAAAVBgYAACACgMDAABQYWAAAAAqDAwAAECFgQEAAKgwMAAAABUGBgAAoPL/NsOEIsekxVgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "    mean = torch.zeros((1, 1, 28, 28)).to(device)\n",
    "    \n",
    "    interpretor = ActivationMaximization(model, [], None, img_shape, img, i,  lr, reg_term, mean, epochs, threshold=0.999)\n",
    "    end_point = interpretor.interpret()\n",
    "    end_point = (end_point - end_point.min()) / (end_point.max() - end_point.min())\n",
    "    \n",
    "    scores = model(end_point.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "\n",
    "    images.append(end_point.reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# III. Activation Maximization in Codespace\n",
    "\n",
    "Activation Maximization in Codespace is an extension to the Activation Maximization idea by introducing a generative model which generates the input to be fed into the main model. Our implementation is based once again on the paper \"Methods for Interpreting and Understanding Deep Neural Networks\n",
    "\" by Montavon et al, which can be accessed [here.](https://arxiv.org/pdf/1706.07979.pdf)\n",
    "\n",
    "![Activation Maximization in Codespace Formula](./assets/activation_maximization_codespace_formula.png)\n",
    "\n",
    " We consider two different starting points:\n",
    "\n",
    "1. Simple GAN\n",
    "2. Pretrained DCGAN (source: https://github.com/csinva/gan-vae-pretrained-pytorch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Simple GAN\n",
    "\n",
    "In the first case, we consider a simple and small GAN that we have trained ourselves."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "classifier_name = 'model_cnn.pt'\n",
    "discriminator_name = 'mnist_discriminator.pt'\n",
    "generator_name = 'mnist_generator.pt'\n",
    "\n",
    "lr = 8e-2\n",
    "epochs = 10000\n",
    "reg_term = 1e-2\n",
    "latent_dim = 100\n",
    "img_shape = (1, 28, 28)\n",
    "classes = [str(i) for i in range(10)]\n",
    "\n",
    "train_classifier = False\n",
    "train_gan = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "classifier = ModelRepository.get_general_mnist_cnn(classifier_name)\n",
    "\n",
    "if train_classifier:\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    optimizer = optim.Adam(classifier.parameters(), lr=0.0005)\n",
    "\n",
    "    classifier.train()\n",
    "    ModelTrainer.train(classifier, criterion, optimizer, mnist_data_loader.trainloader, 1)\n",
    "    \n",
    "    ModelRepository.save(classifier, model_name)\n",
    "\n",
    "generator = ModelRepository.get_mnist_generator(latent_dim, generator_name)\n",
    "generator.eval()\n",
    "\n",
    "if train_gan:\n",
    "    discriminator = ModelRepository.get_mnist_discriminator(discriminator_name)\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    \n",
    "    generator.train()\n",
    "    discriminator.train()\n",
    "    ModelTrainer.train_mnist_gan(generator, discriminator, mnist_data_loader.trainloader, 0.002, latent_dim, epochs=20)\n",
    "    generator.eval()\n",
    "    discriminator.eval()\n",
    "    \n",
    "    ModelRepository.save(generator, generator_name)\n",
    "    ModelRepository.save(discriminator, discriminator_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 28, 28)\n",
    "    mean = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "\n",
    "    interpretor = ActivationMaximizationCodespace(classifier, classes, None, img_shape, i,  lr, reg_term, epochs, generator, latent_dim, mean, threshold=0.99)\n",
    "    end_point = interpretor.interpret()\n",
    "    end_point = (end_point - end_point.min()) / (end_point.max() - end_point.min())\n",
    "    \n",
    "    scores = model(end_point.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "    \n",
    "    images.append(end_point.detach().cpu().numpy().reshape(28, 28))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Pretrained DCGAN\n",
    "\n",
    "In the second and final case, we are using a pretrained DCGAN model that we have found online. The source of this model is the following repository: https://github.com/csinva/gan-vae-pretrained-pytorch."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PretrainedDCGANGenerator(\n",
       "  (main): Sequential(\n",
       "    (0): ConvTranspose2d(100, 512, kernel_size=(4, 4), stride=(1, 1), bias=False)\n",
       "    (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (2): ReLU(inplace=True)\n",
       "    (3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "    (4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (5): ReLU(inplace=True)\n",
       "    (6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "    (7): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (8): ReLU(inplace=True)\n",
       "    (9): ConvTranspose2d(128, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "    (10): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (11): ReLU(inplace=True)\n",
       "    (12): ConvTranspose2d(64, 1, kernel_size=(1, 1), stride=(1, 1), padding=(2, 2), bias=False)\n",
       "    (13): Tanh()\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "classifier_name = 'model_cnn.pt'\n",
    "classifier = ModelRepository.get_general_mnist_cnn(classifier_name)\n",
    "\n",
    "dcgan_generator = ModelRepository.get_pretrained_dcgan_generator()\n",
    "dcgan_generator.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAACXCAYAAABumpNvAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy86wFpkAAAACXBIWXMAAAsTAAALEwEAmpwYAAASKElEQVR4nO3deZBVZXrH8d+DIgZQKHAZccEFtDQEuxwELJIyFEjUODgzpROnXKIiGeMymYojDJMSl0GlrGgSlZqUSKElioooahALTAIGxY24jE5pBlAHQXCwbaRxZJEnf5zD63nPub3c7r7dt7u/n6pbxe+ec9/z3l7efnjPe88xdxcAAIAk9ejoDgAAgOpBYQAAAAIKAwAAEFAYAACAgMIAAAAEFAYAACBok8LAzP7dzG5o632riZndZGbzWvjaS81sZSPbl5jZ35ba18zqzezYlhwXQPfCWNzkaxmLm6HJwsDMPjKzP5rZNjOrM7OXzexKMwuvdfcr3f1XzTlgdl8z+0sz+6SJ4/dKf4A3m1mtmT1rZoc351jpN/ab9Bv6pZm9ZWbnNOe17cndz3L3BxvY1tfd10mSmT1gZjNaehwzO9HM/svMtprZGjP7QW77Fenz9Wb2vJkNamlbmf1uNDM3s/Et7TeAqhiL+5vZg2b2Wfq4qYy+MxZntOdYXE5bezV3xuB77n6ApMGSZkqaKmlOM1/bWv8g6TRJwyUNklQn6Z4yXr/K3ftK6q+kz4+b2YD8Tma2b6t7WsXS9/e0pP+QNEDS30maZ2bHp9tPl3SbpHPT7R9Kmt+StjL7HSfpPEmfVuAtAd1RR47F/yKpt6SjJY2UdLGZXVbG6xmL1b5jcTltRdy90YekjySNzz03UtIeScPS/ICkGZntU5T8Mdgo6QpJLmlIdl9JfST9MW2nPn0MKnH8X0u6I5P/WtIHTfU73fdSSSszuU/alxGSbpL0hKR5kr5M+zlI0jOSaiWtkTQ589q9+z8maZuk/5V0cmb7LyStTbf9VtIPcv14SUlBs1XS+5LGZbYvl3RFA312SUPSb/guSTvTr9Wzkq6XtDD3nu+R9K8lvhbD0tdZ5rmlkn6V/vufJc3KbBuUHvu4ctvKPLdE0tmlfoZ48OBR3qMKxuItkk7N5F9K+p9m9p2x+Nvn220sLqet7KNFawzc/TVJn0j6i/w2MztT0j9KGp9+EU9voI3tks6StNGTKZq+7r6xxK5zJI0xs0Fm1lvShUr+4JQlrayuUPJF/F369LlKfsD6S3pYSSX1iZIv3nmSbjOzcZlmzpW0QEnl9YikRWbWM922VsnXo5+km5VUbYdlXjtK0jpJB0m6UdKTparlhrj7fWkf70i/Vt9T8ot0ppn1z7zHv5H0UKkvQQPPDcv823LblNleTlsys/Ml7XT35xp6TwBap53HYqk4RpQaHxrFWNyuY3E5bQWtWXy4Uck3Je9Hkua6+3vu/pWSb0xr/J+k30vaoKSaPFHSLWW8frSZ1UnaJOnHSqrHrem2Ve6+yN33KPkh+XNJU939a3d/S9L9ki7OtLXa3Z9w912S7pK0v6TRkuTuC9x9o7vvcffHlPzAj8y89jMl1eOudPsHSmY/WszdP5X0oqTz06fOlLTF3VeX2P39tA/Xm1lPM5ugZKDonW5/TtKPzGy4mf2JpOlKKsve5bZlZn2VTF/9rDXvD0CztNdY/LykX5jZAWY2RNLlKj0+NISxONFuY3GZbQWtKQwOVzLNkzdI0vpMXl9in3L8Wsk3faCS6acnVd6MwSvu3t/dD3L30e7+QgN9GySp1t23ZZ77WMn7LOyf/gDvrWhlZpekC2rq0h/+YUp+wPfa4OlcTqbtJheBNMODki5K/32RSleoSn+Bvq/kF2CTpOskPZ6+B7n7fyqpnhemfftIyVRcYUFSU20pGYAecvcPW/fWADRDe43FP1VyyuF3Ss5r7/1ffXMxFqt9x+Jy2spqUWFgZqcq+SaV+tjHp5KOyOQjG2mqObd2PFnSA+5e6+47lJy3GWlmBzXxuubIHn+jpAFmdkDmuaOUzFTsFd5LuhL4CEkbzWywpNmSrpE00N37S3pX8RTO4WaWzUelx2xpf/daJGm4mQ2TdI6SKa7SL3Z/x91Pd/eB7v5Xko6V9Fpm+yx3H+ruhyj5Qdo3fR/ltjVO0k/NbJOZbVLydXvczKaW+X4BNKI9x+J0DL7Q3b/j7n+q5O/Ha029rpkYiyszFpfV1l5lFQZmdmD6EZNHJc1z99+U2O1xSZelH6HorWTqoiGbJQ00s36N7PO6pEvMrF96DukqJefCtqR9Wl7Ox2Ya4u7rJb0s6XYz29/MhkuapPib+10z+2F6/uhnknZIekXfLqT5Q9qny1Q8h3OIkj+WPdPz7ycqmeYpx2Yl3/Rsv79Wcm7uEUmvufvvG3pxOp20v5n1NrOfSzpMyQIkpc8Ps8RRku6T9G/u/kW5bSkpDIZJqkkfGyX9RNKsMt8vgBI6Yiw2s+PMbKCZ7WNmZylZhDcjs52xuMrG4nLb2qu5hcGzZrZNyfTNPyk5p1PyYyruvkTS3ZL+W8lq0lXpph0l9n1fyXTUunTap9R0zs8lfa1k+uoPSla5Zz+neaSSVaZt4cdKPoqzUdJTkm5092WZ7U8rWVDyhZLzXT9Mz1P9VtKdSt7rZkl/VqJPr0oaqmRl762SznP3z8vs3xxJJ6Vfq0WZ5x9Mj1ly6irjYiX/i/hMyR/vM9JZGCk5XfOIkgVBr6XvJVz8xMx+aWZLmtOWu3/u7pv2PiR9I+kLd68v8/0CiHXkWPxdSb9RMhV9u6QL3f29zHbG4iobi5tqqyEWn2ppe2Z2opJpi17uvruN2z5C0gJ3P60t2+1s0krwfUnfcfcvO7o/AKoPY3HldZWxuCKFgSVXXlqsZFrnQUl73P37bX4g7D2/dpekA9398o7uD4DqwVjcfrrSWFypmyj9RMm0/1ol08h/X6HjdGtm1kfJRzjPULLyFACyGIvbQVcbiyt+KgEAAHQe3HYZAAAEFAYAACCo6F2szIzzFF3TFnc/uKM7AaB5GIu7Jncvda+EVmPGAC3xcUd3AABQGRQGAAAgqOiphLaw775xF3fvbtPrcgAAgAxmDAAAQEBhAAAAAgoDAAAQVP0aA9YUAEDrmX37yTaueIvGMGMAAAACCgMAABBQGAAAgIDCAAAABFW/+BAdL7toSWLhEtAZ8XuL5mLGAAAABBQGAAAgoDAAAAABawxa6Nprr43yuHHjonzSSSdFub6+vtDG6NGjo7xz58426l3b4twkAHQfzBgAAICAwgAAAAQUBgAAIGCNgYqf0x8/fnxhn6uvvjrKEyZMiPL27dujPGDAgCj36FGswY466qgor1mzpunOAgBQQcwYAACAgMIAAAAEFAYAACDolmsMBg8eHOV77rknyvn1A1JxDcEjjzwS5euvv77RY86ZM6fwHGsKAADVhhkDAAAQUBgAAICAwgAAAATdYo3B8ccfH+WlS5dGeeDAgVFevnx5oY2LLrooylu2bCmrD3V1dWXtDwAoGjJkSJQfffTRKB966KFRrq2tjXK/fv0Kbb799ttRXrVqVZRnzpxZdj87M2YMAABAQGEAAAACCgMAABBQGAAAgKBbLD48++yzo3zkkUdG+fPPP4/yjTfeWGij3MWG48aNi/K2bdvKej0AdHcTJ04sPDdr1qwoH3LIIVHu2bNnlJtzQ7v8cc4666wo5294d9VVVzXQ466BGQMAABBQGAAAgIDCAAAABObulWvcrHKNl+Hwww+P8sKFC6M8atSoKH/66aeFNi6//PIor1y5Mspjx46N8jXXXBPlc845p9Dmrl27Guhx1Vvt7iM6uhMAmqdaxuI8M4vy9OnTo1xqvdfOnTuj/Nhjj0V58eLFUX755Zej/NVXXxXafOKJJ6KcH883bdoU5Zqamihv3ry50GZ7cHdreq/yMWMAAAACCgMAABBQGAAAgKBbXMdgw4YNUc6f/8/fVOmwww4rtLFkyZIor127Nsp9+/aN8oIFC6LcidcTAEBFTJkyJcr5NQXr1q0rvGbSpElRXrFiRVnH3Hff4p+9MWPGNPqaJ598MsodtaagvTBjAAAAAgoDAAAQUBgAAICgW1zHoCn5a2m//vrrhX2OOeaYKOc/f7t169Yojx8/PspvvPFGa7pYbbiOAdCJVMtYfP7550d5/vz5Uf7kk0+iPHny5EIby5Yta1UfLr300sJzc+fOjfLu3bujfOyxx0Z5/fr1repDS2X/7rg71zEAAACVR2EAAAACCgMAABBQGAAAgKBbXOCoKbW1tVE+4YQTCvs89NBDUb7ggguinL/AURdffAgAZbvkkksa3Z6/wFFLFhrmF4bPmjUryldeeWWTbezZsyfK27dvL7sflVDJDwtkMWMAAAACCgMAABBQGAAAgIA1BiXkL24hSfPmzYtyfo3BPvvsE+VbbrklyvnzXpJ01113RXnHjh1l9RMAqlWpMa+mpqbR1+RvTldKr169onzqqadGefr06VE+44wzmmwzr76+PsrdbWxmxgAAAAQUBgAAIKAwAAAAAWsMmil/849du3ZFeeTIkVFeunRplG+77bZCm/vtt1+Ub7755tZ0EQCqRqnP3G/YsCHKRxxxRJQXLlwY5bq6ukIb+Rva5dd35Y/71FNPRXnQoEGFNvM3STrllFOiXC3XMWgvzBgAAICAwgAAAAQUBgAAILBKXnvZzNrnws5t7NZbby08N3Xq1CjnP9d6wAEHRDm/5uD5558vtNmvX78oT5s2Lcp33HFHlPPX7+5Aq919REd3AkDzmJn36PHt/wM7aiw588wzo3zfffdFOb/moJRt27ZFef78+VHOj9/5tVxvvfVWoc1169ZF+eSTT26yH9XA3YsXi2gDzBgAAICAwgAAAAQUBgAAIOA6BiXkz0lJxc/K5j9vmz9n98orr0R59OjRhTbfe++9KN9+++2NtplfcwAAzVUNa5Tya62GDh0a5VGjRkU5vw5LklauXBnlL774otFjLlu2LMp9+/Yt7DN79uxG2+humDEAAAABhQEAAAgoDAAAQEBhAAAAAi5wVMKrr75aeK6mpibK1113XZTvvffeso8zbNiwKL/zzjtR3rRpU5RHjIivKbRx48ayj9lGuMAR0Il01rG4JfILvVesWBHlUjdEOvjgg6P8zTfftH3HKoALHAEAgIqjMAAAAAGFAQAACLjAkaRevXpF+Zhjjinss2XLlig/88wzrT7uu+++G+X8hTgmTJgQ5WuvvTbK+ZsuAUB3k705lCSddtppUc5fsO6ll14qtNFZ1hS0F2YMAABAQGEAAAACCgMAABCwxkDS7t27o9yzZ8/CPvnzVJW4hkD+3Fd+jUF9fX2bHxMAOrP8De5uuummKG/dujXKkydPrnSXKsbs28sWVPIaRMwYAACAgMIAAAAEFAYAACBgjYGKn2Fdvnx5YZ/x48dHuU+fPlHOn8dqiaFDhza6/eOPP271MQCgKxk1alSUDzzwwCg//fTTUV67dm3F+1QplVxXkMWMAQAACCgMAABAQGEAAAAC1hiUsHjx4sJzEydOjHL+vgUzZswo+zgjRoyI8rnnntvo/jt37iz7GADQlU2aNKnR7fk1BmgaMwYAACCgMAAAAAGFAQAACCgMAABAYJW8YIKZld149iYRUvtd0CFryJAhhedWrVoV5Xw/x4wZ0+j2sWPHFtqcOXNmlPMX5qitrY1yTU1NlNevX19os52sdvcRTe8GoBo0NRb36FH8P+KePXsq1p/WGDhwYJRXr14d5bq6uijnx82uxN2t6b3Kx4wBAAAIKAwAAEBAYQAAAIKqu8BRR6wpyFuzZk3huSlTpkT57rvvjvKLL74Y5fx5sFLn8PLvddGiRVGeNm1alDtwTQGALqxa1xOUcsMNN0R58ODBUZ49e3Z7dqdLYsYAAAAEFAYAACCgMAAAAEHVrTGoVnPnzo3yhx9+GOU777wzyr17947yCy+8UGjzueeeizLnxgBUSnadU2daU5B39NFHR3nHjh1Rfvjhh9uxN10TMwYAACCgMAAAAAGFAQAACKruXgnoFLhXAtCJdNaxOL9WS5I++OCDKL/55ptRnjhxYkX7VE24VwIAAKg4CgMAABBQGAAAgIDCAAAABFzgCAC6AbNv16lVw83qmmP48OGF5w499NAo33///e3VnW6DGQMAABBQGAAAgIDCAAAABFzgCC3BBY6ATsTMvDOuMUDjuMARAACoOAoDAAAQUBgAAICg0tcx2CLp4wofA+1vcEd3AEBZtrg7Y3HXUrFxuKKLDwEAQOfCqQQAABBQGAAAgIDCAAAABBQGAAAgoDAAAAABhQEAAAgoDAAAQEBhAAAAAgoDAAAQ/D/sQhegf9DrOQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = 5e-2\n",
    "epochs = 10000\n",
    "reg_term = 1e-2\n",
    "latent_dim = 100\n",
    "img_shape = (1, 28, 28)\n",
    "classes = [str(i) for i in range(10)]\n",
    "\n",
    "images = []\n",
    "probabilities = []\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 28, 28)\n",
    "    mean = mnist_data_loader.generate_mean_image_for_class(i).expand(1, 1, 28, 28).to(device)\n",
    "\n",
    "    interpretor = ActivationMaximizationCodespace(classifier, classes, None, img_shape, i,  lr, reg_term, epochs, dcgan_generator, latent_dim, mean, threshold=0.99, verbose=False, is_dcgan=True)\n",
    "    end_point = interpretor.interpret()\n",
    "    end_point = (end_point - end_point.min()) / (end_point.max() - end_point.min())\n",
    "    \n",
    "    scores = model(end_point.to(device)).to(device)\n",
    "    prob = torch.nn.functional.softmax(scores, 1)[0][i] * 100\n",
    "    probabilities.append(prob)\n",
    "    \n",
    "    images.append(end_point.detach().cpu().numpy().reshape(28, 28))\n",
    "\n",
    "MnistVisualizer.display_images_with_probabilities_for_each_class(images, probabilities)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}