examples/cf_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Counterfactual instances on MNIST"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Given a test instance $X$, this method can generate counterfactual instances $X^\\prime$ given a desired counterfactual class $t$ which can either be a class specified upfront or any other class that is different from the predicted class of $X$.\n",
    "\n",
    "The loss function for finding counterfactuals is the following: \n",
    "\n",
    "$$L(X^\\prime\\vert X) = (f_t(X^\\prime) - p_t)^2 + \\lambda L_1(X^\\prime, X).$$\n",
    "\n",
    "The first loss term, guides the search towards instances $X^\\prime$ for which the predicted class probability $f_t(X^\\prime)$ is close to a pre-specified target class probability $p_t$ (typically $p_t=1$). The second loss term ensures that the counterfactuals are close in the feature space to the original test instance.\n",
    "\n",
    "In this notebook we illustrate the usage of the basic counterfactual algorithm on the MNIST dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "from keras import backend as K\n",
    "from keras.layers import Conv2D, Dense, Dropout, Flatten, MaxPooling2D, Input\n",
    "from keras.models import Model, load_model\n",
    "from keras.utils import to_categorical\n",
    "import matplotlib\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "import tensorflow as tf\n",
    "from time import time\n",
    "from alibi.explainers import CounterFactual"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load and prepare MNIST data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape: (60000, 28, 28) y_train shape: (60000,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAADYNJREFUeJzt3X+oXPWZx/HPZ20CYouaFLMXYzc16rIqauUqiy2LSzW6S0wMWE3wjyy77O0fFbYYfxGECEuwLNvu7l+BFC9NtLVpuDHGWjYtsmoWTPAqGk2TtkauaTbX3A0pNkGkJnn2j3uy3MY7ZyYzZ+bMzfN+QZiZ88w552HI555z5pw5X0eEAOTzJ3U3AKAehB9IivADSRF+ICnCDyRF+IGkCD+QFOEHkiL8QFKf6+XKbHM5IdBlEeFW3tfRlt/2nbZ/Zfs92491siwAveV2r+23fZ6kX0u6XdJBSa9LWhERvyyZhy0/0GW92PLfLOm9iHg/Iv4g6ceSlnawPAA91En4L5X02ymvDxbT/ojtIdujtkc7WBeAinXyhd90uxaf2a2PiPWS1kvs9gP9pJMt/0FJl015PV/Soc7aAdArnYT/dUlX2v6y7dmSlkvaVk1bALqt7d3+iDhh+wFJ2yWdJ2k4IvZU1hmArmr7VF9bK+OYH+i6nlzkA2DmIvxAUoQfSIrwA0kRfiApwg8kRfiBpAg/kBThB5Ii/EBShB9IivADSRF+IKme3rob7XnooYdK6+eff37D2nXXXVc67z333NNWT6etW7eutP7aa681rD399NMdrRudYcsPJEX4gaQIP5AU4QeSIvxAUoQfSIrwA0lx994+sGnTptJ6p+fi67R///6Gtdtuu6103gMHDlTdTgrcvRdAKcIPJEX4gaQIP5AU4QeSIvxAUoQfSKqj3/PbHpN0TNJJSSciYrCKps41dZ7H37dvX2l9+/btpfXLL7+8tH7XXXeV1hcuXNiwdv/995fO++STT5bW0Zkqbubx1xFxpILlAOghdvuBpDoNf0j6ue03bA9V0RCA3uh0t/+rEXHI9iWSfmF7X0S8OvUNxR8F/jAAfaajLX9EHCoeJyQ9J+nmad6zPiIG+TIQ6C9th9/2Bba/cPq5pEWS3q2qMQDd1clu/zxJz9k+vZwfRcR/VtIVgK5rO/wR8b6k6yvsZcYaHCw/olm2bFlHy9+zZ09pfcmSJQ1rR46Un4U9fvx4aX327Nml9Z07d5bWr7++8X+RuXPnls6L7uJUH5AU4QeSIvxAUoQfSIrwA0kRfiAphuiuwMDAQGm9uBaioWan8u64447S+vj4eGm9E6tWrSqtX3311W0v+8UXX2x7XnSOLT+QFOEHkiL8QFKEH0iK8ANJEX4gKcIPJMV5/gq88MILpfUrrriitH7s2LHS+tGjR8+6p6osX768tD5r1qwedYKqseUHkiL8QFKEH0iK8ANJEX4gKcIPJEX4gaQ4z98DH3zwQd0tNPTwww+X1q+66qqOlr9r1662aug+tvxAUoQfSIrwA0kRfiApwg8kRfiBpAg/kJQjovwN9rCkxZImIuLaYtocSZskLZA0JuneiPhd05XZ5StD5RYvXlxa37x5c2m92RDdExMTpfWy+wG88sorpfOiPRFRPlBEoZUt/w8k3XnGtMckvRQRV0p6qXgNYAZpGv6IeFXSmbeSWSppQ/F8g6S7K+4LQJe1e8w/LyLGJal4vKS6lgD0Qtev7bc9JGmo2+sBcHba3fIftj0gScVjw299ImJ9RAxGxGCb6wLQBe2Gf5uklcXzlZKer6YdAL3SNPy2n5X0mqQ/t33Q9j9I+o6k223/RtLtxWsAM0jTY/6IWNGg9PWKe0EXDA6WH201O4/fzKZNm0rrnMvvX1zhByRF+IGkCD+QFOEHkiL8QFKEH0iKW3efA7Zu3dqwtmjRoo6WvXHjxtL6448/3tHyUR+2/EBShB9IivADSRF+ICnCDyRF+IGkCD+QVNNbd1e6Mm7d3ZaBgYHS+ttvv92wNnfu3NJ5jxw5Ulq/5ZZbSuv79+8vraP3qrx1N4BzEOEHkiL8QFKEH0iK8ANJEX4gKcIPJMXv+WeAkZGR0nqzc/llnnnmmdI65/HPXWz5gaQIP5AU4QeSIvxAUoQfSIrwA0kRfiCppuf5bQ9LWixpIiKuLaY9IekfJf1v8bbVEfGzbjV5rluyZElp/cYbb2x72S+//HJpfc2aNW0vGzNbK1v+H0i6c5rp/xYRNxT/CD4wwzQNf0S8KuloD3oB0EOdHPM/YHu37WHbF1fWEYCeaDf86yQtlHSDpHFJ3230RttDtkdtj7a5LgBd0Fb4I+JwRJyMiFOSvi/p5pL3ro+IwYgYbLdJANVrK/y2p95Odpmkd6tpB0CvtHKq71lJt0r6ou2DktZIutX2DZJC0pikb3axRwBd0DT8EbFimslPdaGXc1az39uvXr26tD5r1qy21/3WW2+V1o8fP972sjGzcYUfkBThB5Ii/EBShB9IivADSRF+IClu3d0Dq1atKq3fdNNNHS1/69atDWv8ZBeNsOUHkiL8QFKEH0iK8ANJEX4gKcIPJEX4gaQcEb1bmd27lfWRTz75pLTeyU92JWn+/PkNa+Pj4x0tGzNPRLiV97HlB5Ii/EBShB9IivADSRF+ICnCDyRF+IGk+D3/OWDOnDkNa59++mkPO/msjz76qGGtWW/Nrn+48MIL2+pJki666KLS+oMPPtj2sltx8uTJhrVHH320dN6PP/64kh7Y8gNJEX4gKcIPJEX4gaQIP5AU4QeSIvxAUk3P89u+TNJGSX8q6ZSk9RHxH7bnSNokaYGkMUn3RsTvutcqGtm9e3fdLTS0efPmhrVm9xqYN29eaf2+++5rq6d+9+GHH5bW165dW8l6Wtnyn5C0KiL+QtJfSvqW7aslPSbppYi4UtJLxWsAM0TT8EfEeES8WTw/JmmvpEslLZW0oXjbBkl3d6tJANU7q2N+2wskfUXSLknzImJcmvwDIemSqpsD0D0tX9tv+/OSRiR9OyJ+b7d0mzDZHpI01F57ALqlpS2/7VmaDP4PI2JLMfmw7YGiPiBpYrp5I2J9RAxGxGAVDQOoRtPwe3IT/5SkvRHxvSmlbZJWFs9XSnq++vYAdEvTW3fb/pqkHZLe0eSpPklarcnj/p9I+pKkA5K+ERFHmywr5a27t2zZUlpfunRpjzrJ5cSJEw1rp06dalhrxbZt20rro6OjbS97x44dpfWdO3eW1lu9dXfTY/6I+G9JjRb29VZWAqD/cIUfkBThB5Ii/EBShB9IivADSRF+ICmG6O4DjzzySGm90yG8y1xzzTWl9W7+bHZ4eLi0PjY21tHyR0ZGGtb27dvX0bL7GUN0AyhF+IGkCD+QFOEHkiL8QFKEH0iK8ANJcZ4fOMdwnh9AKcIPJEX4gaQIP5AU4QeSIvxAUoQfSIrwA0kRfiApwg8kRfiBpAg/kBThB5Ii/EBShB9Iqmn4bV9m+79s77W9x/Y/FdOfsP0/tt8q/v1t99sFUJWmN/OwPSBpICLetP0FSW9IulvSvZKOR8S/trwybuYBdF2rN/P4XAsLGpc0Xjw/ZnuvpEs7aw9A3c7qmN/2AklfkbSrmPSA7d22h21f3GCeIdujtkc76hRApVq+h5/tz0t6RdLaiNhie56kI5JC0j9r8tDg75ssg91+oMta3e1vKfy2Z0n6qaTtEfG9aeoLJP00Iq5tshzCD3RZZTfwtG1JT0naOzX4xReBpy2T9O7ZNgmgPq182/81STskvSPpVDF5taQVkm7Q5G7/mKRvFl8Oli2LLT/QZZXu9leF8APdx337AZQi/EBShB9IivADSRF+ICnCDyRF+IGkCD+QFOEHkiL8QFKEH0iK8ANJEX4gKcIPJNX0Bp4VOyLpgymvv1hM60f92lu/9iXRW7uq7O3PWn1jT3/P/5mV26MRMVhbAyX6tbd+7Uuit3bV1Ru7/UBShB9Iqu7wr695/WX6tbd+7Uuit3bV0lutx/wA6lP3lh9ATWoJv+07bf/K9nu2H6ujh0Zsj9l+pxh5uNYhxoph0CZsvztl2hzbv7D9m+Jx2mHSauqtL0ZuLhlZutbPrt9GvO75br/t8yT9WtLtkg5Kel3Sioj4ZU8bacD2mKTBiKj9nLDtv5J0XNLG06Mh2f4XSUcj4jvFH86LI+LRPuntCZ3lyM1d6q3RyNJ/pxo/uypHvK5CHVv+myW9FxHvR8QfJP1Y0tIa+uh7EfGqpKNnTF4qaUPxfIMm//P0XIPe+kJEjEfEm8XzY5JOjyxd62dX0lct6gj/pZJ+O+X1QfXXkN8h6ee237A9VHcz05h3emSk4vGSmvs5U9ORm3vpjJGl++aza2fE66rVEf7pRhPpp1MOX42IGyX9jaRvFbu3aM06SQs1OYzbuKTv1tlMMbL0iKRvR8Tv6+xlqmn6quVzqyP8ByVdNuX1fEmHauhjWhFxqHickPScJg9T+snh04OkFo8TNffz/yLicEScjIhTkr6vGj+7YmTpEUk/jIgtxeTaP7vp+qrrc6sj/K9LutL2l23PlrRc0rYa+vgM2xcUX8TI9gWSFqn/Rh/eJmll8XylpOdr7OWP9MvIzY1GllbNn12/jXhdy0U+xamMf5d0nqThiFjb8yamYftyTW7tpclfPP6ozt5sPyvpVk3+6uuwpDWStkr6iaQvSTog6RsR0fMv3hr0dqvOcuTmLvXWaGTpXarxs6tyxOtK+uEKPyAnrvADkiL8QFKEH0iK8ANJEX4gKcIPJEX4gaQIP5DU/wG6SwYLYCwMKQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()\n",
    "print('x_train shape:', x_train.shape, 'y_train shape:', y_train.shape)\n",
    "plt.gray()\n",
    "plt.imshow(x_test[1]);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prepare data: scale, reshape and categorize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape: (60000, 28, 28, 1) x_test shape: (10000, 28, 28, 1)\n",
      "y_train shape: (60000, 10) y_test shape: (10000, 10)\n"
     ]
    }
   ],
   "source": [
    "x_train = x_train.astype('float32') / 255\n",
    "x_test = x_test.astype('float32') / 255\n",
    "x_train = np.reshape(x_train, x_train.shape + (1,))\n",
    "x_test = np.reshape(x_test, x_test.shape + (1,))\n",
    "print('x_train shape:', x_train.shape, 'x_test shape:', x_test.shape)\n",
    "y_train = to_categorical(y_train)\n",
    "y_test = to_categorical(y_test)\n",
    "print('y_train shape:', y_train.shape, 'y_test shape:', y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "xmin, xmax = -.5, .5\n",
    "x_train = ((x_train - x_train.min()) / (x_train.max() - x_train.min())) * (xmax - xmin) + xmin\n",
    "x_test = ((x_test - x_test.min()) / (x_test.max() - x_test.min())) * (xmax - xmin) + xmin"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define and train CNN model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cnn_model():\n",
    "    x_in = Input(shape=(28, 28, 1))\n",
    "    x = Conv2D(filters=64, kernel_size=2, padding='same', activation='relu')(x_in)\n",
    "    x = MaxPooling2D(pool_size=2)(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    \n",
    "    x = Conv2D(filters=32, kernel_size=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=2)(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    \n",
    "    x = Flatten()(x)\n",
    "    x = Dense(256, activation='relu')(x)\n",
    "    x = Dropout(0.5)(x)\n",
    "    x_out = Dense(10, activation='softmax')(x)\n",
    "    \n",
    "    cnn = Model(inputs=x_in, outputs=x_out)\n",
    "    cnn.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "    \n",
    "    return cnn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/janis/.conda/envs/py36dev/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From /home/janis/.conda/envs/py36dev/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 28, 28, 1)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 64)        320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 14, 14, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 14, 14, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 14, 14, 32)        8224      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 7, 7, 32)          0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 7, 7, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 1568)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 256)               401664    \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                2570      \n",
      "=================================================================\n",
      "Total params: 412,778\n",
      "Trainable params: 412,778\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "WARNING:tensorflow:From /home/janis/.conda/envs/py36dev/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Epoch 1/3\n",
      "60000/60000 [==============================] - 53s 875us/step - loss: 0.3319 - acc: 0.8940\n",
      "Epoch 2/3\n",
      "60000/60000 [==============================] - 43s 712us/step - loss: 0.1187 - acc: 0.9629\n",
      "Epoch 3/3\n",
      "60000/60000 [==============================] - 44s 733us/step - loss: 0.0930 - acc: 0.9719\n"
     ]
    }
   ],
   "source": [
    "cnn = cnn_model()\n",
    "cnn.summary()\n",
    "cnn.fit(x_train, y_train, batch_size=64, epochs=3, verbose=1)\n",
    "cnn.save('mnist_cnn.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluate the model on test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy:  0.987\n"
     ]
    }
   ],
   "source": [
    "score = cnn.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test accuracy: ', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generate counterfactuals"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Original instance:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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": [
    "X = x_test[0].reshape((1,) + x_test[0].shape)\n",
    "plt.imshow(X.reshape(28, 28));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Counterfactual parameters:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "shape = (1,) + x_train.shape[1:]\n",
    "target_proba = 1.0\n",
    "tol = 0.01 # want counterfactuals with p(class)>0.99\n",
    "target_class = 'other' # any class other than 7 will do\n",
    "max_iter = 1000\n",
    "lam_init = 1e-1\n",
    "max_lam_steps = 10\n",
    "learning_rate_init = 0.1\n",
    "feature_range = (x_train.min(),x_train.max())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run counterfactual:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/janis/.conda/envs/py36dev/lib/python3.6/site-packages/tensorflow/python/training/learning_rate_decay_v2.py:321: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Deprecated in favor of operator or tf.math.divide.\n",
      "Explanation took 6.444 sec\n"
     ]
    }
   ],
   "source": [
    "# set random seed\n",
    "np.random.seed(1)\n",
    "tf.set_random_seed(1)\n",
    "\n",
    "sess = K.get_session()\n",
    "\n",
    "# initialize explainer\n",
    "cf = CounterFactual(sess, cnn, shape=shape, target_proba=target_proba, tol=tol,\n",
    "                    target_class=target_class, max_iter=max_iter, lam_init=lam_init,\n",
    "                    max_lam_steps=max_lam_steps, learning_rate_init=learning_rate_init,\n",
    "                    feature_range=feature_range)\n",
    "\n",
    "start_time = time()\n",
    "explanation = cf.explain(X)\n",
    "print('Explanation took {:.3f} sec'.format(time() - start_time))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counterfactual prediction: 9 with probability 0.9900996088981628\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAADZFJREFUeJzt3WuIXPUZx/Hf020i5iIoQV1yadoqJSUv1rKYeCtKSI1aiH2hJGCJWF1fRGxBoRIEg1iQkl58IcqKoRHUttpLImibsBRjtF7WJcSYNG1M0u3WkChRTEBN3H36Yk9kG/f8ZzJzzpwZn+8HZGfOM2fO4+hvzzl7zn/+5u4CEM9Xqm4AQDUIPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoL7ayo2ZGbcTAiVzd6vndU3t+c1smZntMbO9ZnZPM+8FoLWs0Xv7zaxL0j8lLZU0IukNSSvdfVdiHfb8QMlasee/WNJed9/n7scl/VbS8ibeD0ALNRP+2ZL+M+H5SLbs/5hZn5kNmtlgE9sCULBm/uA32aHFFw7r3b1fUr/EYT/QTprZ849Imjvh+RxJ7zbXDoBWaSb8b0i60My+bmZTJa2QtKmYtgCUreHDfnf/zMzukPRXSV2S1rv724V1BqBUDV/qa2hjnPMDpWvJTT4AOhfhB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoAg/EBThB4Ii/EBQhB8IivADQTU8RbckmdkBSUcljUr6zN17i2gKQPmaCn/mKnd/v4D3AdBCHPYDQTUbfpe02czeNLO+IhoC0BrNHvZf5u7vmtm5kraY2T/cfevEF2S/FPjFALQZc/di3shsraRj7r4u8ZpiNgYgl7tbPa9r+LDfzKab2cyTjyV9T9LORt8PQGs1c9h/nqQ/mdnJ93nK3f9SSFcASlfYYX9dG+OwHyhd6Yf9ADob4QeCIvxAUIQfCIrwA0ERfiCoIkb1hbB48eLc2hVXXJFcd+HChcn6rl27kvXBwcFkff/+/bm1ffv2JddFXOz5gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAohvTWaWhoKLc2e/bs5LojIyPJeldXV7I+b968ZP2TTz7Jre3Zsye57plnnpmsj42NJesnTpxI1qdPn55bGx0dTa47derUZP3TTz9N1nfuzP9umUcffTS5bq17K9oZQ3oBJBF+ICjCDwRF+IGgCD8QFOEHgiL8QFCM589MmTIlWV+zZk1u7YUXXkiue9NNNyXrPT09yfqCBQuS9WXLluXWUt9DIEnvv5+eYHnOnDnJejOOHz+erH/wwQfJ+qxZs5L1RYsW5dZq/Xt38nX+erHnB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGgao7nN7P1kr4v6bC7L8yWnSPpd5LmSzog6UZ3T1+UVWeP50+p4zOsbPvXXHNNct2BgYFk/ZJLLknWp02blqynxtx//PHHyXVrfQ/CK6+8kqynvmfhzjvvTK778MMPJ+vtrMjx/L+RdOpdJPdIGnD3CyUNZM8BdJCa4Xf3rZKOnLJ4uaQN2eMNkq4vuC8AJWv0nP88dz8oSdnPc4trCUArlH5vv5n1SeorezsATk+je/5DZtYtSdnPw3kvdPd+d+91994GtwWgBI2Gf5OkVdnjVZI2FtMOgFapGX4ze1rS3yV9y8xGzOxHkh6UtNTM/iVpafYcQAepec7v7itzSksK7qVjPfDAA8l6rev8zc6dkHr/Mt+7bCtWrEjW586dm6zv2LEjt/bMM8801NOXCXf4AUERfiAowg8ERfiBoAg/EBThB4Jiiu4WKPtSX6fq7u5O1mtNLz5z5sxk/eqrr86tbd68ObluJ2OKbgBJhB8IivADQRF+ICjCDwRF+IGgCD8QFFN0t0DU6/i13HLLLcn6jBkzkvUPP/wwWR8eHj7tniJhzw8ERfiBoAg/EBThB4Ii/EBQhB8IivADQTGevwM089+o7O8SeOedd5L1m2++Obe2ZcuW5LpTp05N1q+66qpkfevWrcn6lxXj+QEkEX4gKMIPBEX4gaAIPxAU4QeCIvxAUDXH85vZeknfl3TY3Rdmy9ZKuk3Se9nL1rj782U1GV2Z02SXPQX3/fffn1urdR1/YGAgWX/99dcb6gnj6tnz/0bSskmW/8rde7J/CD7QYWqG3923SjrSgl4AtFAz5/x3mNkOM1tvZmcX1hGAlmg0/I9I+qakHkkHJf0i74Vm1mdmg2Y22OC2AJSgofC7+yF3H3X3MUmPSbo48dp+d+91995GmwRQvIbCb2YTp1f9gaSdxbQDoFXqudT3tKQrJc0ysxFJ90m60sx6JLmkA5JuL7FHACVgPD+aMm3atGT95Zdfzq0tWLAgue7ixYuT9e3btyfrUTGeH0AS4QeCIvxAUIQfCIrwA0ERfiAopuhGU+6+++5kvaenJ7f2/PPpwaBcyisXe34gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIohvUi67rrrkvVnn302WT9x4kRu7dprr02uu23btmQdk2NIL4Akwg8ERfiBoAg/EBThB4Ii/EBQhB8IivH8wZ111lnJ+rp165L1M844I1nfuHFjbo3r+NVizw8ERfiBoAg/EBThB4Ii/EBQhB8IivADQdUcz29mcyU9Iel8SWOS+t39ITM7R9LvJM2XdEDSje7+QY33Yjx/i02ZMiVZf/XVV5P11PfuS9LevXuT9aVLl+bWhoeHk+uiMUWO5/9M0l3uvkDSYkmrzezbku6RNODuF0oayJ4D6BA1w+/uB919KHt8VNJuSbMlLZe0IXvZBknXl9UkgOKd1jm/mc2XdJGk1ySd5+4HpfFfEJLOLbo5AOWp+95+M5sh6Q+SfuLuH5nVdVohM+uT1NdYewDKUtee38ymaDz4T7r7H7PFh8ysO6t3Szo82bru3u/uve7eW0TDAIpRM/w2vot/XNJud//lhNImSauyx6sk5Q/fAtB26rnUd7mklyS9pfFLfZK0RuPn/b+XNE/SsKQb3P1IjffiUl+LzZ8/P1nfv39/sl7r/4/ly5cn688991yyjuLVe6mv5jm/u2+TlPdmS06nKQDtgzv8gKAIPxAU4QeCIvxAUIQfCIrwA0Hx1d1fAhdccEFu7cUXX0yuOzY2lqzfe++9yTrX8TsXe34gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIrr/F8Ct956a26tu7u7qfceGBhoan20L/b8QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxAU1/k7wKJFi5L11atX59bqnVYtT63x/uhc7PmBoAg/EBThB4Ii/EBQhB8IivADQRF+IKia1/nNbK6kJySdL2lMUr+7P2RmayXdJum97KVr3P35shqN7NJLL03WZ8yYkVtz9+S6w8PDyfqxY8eSdXSuem7y+UzSXe4+ZGYzJb1pZluy2q/cfV157QEoS83wu/tBSQezx0fNbLek2WU3BqBcp3XOb2bzJV0k6bVs0R1mtsPM1pvZ2Tnr9JnZoJkNNtUpgELVHX4zmyHpD5J+4u4fSXpE0jcl9Wj8yOAXk63n7v3u3uvuvQX0C6AgdYXfzKZoPPhPuvsfJcndD7n7qLuPSXpM0sXltQmgaDXDb+PDwh6XtNvdfzlh+cSvhf2BpJ3FtwegLPX8tf8yST+U9JaZbc+WrZG00sx6JLmkA5JuL6VDqKurK1kfHR3NrQ0NDSXXXbJkSbJ+9OjRZB2dq56/9m+TNNmgcK7pAx2MO/yAoAg/EBThB4Ii/EBQhB8IivADQVmtIZ+FbsysdRsDgnL3ur6vnT0/EBThB4Ii/EBQhB8IivADQRF+ICjCDwTV6im635f07wnPZ2XL2lG79taufUn01qgie/tavS9s6U0+X9i42WC7frdfu/bWrn1J9NaoqnrjsB8IivADQVUd/v6Kt5/Srr21a18SvTWqkt4qPecHUJ2q9/wAKlJJ+M1smZntMbO9ZnZPFT3kMbMDZvaWmW2veoqxbBq0w2a2c8Kyc8xsi5n9K/s56TRpFfW21sz+m312283s2op6m2tmfzOz3Wb2tpn9OFte6WeX6KuSz63lh/1m1iXpn5KWShqR9Iakle6+q6WN5DCzA5J63b3ya8Jm9l1JxyQ94e4Ls2U/l3TE3R/MfnGe7e4/bZPe1ko6VvXMzdmEMt0TZ5aWdL2km1XhZ5fo60ZV8LlVsee/WNJed9/n7scl/VbS8gr6aHvuvlXSkVMWL5e0IXu8QeP/87RcTm9twd0PuvtQ9viopJMzS1f62SX6qkQV4Z8t6T8Tno+ovab8dkmbzexNM+uruplJnJdNm35y+vRzK+7nVDVnbm6lU2aWbpvPrpEZr4tWRfgn+4qhdrrkcJm7f0fSNZJWZ4e3qE9dMze3yiQzS7eFRme8LloV4R+RNHfC8zmS3q2gj0m5+7vZz8OS/qT2m3340MlJUrOfhyvu53PtNHPzZDNLqw0+u3aa8bqK8L8h6UIz+7qZTZW0QtKmCvr4AjObnv0hRmY2XdL31H6zD2+StCp7vErSxgp7+T/tMnNz3szSqviza7cZryu5ySe7lPFrSV2S1rv7z1rexCTM7Bsa39tL4yMen6qyNzN7WtKVGh/1dUjSfZL+LOn3kuZJGpZ0g7u3/A9vOb1dqfFD189nbj55jt3i3i6X9JKktySNZYvXaPz8urLPLtHXSlXwuXGHHxAUd/gBQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFCEHwjqf+hLHZWDVnX2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pred_class = explanation['cf']['class']\n",
    "proba = explanation['cf']['proba'][0][pred_class]\n",
    "\n",
    "print(f'Counterfactual prediction: {pred_class} with probability {proba}')\n",
    "plt.imshow(explanation['cf']['X'].reshape(28, 28));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The counterfactual starting from a 7 moves towards the closest class as determined by the model and the data: a 9. The evolution of the counterfactual during the iterations over $\\lambda$ can be seen below (note that all of the following examples satisfy the counterfactual condition):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_cfs = np.array([len(explanation['all'][iter_cf]) for iter_cf in range(max_lam_steps)])\n",
    "examples = {}\n",
    "for ix, n in enumerate(n_cfs):\n",
    "    if n>0:\n",
    "        examples[ix] = {'ix': ix, 'lambda': explanation['all'][ix][0]['lambda'],\n",
    "                       'X': explanation['all'][ix][0]['X']}\n",
    "columns = len(examples) + 1\n",
    "rows = 1\n",
    "\n",
    "fig = plt.figure(figsize=(16,6))\n",
    "\n",
    "for i, key in enumerate(examples.keys()):\n",
    "    ax = plt.subplot(rows, columns, i+1)\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "    plt.imshow(examples[key]['X'].reshape(28,28))\n",
    "    plt.title(f'Iteration: {key}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Typically, the first few iterations find counterfactuals that are out of distribution, while the later iterations make the counterfactual more sparse and interpretable."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's now try to steer the counterfactual to a specific class:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Explanation took 4.563 sec\n"
     ]
    }
   ],
   "source": [
    "target_class = 1\n",
    "\n",
    "cf = CounterFactual(sess, cnn, shape=shape, target_proba=target_proba, tol=tol,\n",
    "                    target_class=target_class, max_iter=max_iter, lam_init=lam_init,\n",
    "                    max_lam_steps=max_lam_steps, learning_rate_init=learning_rate_init,\n",
    "                    feature_range=feature_range)\n",
    "\n",
    "explanation = start_time = time()\n",
    "explanation = cf.explain(X)\n",
    "print('Explanation took {:.3f} sec'.format(time() - start_time))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counterfactual prediction: 1 with probability 0.9983848333358765\n"
     ]
    },
    {
     "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": [
    "pred_class = explanation['cf']['class']\n",
    "proba = explanation['cf']['proba'][0][pred_class]\n",
    "\n",
    "print(f'Counterfactual prediction: {pred_class} with probability {proba}')\n",
    "plt.imshow(explanation['cf']['X'].reshape(28, 28));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As you can see, by specifying a class, the search process can't go towards the closest class to the test instance (in this case a 9 as we saw previously), so the resulting counterfactual might be less interpretable. We can gain more insight by looking at the difference between the counterfactual and the original instance:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAADk9JREFUeJzt3W+IXfWdx/HP18mfC0keRGLMYKN2iywNhU3LEFdcFkux2LIQy1BpHpQUwqaDVbZQoZIn9cmClP5TWBumTWiE1rYwVfNA2kRZ0EIpjhJqanatSLZNMyajKclEucn8+e6DOWmn8d7fubnn3535vl8wzL3nd+493zn3fuace3/nnJ+5uwDEc13TBQBoBuEHgiL8QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxDUqjoX1mq1fMOGDXUuEjnMLNnOEaDLy8zMjNrtdvpFzRQKv5ndI+kxSUOSfujuj6bm37Bhg0ZHR4ssEiUj/CvLxMREz/P2vdtvZkOS/kvSZyRtk7TLzLb1+3wA6lXkM/8OSW+6+1vuflnSTyXtLKcsAFUrEv6bJP1pyf1T2bS/Y2Z7zWzSzCbb7XaBxQEoU5Hwd/qw+IEPiO4+7u4j7j7SarUKLA5AmYqE/5SkrUvuf0jS6WLlAKhLkfC/LOk2M/uwma2R9AVJh8spC0DV+u7qc/c5M3tA0q+02NV30N1/X1plqAVdeXEV6ud39+ckPVdSLQBqxOG9QFCEHwiK8ANBEX4gKMIPBEX4gaBqPZ8fKw+nBC9fbPmBoAg/EBThB4Ii/EBQhB8IivADQdHVF1zRrrrHH3882f7ggw9ec02oB1t+ICjCDwRF+IGgCD8QFOEHgiL8QFCEHwiKfv7gip5yu3r16pIqQd3Y8gNBEX4gKMIPBEX4gaAIPxAU4QeCIvxAUIX6+c3spKQZSfOS5tx9pIyiMDjyzvfH8lXGQT6fdPd3SngeADVitx8Iqmj4XdIRM3vFzPaWURCAehTd7b/T3U+b2WZJR83sf9z9xaUzZP8U9krS+vXrCy4OQFkKbfnd/XT2+6ykpyXt6DDPuLuPuPtIq9UqsjgAJeo7/Ga2zsw2XLkt6dOSjpdVGIBqFdntv1HS01lX0CpJP3H3X5ZSFYDK9R1+d39L0j+VWMuKxTDWuFrqPVHX+4GuPiAowg8ERfiBoAg/EBThB4Ii/EBQXLq7R9dd1/3/5OzsbKHnzrv89cLCQrK9SLdRXjfkqlXVvUWq7gIdHh7u2jY1NVXoufPs378/2f7QQw91bbt48WLZ5XTElh8IivADQRF+ICjCDwRF+IGgCD8QFOEHgqKfv0epvvbUMQCSdO7cuWT7xo0bk+1HjhxJtp8/f75r2x133JF87M0335xsf/fdd5Pt999/f7L9pZde6to2MzOTfOzatWuT7bfcckuyPfW65L1mecdW5BkbGyv0+Dqw5QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoOjnz+SdW57qc7506VLysZs2bSq07Ntvvz3ZfvTo0a5tb7zxRvKxW7ZsSba//fbbyfb33nsv2f766693bRsaGko+Nq+vfdu2bcn2d97pPnh03rEVEYYmZ8sPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0FZD9d1Pyjp3ySddfePZdOul/QzSbdKOinpPnf/S97CbrjhBh8dHS1Y8spTZZ9y3uubd878unXrku155/un/ra8c+rXrFmTbL9w4UKyfXp6umtb6pr+Un5tg2piYkLT09M9vaF6+Qt/JOmeq6Y9LOkFd79N0gvZfQDLSG743f1FSVdfimanpEPZ7UOS7i25LgAV63ff5kZ3n5Kk7Pfm8koCUIfKP9iY2V4zmzSzyXa7XfXiAPSo3/CfMbNhScp+n+02o7uPu/uIu4+0Wq0+FwegbP2G/7Ck3dnt3ZKeLaccAHXJDb+ZPSXpN5L+0cxOmdkeSY9KutvM/iDp7uw+gGUk93x+d9/VpelTJddSqarHgh/UZefJuxZBXnuRYxTyztfPW3beMQqpvvy8Ywjm5uaS7SvB8jySAUBhhB8IivADQRF+ICjCDwRF+IGgwly6u8nutCaXPcjyugnzugLHx8eT7alhsiN05eVhyw8ERfiBoAg/EBThB4Ii/EBQhB8IivADQYXp50czUpfALnp57D179iTb5+fnK1v2SsAaAIIi/EBQhB8IivADQRF+ICjCDwRF+IGg6OdHpYaGhrq2zc7OJh+bdz5/6rmlaoc+XwnY8gNBEX4gKMIPBEX4gaAIPxAU4QeCIvxAULnhN7ODZnbWzI4vmfaImf3ZzI5lP5+ttkwsV/Pz811/8gwNDRX6QVovW/4fSbqnw/Tvuvv27Oe5cssCULXc8Lv7i5LO1VALgBoV+cz/gJn9LvtYsLG0igDUot/wf1/SRyRtlzQl6dvdZjSzvWY2aWaT7Xa7z8UBKFtf4Xf3M+4+7+4Lkn4gaUdi3nF3H3H3kVar1W+dAErWV/jNbHjJ3c9JOt5tXgCDKfeUXjN7StJdkjaZ2SlJ35B0l5ltl+SSTkr6coU1AqhAbvjdfVeHyQcqqKVReed+u3tNlSwvede/T623vHWadz4/194vhrUHBEX4gaAIPxAU4QeCIvxAUIQfCIpLd6OQVavSb6HLly93bSvSTYji2PIDQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFArpp+/6Cm59Cn359KlS8n21CW0WefNYssPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0Etq37+VF8+fcad5R3/kKfK9cpr1iy2/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QVG4/v5ltlfSkpC2SFiSNu/tjZna9pJ9JulXSSUn3uftfqiu1WL9w3jXin3jiib6f+8CB9Ijlk5OTyfYq+7vz+vnn5+eT7UWHwZ6dne3aljrXH9Xr5ZWdk/Q1d/+opH+W9BUz2ybpYUkvuPttkl7I7gNYJnLD7+5T7v5qdntG0glJN0naKelQNtshSfdWVSSA8l3TPp2Z3Srp45J+K+lGd5+SFv9BSNpcdnEAqtNz+M1svaQJSV919wvX8Li9ZjZpZpPtdrufGgFUoKfwm9lqLQb/x+7+i2zyGTMbztqHJZ3t9Fh3H3f3EXcfabVaZdQMoAS54bfFr4sPSDrh7t9Z0nRY0u7s9m5Jz5ZfHoCq9HJK752SvijpNTM7lk3bJ+lRST83sz2S/ijp89WU+DdFTk/N604bGxvr+7kH2cLCQrK96CXP84bo3r9/f9e2lbrOl4vc8Lv7ryV1e4d8qtxyANSFI/yAoAg/EBThB4Ii/EBQhB8IivADQS2rS3en+pzz+qv37duXbN+6dWtfNfWi6v7s1N+e18+fd8pu0VOCU3976hiAMnAcQRpbfiAowg8ERfiBoAg/EBThB4Ii/EBQhB8IyuocJnnz5s0+OjratZ0hm8s3yEN0r2RFr5PQr4mJCU1PT/f0orPlB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGgaj2f393pN+7DmjVrku3vv/9+17a8YbBTQ2hL0tq1a5PtvJ6dLYf1wpYfCIrwA0ERfiAowg8ERfiBoAg/EBThB4LK7ec3s62SnpS0RdKCpHF3f8zMHpH075Kms1n3uftzVRUa2eXLl5Ptq1ev7tqWd93+vGMI8q7Ln3fd/6iaOp//WvRykM+cpK+5+6tmtkHSK2Z2NGv7rrt/q7ryAFQlN/zuPiVpKrs9Y2YnJN1UdWEAqnVN+2xmdqukj0v6bTbpATP7nZkdNLONXR6z18wmzWyy3W4XKhZAeXoOv5mtlzQh6avufkHS9yV9RNJ2Le4ZfLvT49x93N1H3H2k1WqVUDKAMvQUfjNbrcXg/9jdfyFJ7n7G3efdfUHSDyTtqK5MAGXLDb8tfm15QNIJd//OkunDS2b7nKTj5ZcHoCq9fNt/p6QvSnrNzI5l0/ZJ2mVm2yW5pJOSvlxJhcjtFkp1561alX6J87rq5ubmku3oLO81KzI8+TPPPNO17fnnn+/5eXr5tv/Xkjp1WtKnDyxjHKEBBEX4gaAIPxAU4QeCIvxAUIQfCKrWS3c3aTmcYtlNXu2p9rxTevPaUY2xsbFKnvf8+fM9z8uWHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeCsjr7t81sWtL/LZm0SdI7tRVwbQa1tkGtS6K2fpVZ2y3ufkMvM9Ya/g8s3GzS3UcaKyBhUGsb1LokautXU7Wx2w8ERfiBoJoO/3jDy08Z1NoGtS6J2vrVSG2NfuYH0Jymt/wAGtJI+M3sHjP7XzN708webqKGbszspJm9ZmbHzGyy4VoOmtlZMzu+ZNr1ZnbUzP6Q/e44TFpDtT1iZn/O1t0xM/tsQ7VtNbP/NrMTZvZ7M/uPbHqj6y5RVyPrrfbdfjMbkvSGpLslnZL0sqRd7v56rYV0YWYnJY24e+N9wmb2r5IuSnrS3T+WTfumpHPu/mj2j3Oju399QGp7RNLFpkduzgaUGV46srSkeyV9SQ2uu0Rd96mB9dbEln+HpDfd/S13vyzpp5J2NlDHwHP3FyWdu2ryTkmHstuHtPjmqV2X2gaCu0+5+6vZ7RlJV0aWbnTdJepqRBPhv0nSn5bcP6XBGvLbJR0xs1fMbG/TxXRwYzZs+pXh0zc3XM/VckdurtNVI0sPzLrrZ8TrsjUR/k7XnBqkLoc73f0Tkj4j6SvZ7i1609PIzXXpMLL0QOh3xOuyNRH+U5K2Lrn/IUmnG6ijI3c/nf0+K+lpDd7ow2euDJKa/T7bcD1/NUgjN3caWVoDsO4GacTrJsL/sqTbzOzDZrZG0hckHW6gjg8ws3XZFzEys3WSPq3BG334sKTd2e3dkp5tsJa/MygjN3cbWVoNr7tBG/G6kYN8sq6M70kaknTQ3f+z9iI6MLN/0OLWXlq8svFPmqzNzJ6SdJcWz/o6I+kbkp6R9HNJN0v6o6TPu3vtX7x1qe0uLe66/nXk5iufsWuu7V8kvSTpNUlXLk+8T4ufrxtbd4m6dqmB9cYRfkBQHOEHBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiCo/wc3qnTvaM6m2AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow((explanation['cf']['X'] - X).reshape(28, 28));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This shows that the counterfactual is stripping out the top part of the 7 to make to result in a prediction of 1 - not very surprising as the dataset has a lot of examples of diagonally slanted 1's."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clean up:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.remove('mnist_cnn.h5')"
   ]
  }
 ],
 "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"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}