8 - TensorFlow CNN Bayesian.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 8 - TensorFlow CNN with Bayesian Convolutional Layer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we've trained a CNN including an estimator of aleatoric uncertainty, let's take a look at how to train that same network with support for _epistemic_ uncertainty -- uncertainty in the shape of the model itself. \n",
    "\n",
    "We've so far only used the `tfp.layers.DenseVariational` for epistemic uncertainty in neural networks, but TensorFlow Probability ships with a full complement of Bayesian layers that define distributions over their weights. Check out the full list in the documentation, [here](https://www.tensorflow.org/probability/api_docs/python/tfp/layers)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.0.0\n"
     ]
    }
   ],
   "source": [
    "from __future__ import absolute_import, division, print_function, unicode_literals\n",
    "\n",
    "import os\n",
    "import pickle\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import datasets, layers, models\n",
    "import tensorflow_probability as tfp\n",
    "\n",
    "print(tf.__version__)\n",
    "tf.get_logger().setLevel('ERROR')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll use the same `plot_image` and `plot_value_array` functions that we've defined for all of the image classification problems."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_image(i: int, predictions_array: np.ndarray, true_label: np.ndarray, img: np.ndarray):\n",
    "    \"\"\"\n",
    "    Plot an image from the passed set, along with its true and predicted label.\n",
    "    \"\"\"\n",
    "    \n",
    "    predictions_array, true_label, img = predictions_array, true_label[i], img[i]\n",
    "    plt.grid(False)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "\n",
    "    plt.imshow(img, cmap=plt.cm.binary)\n",
    "\n",
    "    true_label = true_label.squeeze()\n",
    "    \n",
    "    predicted_label = np.argmax(predictions_array)\n",
    "    if predicted_label == true_label.squeeze():\n",
    "        color = 'blue'\n",
    "    else:\n",
    "        color = 'red'\n",
    "\n",
    "\n",
    "    plt.xlabel(\"{} {:2.0f}% ({})\".format(class_names[predicted_label],\n",
    "                                100*np.max(predictions_array),\n",
    "                                class_names[true_label]),\n",
    "                                color=color)\n",
    "\n",
    "def plot_value_array(i: int, predictions_array: np.ndarray, true_label: np.ndarray):\n",
    "    \"\"\"\n",
    "    Plot a bar chart of the model activations for each class pertaining to a given image.\n",
    "    \"\"\"\n",
    "    \n",
    "    predictions_array, true_label = predictions_array, true_label[i]\n",
    "    \n",
    "    true_label = true_label.squeeze()\n",
    "    \n",
    "    plt.grid(False)\n",
    "    plt.xticks(range(2))\n",
    "    plt.yticks([])\n",
    "    thisplot = plt.bar(range(2), predictions_array, color=\"#777777\")\n",
    "    plt.ylim([0, 1])\n",
    "    predicted_label = np.argmax(predictions_array)\n",
    "\n",
    "    thisplot[predicted_label].set_color('red')\n",
    "    thisplot[true_label].set_color('blue')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load training data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Once again, we'll use CIFAR-10 as our training data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()\n",
    "\n",
    "# Normalize pixel values to be between 0 and 1\n",
    "train_images, test_images = train_images / 255.0, test_images / 255.0\n",
    "\n",
    "train_images = np.array(train_images[[i for (i, label) in enumerate(train_labels) if label in [0, 2]]])\n",
    "train_labels = np.array([label for label in train_labels if label in [0, 2]]) // 2\n",
    "test_images = np.array(test_images[[i for (i, label) in enumerate(test_labels) if label in [0, 2]]])\n",
    "test_labels = np.array([label for label in test_labels if label in [0, 2]]) // 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 25 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "class_names = ['airplane', 'bird']\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "for i in range(25):\n",
    "    plt.subplot(5,5,i+1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.grid(False)\n",
    "    plt.imshow(train_images[i], cmap=plt.cm.binary)\n",
    "    # The CIFAR labels happen to be arrays, \n",
    "    # which is why you need the extra index\n",
    "    plt.xlabel(class_names[train_labels[i][0]])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define model\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll define the model in the exact same fashion as before, only this time replacing the first convolutional layer in the model with a `tfp.layers.Convolution2DFlipout` layer object. The layer by default applies a normal prior distribution over its weights, so we don't need to define a prior ourselves."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = models.Sequential()\n",
    "model.add(tfp.layers.Convolution2DFlipout(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_flipout (Conv2DFlipou (None, 30, 30, 32)        1760      \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d (Conv2D)              (None, 13, 13, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 4, 4, 64)          36928     \n",
      "=================================================================\n",
      "Total params: 57,184\n",
      "Trainable params: 57,184\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(2, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_flipout (Conv2DFlipou (None, 30, 30, 32)        1760      \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d (Conv2D)              (None, 13, 13, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 4, 4, 64)          36928     \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                65600     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 2)                 130       \n",
      "=================================================================\n",
      "Total params: 122,914\n",
      "Trainable params: 122,914\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "checkpoint_path = 'saved_models/bayesian_cnn_conv2dflipout.ckpt'\n",
    "\n",
    "model.compile(optimizer=tf.optimizers.Adam(0.001),\n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'],\n",
    "              experimental_run_tf_function=False  # Required argument for this probabilistic layer (and all Flipouts)\n",
    "             )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's again load this model's trained weights from disk -- it takes about 10 minutes to train on a CPU."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # Run this to train the model from scratch\n",
    "\n",
    "# # \n",
    "\n",
    "# cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,\n",
    "#                                                  save_weights_only=True,\n",
    "#                                                  verbose=1)\n",
    "\n",
    "# history = model.fit(train_images, train_labels, epochs=20, \n",
    "#                     validation_data=(test_images, test_labels), callbacks=[cp_callback])\n",
    "\n",
    "# history_dict = history.history\n",
    "\n",
    "# with open(os.path.join(os.getcwd(), 'saved_models', 'bayes_cnn_history.pkl'), 'wb') as file:\n",
    "#     pickle.dump(history_dict, file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run this to instead load pre-trained weights\n",
    "model.load_weights(checkpoint_path)\n",
    "\n",
    "with open(os.path.join(os.getcwd(), 'saved_models', 'bayes_cnn_history.pkl'), 'rb') as file:\n",
    "    history_dict = pickle.load(file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2000/2000 - 1s - loss: 0.4588 - accuracy: 0.7965\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAgAElEQVR4nO3deZwV5Zn3/8/V+0rTDTRbg4Ci7Ih2XBMlLglmjCbmIeiYRNGEMQmOyzNjjFn05/jLmOVJfvqMY4IZt5n4mMSMCfExmrhFjZrQuKACCrJIs/UC9ELT+/X7o6qbQ3MaDnTX6Ybzfb9e9artPlXXqT59X+fcVXWXuTsiIpK60gY6ABERGVhKBCIiKU6JQEQkxSkRiIikOCUCEZEUp0QgIpLiIksEZna/mVWZ2Tu9rDczu9vM1prZCjM7KapYRESkd1H+IngQmHeA9RcAk8NhEXBvhLGIiEgvIksE7v4isOMARS4GHvbAa8BQMxsdVTwiIhJfxgDueyywKWa+Mly2tWdBM1tE8KuB/Pz8k6dMmZKUAEVEjhbLly+vcfcR8dYNZCJImLsvAZYAlJeXe0VFxQBHJCJyZDGzjb2tG8irhjYD42Lmy8JlIiKSRAOZCJYCXwqvHjoNqHP3/ZqFREQkWpE1DZnZ/wHmAsPNrBK4FcgEcPefAk8CnwLWAk3AwqhiERGR3kWWCNz9soOsd+DrUe1fREQSozuLRURSnBKBiEiKUyIQEUlxSgQiIilOiUBEJMUpEYiIpDglAhGRFKdEICKS4pQIRERSnBKBiEiKUyIQEUlxSgQiIilOiUBEJMUpEYiIpDglAhGRFKdEICKS4pQIRERSnBKBiEiKizQRmNk8M3vPzNaa2c1x1h9jZs+a2Qoze8HMyqKMR0RE9hdZIjCzdOAe4AJgGnCZmU3rUexHwMPuPgu4HfjXqOIREZH4ovxFcAqw1t3XuXsr8ChwcY8y04Dnwunn46wXEZGIRZkIxgKbYuYrw2Wx3gIuCac/CxSa2bAIYxIRkR4G+mTxPwFnm9kbwNnAZqCjZyEzW2RmFWZWUV1dnewYRUSOalEmgs3AuJj5snBZN3ff4u6XuPsc4Fvhsl09N+TuS9y93N3LR4wYEWHIIiKpJ8pEsAyYbGYTzSwLuBRYGlvAzIabWVcM3wTujzAeERGJI7JE4O7twGLgaWAV8Ct3f9fMbjezi8Jic4H3zOx9YCTw/0YVj4iIxGfuPtAxHJLy8nKvqKgY6DBERI4oZrbc3cvjrRvok8UiIjLAlAhERFKcEoGISIpTIhARSXFKBCIiKU6JQEQkxSkRiIikOCUCEZEUp0QgIpLiMgY6ABER2au5rYPNu/ZQuXMPm3Y0BeOdTVTuaOKrc49j3oxR/b5PJQIRkSRq6+hk667moHLf2cSmHXuCcVjxVzW07FM+M90YOzSXcSV5ZGVYJDEpEYiIRMDd2VjbxIrNdbxduYt3Ntfz4Y4mttbtoTOmi7f0NGN0UQ5lxbmcffwIyorzGFcSVPxlxbmMLMwhLS2aBNBFiUBEpI/cnc279vB2ZV1Y8dexonIX9c3tAGRlpDF19BBOmVjCuOJcyorzKCvJZVxxHqOLcshIH9jTtUoEIiKHaHt9MyvCyn5FZR1vb65jx+5WIGjKmTJqCBfOHsOssUXMLCvi+JGFZA5wZX8gSgQiIjHcncaWdmobW6lpbKEmHFfVN7Nyaz0rKuu62/HT04zJpQWcN7WUmWVDmTW2iBNGFZKTmT7A7+LQKBGIyFGvpb2DhuZ2duxupaahherGlu6Kfm+Fv7fSb2nv3G8bZnDciAI+Onl4+E1/KNNGDyE368iq9ONRIhCRQW9Pawdb6vZQt6eNhuZ2GpqDcWM4Xd/cTmPL3uUNMfP1ze20xqnYIfhGPyw/i+EF2QwryOLYEQUML8zeZ9nwgmyGF2RTkp9FVsbgbd7pCyUCERlwzW0dVO4MLqMMxuG18zv3sHlnEzWNrb2+1gwKsjIozMmgMCeTgpwMhhVkMWF4frAse++6krCCH1GYxbD8bIpyMyO/IudIoEQgIpHp7HSa2jrYHX4737yred/KPrxhqqYx/rXzZcV5nDd1JONK8hgzNIeheVlhxZ4ZVu4Z5GdlqDLvo0gTgZnNA+4C0oGfu/udPdaPBx4ChoZlbnb3J6OMSUR611VxN7W0s7s1qMCbWjvY3drO7pZgaGzpiJned1ljj+VNbR3Eeyx6ZroxZmguZcW5nDullLLivdfNlxXnUVqYrco9iSJLBGaWDtwDnA9UAsvMbKm7r4wp9m3gV+5+r5lNA54EJkQVk0gqaevopHLnHjbU7GZdzW427Wiiobmdptagku+q7Ju6K/kO9rR1JLz97Iw0CrIzyA+Hgux0hhVkMX5YHgVZe5d1rS/MyWB0US7jSnIpLcwhXRV9YtyhpQEatkHeMMgf1u+7iPIXwSnAWndfB2BmjwIXA7GJwIEh4XQRsCXCeESOOp2dzpa6Payv2d1d4W+o2c2G2iY27WiiPeYW1oLsDIpyM8nPTicvK4P87HSK87PIz0onLzuD/Kyw0s7KIC87PRiHy7rG+dkZFITrB/N18UeM1iZo2BpU8vuNY6bbdgflL/wJlF/V72FEmQjGApti5iuBU3uUuQ34o5ldC+QD58XbkJktAhYBjB8/vt8DFRksOjqd5rYOWto7aWnvoKWtk+ZwvLulnY07mthQs5v14bBxR9M+V8TkZqYzYXg+00YP4VMzRzFhWD6TRuQzYVg+JflZmOlb+GHpaIfGbUHF3dEKHS3Q0Qbt4bijJVje3tr7+vZm2F2zb4XfXLf/vjJyoHB0MIyeDcfPg8JRwXxZeSRvb6BPFl8GPOju/8vMTgf+08xmuPs+13q5+xJgCUB5eXmcFkeRwWfH7lbWbG9gTVUja6sa2VC7m6bWsJJv2ztujpmP/Qbfm6z0NMYPy2Pi8Hw+PqWUicODin7i8HxGDskevJV9eyukZwaX+Qw2nZ1BRb/rQ9i5MRjv2rB3vn4zdLb3bR8ZOZA/IqjUh0+GiWftreBjxzlDk36MokwEm4FxMfNl4bJYVwPzANz9VTPLAYYDVRHGJdJv3J2axlbWVDWwtqqR97c3sGZ7UPHX7t57yWN+VjoTR+RTkJ3B0NxMcoZkk52RTnZGGtmZaeRkpJOdmUZ2Rjo54bjnupzMdMYV5zFmaG7i7et7dsGHr8HGl2HrW5A9BApKoWBkzDiczi+FzJzDOxAtjVC/Beorw/GWoPKs27x3unkXYJBVANkFPcaFCcznQ1oGWHpQUVra3iEtfd95s7BczDI8iGXXxpjKvmu8KfjWHqtgJAw9Bso+AsX/A4rGBXGkZ0J6djDOyIb0rL1DRnb89WkZgzMBhqJMBMuAyWY2kSABXAr8fY8yHwLnAg+a2VQgB6iOMCaRw9LR6VQ3tOyt7KsaWVsVjHc1tXWXK8zJCLscGMnkkQVMHlnI5NICRhflJOebetMO2PgKbPwLbHgZtr0NeFAZjZwRNE18+Co01cZ/fXZRnEQRjnOLoakmTiW/BVriNHHkj4AhY6B4AhxzRrCNjlZobQxOfrY2BgmktTGoiFsb9s63N0d5lAK5JVB8THBcTvhUMD20axgHmbnRxzBIRJYI3L3dzBYDTxNcGnq/u79rZrcDFe6+FPifwH1mdgPBieMr3eNdbCZyGNr2wJY3oHJZMFStDr7x5gwNvhnnDIHsIbRlFlDveezszKW2PYeq1my2tWRR2ZTJh7szWN+YzqZGpyOmwbIoN5PjRxZwwYzRHD+ygMmlhUweWUBpYZKbZhqrg0p/419gw1+g6t1geUZO8E127s1wzJlB23JsxdbRFiSFxu3QWBWOY6ergl8QjVVBBd1Twcigkh92LEz8GAwZGw5j9g4Z2Yf/vjra9k0UXePODvDOHkPsMt873bMsDgWjwgp/fPDtXgCwI63eLS8v94qKioEOQwYbd9i5YW+lv+lvsP2d7nbdPQXjqcqfTFtrC9ZST0ZbIzntDeR27qaAPaTZgf8POiyD1swiWosnkzlmJrllM7FRM2DEVMjKS8IbDDVsC77pd1X8Ne8FyzPzYNypMOFMOOajMPakvlXEsVqbYHdV8Csib3jQlp2R1T/blqQxs+XuHvds80CfLBY5PC2N4bf9v9G28W/Y5goy9tQEq9JyWZtxPMvtYl5sm8AbHcdR21wENUHfMiMKshk5NIdRQ7IZNSSHkUOyKMvrZExOKyOzWxmW0Uxe525org+aPJrrSW+uI7ephtyq1fDOL+CNpjAQg5JJMHJa0MQwcjqUToPiiZB2iJdXugdXkcS2sXcPW2DHeti5PiibVQjjT4MTLwsq/jEnBm3SUcjKg6wJQROPHJWUCGTQc3dqK9dQt/oFvHIZQ6rfYHjTB6QRtNV82DmaN3wab3Qex+udk6nNncSYokKOGZbHtJI8PlmSxzHD8hlfkseIwuy+38jU2RlcUbL9Xdi+Mvjlsf1dWPUEQQsnkJkPpVOCxDByRpAchh0btOF3V/Jb9q3o67cEzR/7sOBKkiFjYNRM+MjVQVPPqFmQrn9f6R9qGpJBwd3Z2dTWfX18zeZ15G55hdE7ljGt5S3KLLiGoN5zecuP44OsKVQXzaZ51BxKS0czviSP8cPyGF+SR2FORN+MD6Z1N1Sv3j9B7NkRv7ylBc0s3e3qse3s4bhwVHTf9CWlqGlIBo365rZ9bojaULOb9bVN1FVXMrN1BaenreS0tJVMStsGQGNaIZuGzuFvoxeSNvEsRkyazWklBXxsMN7VmpUPY08Ohi7uwcnX7e8E5zDyhu+t5AtG6lu9DAr6FEqkPqxt4qW11by8poZlG3Z29zJZTD2npa/mvJz3+Gd7l7F8CFnQnllAy9jT6Ji8mPRJZ1EwcgZT047gB39Y2LRTOGqgIxHplRKB9Ku6pjZe+aCGl9bW8PKaGj7cEZxUPWFIG18bXUm5v8PEhtcprAuvdknLD056TrwKJp5FxqjZZOhbskhS6T9O+qS1vZPXP9zJy2uCyv/tyl10OhyXvYsvlFZydskaJjatIGvHe0HPUxk5wWWOJ8+HCWcFlzmqDVxkQCkRyCFxd9ZUNfLSmhpeXlPNX9fvoKm1nRPStvDZYRu5s2wtk5pWkL17S3CPeFYhjDsFTpwP488Ibmzqr+vbRaRfKBFIrzo6ncqdTayr3s0H1Y2s3FrPX9bWUFu/mxm2gU8WruOmoWs5tvkdslp3QQPQWQrHnB5U+secHlw6eSS38YukACUCoa6pjQ9qGvmgqpF1NbtZV93IuurdbKxtorWjk2HUMTNtHWdkb+AfctZybP5qMjr2QCtQMAmmXRhW/qcHN1cN4s61RGR/SgQpoqm1na11zayr3lvRr6sJxrG9ZA5L2805RVu4JudDppd+wLjm98jfszVcmwZFM+CYK4JKf/xpuhpG5CigRHAEc3caW9rZXt9CVUMz1Q0tVIXTXcuqGlqorm+hoWXfvtSHF2QxvcT4h/FbmJW+jgkt7zGsbiWZ9RthD8FQMgmOPQPGnARj5sDoWeqoS+QopEQwyO1p7eC97Q2s3FLPmqqG7oq+Kqz04z1jNiczjdLCHEoLs5kyqpCzJo+gtDCTaR3vcVzb+5Q2rCRr+1tQtWbvi4aOh7I5MGZhWOnPDrodFpGjnhLBIFLV0MyqrUGlv3JrPSu31LG+ZjddD63KzUxnVFFQwc8qG0ppYTalhdmMHBIsKx2SzYjCHIbkZOzbFXJdJTx+DWx4KZgvHBNU9rMWBOMxJ0L+8OS/YREZFFInEexYFww5xZA7NPi2m1M0IFe0dHQ662saWblPpV/ffdctwNihuUwbM4S/mzWGaaOHMH3MEMqKcw+9r/t3fgNP3BA8c/VTP4Kpn1a7vojsI3USwcrfwTO37b88uwhyi8LEMDQmSfScLt77mLq0zODRc2np4XwGHZZBYys0tDp1rU5di1Pf4tQ3t1G/p4365naqG5pZubWB97bV09wW9JyZmW5MLi1k7gkjmDZ6CFNHD2Ha6CEU5fXxJqvmOnjyJljxaPCAkkuWBG3+IiI9pE4iOPHy4Nr2PTuDZ6fu2Rk8z7XndNWqYHrPTuhsO/h2Q+lAUTiUhcs63WgnjQ7SaSOdOhvC80WfZfNHFnD82FKmjRnCsSMKyMro5w7UNr4C//0PQffGc78JH/sndW4mIr1KndqhoDQYEuUObU1sr9rGQ8++yfLV6ynJ2ENxThr5mVCYBQUZkJ8J+ZlOXgbkZTi56Z3kpkNOupOT3klOWifZaZ0UWCdDtq3gSxt/Cu//FkpvgOFXQH8mgfZW+POd8PJPgueuXvU0jPtI/21fRI5KkSYCM5sH3EXwhfnn7n5nj/U/AT4ezuYBpe4+NMqYEtXU1sFP/7yZJS9+QKfn8eWzL+RrHz+Oguw+HrL1L8Gfvw9PfSOosD96A5x8Rd8flF39Pvz3V2DrmzDnizDvX3Wpp4gkJLIH05hZOvA+cD5QCSwDLnP3lb2UvxaY4+5XHWi7UT+YprPT+e83NvPDp1ezvb6FC2eN5hvzpjCupJ+fS7v+JXjhTtj4cvBA7cNNCO5Q8R/w9LeD1150d3BCWEQkxkA9mOYUYK27rwuDeBS4GIibCIDLgFsjjOeg/rZ+B3f835WsqKxjdlkR9/z9SZRPKIlmZxM/FgxdCeFwfiE0VsHvFsOap+HYc+Ez/64rgkTkkEWZCMYSdDzcpRI4NV5BMzsGmAg818v6RcAigPHjx/dvlAQPT7nzqVU8+fY2Rhfl8JMFs7l49ljS+vps20TskxD+dW9C+NiNcNIVkJkT/3Xv/SFIAq2NcMEP4JRF6uNHRA7LYDlZfCnwmLvvf5ss4O5LgCUQNA31104bmtv4t+fX8sDLG0hPM24473gWnTWJ3KwB6C2zZ0L4w03w0o/3Twitu+Hpb8HyB2DkTPjcfVA6NfnxishRI8pEsBkYFzNfFi6L51Lg6xHGso/2jk5+WbGJH//xfWp3t/K5k8r450+ewKiiXr59J1O8hNDVZDRqFixdDLUfwJnXwce/pb79RaTPokwEy4DJZjaRIAFcCvx9z0JmNgUoBl6NMJZuL62p5o4nVvHe9gZOmVDCgwunMbOsKBm7PjQTPwYTPhp0C/HCnUFCABhSBlf8PlgvItIPIksE7t5uZouBpwkuH73f3d81s9uBCndfGha9FHjUo7p8KbS2qpHvPbmK51ZXMa4kl3svP4l5M0YdepcNyWQGE8+CCR8LEsLm14MTyeoMTkT6UWSXj0blcC8fve/Fddz17BoWn3McV54xgZxMPTVLRFLHQF0+OqhcccYEPnvSWIYXqE1dRCTWQfs3MLNrzeyIb4vIykhTEhARiSORjm5GAsvM7FdmNs8GdaO6iIgcqoMmAnf/NjAZ+A/gSmCNmX3PzI6NODYREUmChLq+DK/o2RYO7QSXez5mZj+IMDYREUmCg54sNrPrgC8BNcDPgX929zYzSwPWADdFG6KIiEQpkauGSoBL3H1j7EJ37zSzC6MJS0REkiWRpqE/ADu6ZsxsiJmdCuDuq6IKTEREkiORRHAv0Bgz3xguExGRo0AiicBiu39w905S6EY0EZGjXSKJYJ2Z/aOZZYbDdcC6qAMTEZHkSCQRXAOcQdCDaNfDZRZFGZSIiCTPQZt43L2KoIdQERE5CiVyH0EOcDUwHeh+csvBHjIvIiJHhkSahv4TGAV8EvgzwZPGGqIMSkREkieRRHCcu38H2O3uDwF/Ry8PoRcRkSNPIomgLRzvMrMZQBFQGl1IIiKSTIncD7AkfB7Bt4GlQAHwnUijEhGRpDngL4KwY7l6d9/p7i+6+yR3L3X3nyWy8fD5Be+Z2Vozu7mXMp83s5Vm9q6ZPXIY70FERPrggIkgvIv4sHoXNbN04B7gAmAacJmZTetRZjLwTeBMd58OXH84+xIRkcOXyDmCZ8zsn8xsnJmVdA0JvO4UYK27r3P3VuBR4OIeZb4C3OPuO6H7ngUREUmiRM4RLAjHX49Z5sCkg7xuLLApZr7rruRYxwOY2V+AdOA2d3+q54bMbBHh3czjx49PIGQREUlUIncWT4x4/5OBuQT3J7xoZjPdfVePGJYASwDKy8u950ZEROTwJXJn8ZfiLXf3hw/y0s3AuJj5snBZrErgr+7eBqw3s/cJEsOyg8UlIiL9I5GmoY/ETOcA5wKvAwdLBMuAyWY2kSABXAr8fY8yvwUuAx4ws+EETUXq2VREJIkSaRq6NnbezIYSnPg92OvazWwx8DRB+//97v6umd0OVLj70nDdJ8xsJdBB8Dzk2sN4HyIicpgs5pkzib3ALBN4x91PiCakAysvL/eKioqB2LWIyBHLzJa7e3m8dYmcI/g9wVVCEFxuOg34Vf+FJyIiAymRcwQ/ipluBza6e2VE8YiISJIlkgg+BLa6ezOAmeWa2QR33xBpZCIikhSJ3Fn8a6AzZr4jXCYiIkeBRBJBRthFBADhdFZ0IYmISDIlkgiqzeyirhkzuxioiS4kERFJpkTOEVwD/MLM/i2crwTi3m0sIiJHnkRuKPsAOM3MCsL5xsijEhGRpDlo05CZfc/Mhrp7o7s3mlmxmd2RjOBERCR6iZwjuCC2N9Dw2QGfii4kERFJpkQSQbqZZXfNmFkukH2A8iIicgRJ5GTxL4BnzewBwIArgYeiDEpERJInkZPF3zezt4DzCPoceho4JurAREQkORJpGgLYTpAE5gPnAKsii0hERJKq118EZnY8wUNjLiO4geyXBN1WfzxJsYmISBIcqGloNfAScKG7rwUwsxuSEpWIiCTNgZqGLgG2As+b2X1mdi7ByWIRETmK9JoI3P237n4pMAV4HrgeKDWze83sE8kKUEREonXQk8XuvtvdH3H3TwNlwBvANxLZuJnNM7P3zGytmd0cZ/2VZlZtZm+Gw5cP+R2IiEifJHIfQbfwruIl4XBAZpYO3AOcT9BR3TIzW+ruK3sU/aW7Lz6UOEREpP8kevno4TgFWOvu68JnGDwKXBzh/kRE5DBEmQjGApti5ivDZT19zsxWmNljZjYu3obMbJGZVZhZRXV1dRSxioikrCgTQSJ+D0xw91nAn+il6wp3X+Lu5e5ePmLEiKQGKCJytIsyEWwGYr/hl4XLurl7rbu3hLM/B06OMB4REYkjykSwDJhsZhPNLAu4FFgaW8DMRsfMXoS6rhARSbpDumroULh7u5ktJuikLh24393fNbPbgQp3Xwr8Y/g85HZgB0HPpiIikkTm7gMdwyEpLy/3ioqKgQ5DROSIYmbL3b083rqBPlksIiIDTIlARCTFKRGIiKQ4JQIRkRSnRCAikuKUCEREUpwSgYhIilMiEBFJcUoEIiIpTolARCTFKRGIiKQ4JQIRkRSnRCAikuKUCEREUpwSgYhIilMiEBFJcUoEIiIpTolARCTFRZoIzGyemb1nZmvN7OYDlPucmbmZxX2MmoiIRCeyRGBm6cA9wAXANOAyM5sWp1whcB3w16hiERGR3kX5i+AUYK27r3P3VuBR4OI45f4F+D7QHGEsIiLSiygTwVhgU8x8Zbism5mdBIxz9/97oA2Z2SIzqzCziurq6v6PVEQkhQ3YyWIzSwN+DPzPg5V19yXuXu7u5SNGjIg+OBGRFBJlItgMjIuZLwuXdSkEZgAvmNkG4DRgqU4Yi4gkV5SJYBkw2cwmmlkWcCmwtGulu9e5+3B3n+DuE4DXgIvcvSLCmEREpIfIEoG7twOLgaeBVcCv3P1dM7vdzC6Kar8iInJoMqLcuLs/CTzZY9l3eyk7N8pYREQkPt1ZLCKS4pQIRERSnBKBiEiKUyIQEUlxSgQiIilOiUBEJMUpEYiIpDglAhGRFKdEICKS4pQIRERSXKRdTIjI0a+trY3Kykqam/VsqcEgJyeHsrIyMjMzE36NEoGI9EllZSWFhYVMmDABMxvocFKau1NbW0tlZSUTJ05M+HVqGhKRPmlubmbYsGFKAoOAmTFs2LBD/nWmRCAifaYkMHgczt9CiUBEJMUpEYiIpDglAhGRBLW3tw90CJHQVUMi0m/+n9+/y8ot9f26zWljhnDrp6cftNxnPvMZNm3aRHNzM9dddx2LFi3iqaee4pZbbqGjo4Phw4fz7LPP0tjYyLXXXktFRQVmxq233srnPvc5CgoKaGxsBOCxxx7jiSee4MEHH+TKK68kJyeHN954gzPPPJNLL72U6667jubmZnJzc3nggQc44YQT6Ojo4Bvf+AZPPfUUaWlpfOUrX2H69Oncfffd/Pa3vwXgT3/6E//+7//O448/3q/HqK8iTQRmNg+4C0gHfu7ud/ZYfw3wdaADaAQWufvKKGMSkaPT/fffT0lJCXv27OEjH/kIF198MV/5yld48cUXmThxIjt27ADgX/7lXygqKuLtt98GYOfOnQfddmVlJa+88grp6enU19fz0ksvkZGRwTPPPMMtt9zCb37zG5YsWcKGDRt48803ycjIYMeOHRQXF/O1r32N6upqRowYwQMPPMBVV10V6XE4HJElAjNLB+4BzgcqgWVmtrRHRf+Iu/80LH8R8GNgXlQxiUi0EvnmHpW77767+5v2pk2bWLJkCWeddVb39fQlJSUAPPPMMzz66KPdrysuLj7otufPn096ejoAdXV1XHHFFaxZswYzo62trXu711xzDRkZGfvs74tf/CL/9V//xcKFC3n11Vd5+OGH++kd958ofxGcAqx193UAZvYocDHQnQjcPfY3ZD7gEcYjIkepF154gWeeeYZXX32VvLw85s6dy4knnsjq1asT3kbsZZc9r8PPz8/vnv7Od77Dxz/+cR5//HE2bNjA3LlzD7jdhQsX8ulPf5qcnBzmz5/fnSgGkyhPFo8FNsXMV4bL9mFmXzezD4AfAP8YYTwicpSqq6ujuLiYvLw8Vq9ezWuvvUZzczMvvvgi69evB+huGjr//PO55557ul/b1TQ0cuRIVq1aRWdn5wHb8Ovq6hg7NqjKHnzwwe7l559/Pj/72c+6Tyh37W/MmDGMGTOGO+64g4ULF/bfm+5HA37VkLvf4+7HAt8Avh2vjJktMrMKM6uorq5OboAiMujNmzeP9vZ2pk6dys0338xpp53GiBEjWLJkCVhNBasAAAzASURBVJdccgmzZ89mwYIFAHz7299m586dzJgxg9mzZ/P8888DcOedd3LhhRdyxhlnMHr06F73ddNNN/HNb36TOXPm7HMV0Ze//GXGjx/PrFmzmD17No888kj3ussvv5xx48YxderUiI5A35h7NK0xZnY6cJu7fzKc/yaAu/9rL+XTgJ3uXnSg7ZaXl3tFRUV/hysih2nVqlWDtoIbLBYvXsycOXO4+uqrk7K/eH8TM1vu7uXxykf5i2AZMNnMJppZFnApsLRHYJNjZv8OWBNhPCIiSXfyySezYsUKvvCFLwx0KL2K7KyFu7eb2WLgaYLLR+9393fN7Hagwt2XAovN7DygDdgJXBFVPCIiA2H58uUDHcJBRXr62t2fBJ7ssey7MdPXRbl/ERE5uAE/WSwiIgNLiUBEJMUpEYiIpDglAhGRFKdEICIppaCgYKBDGHQGX6cXInLk+sPNsO3t/t3mqJlwwZ0HL3eEaW9vHzT9DukXgYgc0W6++eZ9+g667bbbuOOOOzj33HM56aSTmDlzJr/73e8S2lZjY2Ovr3v44Ye7u4/44he/CMD27dv57Gc/y+zZs5k9ezavvPIKGzZsYMaMGd2v+9GPfsRtt90GwNy5c7n++uspLy/nrrvu4ve//z2nnnoqc+bM4bzzzmP79u3dcSxcuJCZM2cya9YsfvOb33D//fdz/fXXd2/3vvvu44Ybbjjs47YPdz+ihpNPPtlFZPBYuXLlgO7/9ddf97POOqt7furUqf7hhx96XV2du7tXV1f7scce652dne7unp+f3+u22tra4r7unXfe8cmTJ3t1dbW7u9fW1rq7++c//3n/yU9+4u7u7e3tvmvXLl+/fr1Pnz69e5s//OEP/dZbb3V397PPPtu/+tWvdq/bsWNHd1z33Xef33jjje7uftNNN/l11123T7mGhgafNGmSt7a2urv76aef7itWrIj7PuL9TQhu5I1brw6O3yUiIodpzpw5VFVVsWXLFqqrqykuLmbUqFHccMMNvPjii6SlpbF582a2b9/OqFGjDrgtd+eWW27Z73XPPfcc8+fPZ/jw4cDeZw0899xz3c8XSE9Pp6io6KAPuunq/A6CB94sWLCArVu30tra2v3shN6emXDOOefwxBNPMHXqVNra2pg5c+YhHq34lAhE5Ig3f/58HnvsMbZt28aCBQv4xS9+QXV1NcuXLyczM5MJEybs94yBeA73dbEyMjLo7Ozsnj/Qsw2uvfZabrzxRi666CJeeOGF7iak3nz5y1/me9/7HlOmTOnXLq11jkBEjngLFizg0Ucf5bHHHmP+/PnU1dVRWlpKZmYmzz//PBs3bkxoO7297pxzzuHXv/41tbW1wN5nDZx77rnce++9AHR0dFBXV8fIkSOpqqqitraWlpYWnnjiiQPur+vZBg899FD38t6emXDqqaeyadMmHnnkES677LJED89BKRGIyBFv+vTpNDQ0MHbsWEaPHs3ll19ORUUFM2fO5OGHH2bKlCkJbae3102fPp1vfetbnH322cyePZsbb7wRgLvuuovnn3+emTNncvLJJ7Ny5UoyMzP57ne/yymnnML5559/wH3fdtttzJ8/n5NPPrm72Ql6f2YCwOc//3nOPPPMhB6xmajInkcQFT2PQGRw0fMIkuvCCy/khhtu4Nxzz+21zGB6HoGIiPSTXbt2cfzxx5Obm3vAJHA4dLJYRFLO22+/3X0vQJfs7Gz++te/DlBEBzd06FDef//9SLatRCAifebumNlAh5GwmTNn8uabbw50GJE4nOZ+NQ2JSJ/k5ORQW1t7WBWQ9C93p7a2lpycnEN6nX4RiEiflJWVUVlZSXV19UCHIgSJuays7JBeo0QgIn2SmZnZfUesHJkibRoys3lm9p6ZrTWzm+Osv9HMVprZCjN71syOiTIeERHZX2SJwMzSgXuAC4BpwGVmNq1HsTeAcnefBTwG/CCqeEREJL4ofxGcAqx193Xu3go8ClwcW8Ddn3f3pnD2NeDQGrZERKTPojxHMBbYFDNfCZx6gPJXA3+It8LMFgGLwtlGM3vvMGMaDtQc5muTQfH1jeLru8Eeo+I7fL02vQ+Kk8Vm9gWgHDg73np3XwIs6Yf9VPR2i/VgoPj6RvH13WCPUfFFI8pEsBkYFzNfFi7bh5mdB3wLONvdWyKMR0RE4ojyHMEyYLKZTTSzLOBSYGlsATObA/wMuMjdqyKMRUREehFZInD3dmAx8DSwCviVu79rZreb2UVhsR8CBcCvzexNM1vay+b6S5+blyKm+PpG8fXdYI9R8UXgiOuGWkRE+pf6GhIRSXFKBCIiKe6oTAQJdG2RbWa/DNf/1cwmJDG2cWb2fNi1xrtmdl2cMnPNrC48b/KmmX03WfGF+99gZm+H+97vcXAWuDs8fivM7KQkxnZCzHF508zqzez6HmWSfvzM7H4zqzKzd2KWlZjZn8xsTTiO+2xBM7siLLPGzK5IUmw/NLPV4d/vcTMb2strD/hZiDjG28xsc8zf8VO9vPaA/+8RxvfLmNg2mFncfq2TdQz7xN2PqgFIBz4AJgFZwFvAtB5lvgb8NJy+FPhlEuMbDZwUThcC78eJby7wxAAeww3A8AOs/xTBzX8GnAb8dQD/1tuAYwb6+AFnAScB78Qs+wFwczh9M/D9OK8rAdaF4+JwujgJsX0CyAinvx8vtkQ+CxHHeBvwTwl8Bg74/x5VfD3W/y/guwN5DPsyHI2/CA7atUU4/1A4/RhwriXpqRruvtXdXw+nGwiuqBqbjH33o4uBhz3wGjDUzEYPQBznAh+4+8YB2Pc+3P1FYEePxbGfs4eAz8R56SeBP7n7DnffCfwJmBd1bO7+Rw+u7INB0L1LL8cvEYn8v/fZgeIL647PA/+nv/ebLEdjIojXtUXPira7TPjPUAcMS0p0McImqTlAvOfjnW5mb5nZH8xselIDAwf+aGbLw+49ekrkGCfDpfT+zzeQx6/LSHffGk5vA0bGKTMYjuVV9NK9Cwf/LERtcdh8dX8vTWuD4fh9DNju7mt6WT/Qx/CgjsZEcEQwswLgN8D17l7fY/XrBM0ds4H/Dfw2yeF91N1PIug59utmdlaS939Q4U2KFwG/jrN6oI/ffjxoIxh012qb2beAduAXvRQZyM/CvcCxwInAVoLml8HoMg78a2DQ/z8djYkgka4tusuYWQZQBNQmJbpgn5kESeAX7v7fPde7e727N4bTTwKZZjY8WfG5++ZwXAU8TvDzO1ZC3YdE7ALgdXff3nPFQB+/GNu7mszCcby75wfsWJrZlcCFwOVhotpPAp+FyLj7dnfvcPdO4L5e9j2gn8Ww/rgE+GVvZQbyGCbqaEwEB+3aIpzvujrjfwDP9faP0N/C9sT/AFa5+497KTOq65yFmZ1C8HdKSqIys3wzK+yaJjip+E6PYkuBL4VXD50G1MU0gSRLr9/CBvL49RD7ObsC+F2cMk8DnzCz4rDp4xPhskiZ2TzgJoLuXZp6KZPIZyHKGGPPO322l30n8v8epfOA1e5eGW/lQB/DhA302eooBoKrWt4nuJrgW+Gy2wk+9AA5BE0Ka4G/AZOSGNtHCZoIVgBvhsOngGuAa8Iyi4F3Ca6AeA04I4nxTQr3+1YYQ9fxi43PCB469AHwNsHDhZL5980nqNiLYpYN6PEjSEpbgTaCduqrCc47PQusAZ4BSsKy5cDPY157VfhZXAssTFJsawna1rs+g11X0Y0BnjzQZyGJx+8/w8/XCoLKfXTPGMP5/f7fkxFfuPzBrs9dTNkBOYZ9GdTFhIhIijsam4ZEROQQKBGIiKQ4JQIRkRSnRCAikuKUCEREUpwSgUgPZtbRo4fTfuvR0swmxPZgKTIYRPnwepEj1R53P3GggxBJFv0iEElQ2K/8D8K+5f9mZseFyyeY2XNh52jPmtn4cPnIsK//t8LhjHBT6WZ2nwXPo/ijmeUO2JsSQYlAJJ7cHk1DC2LW1bn7TODfgP8vXPa/gYfcfRZB5213h8vvBv7sQed3JxHcWQowGbjH3acDu4DPRfx+RA5IdxaL9GBmje5eEGf5BuAcd18Xdhy4zd2HmVkNQfcHbeHyre4+3MyqgTJ3b4nZxgSC5w9MDue/AWS6+x3RvzOR+PSLQOTQeC/Th6IlZroDnauTAaZEIHJoFsSMXw2nXyHo9RLgcuClcPpZ4KsAZpZuZkXJClLkUOibiMj+cns8iPwpd++6hLTYzFYQfKu/LFx2LfCAmf0zUA0sDJdfBywxs6sJvvl/laAHS5FBRecIRBIUniMod/eagY5FpD+paUhEJMXpF4GISIrTLwIRkRSnRCAikuKUCEREUpwSgYhIilMiEBFJcf8/5UKL21mb/rkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history_dict['accuracy'], label='accuracy')\n",
    "plt.plot(history_dict['val_accuracy'], label = 'val_accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.ylim([0.2, 1])\n",
    "plt.legend(loc='lower right')\n",
    "\n",
    "test_loss, test_acc = model.evaluate(test_images,  test_labels, verbose=2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Much less tendency to overfit here!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reasoning with uncertainty"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's find what this model sees as the area of greatest uncertainty in the input images. This don't be uncertainty due to the data itself, but rather the test images most affected by the probabilistic variance in our first convolutional layer's weights."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Start by predicting over the entire test set 10 times."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "yhats = [model.predict(test_images) for _ in range(10)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now collect like sample predictions together, so we have a 10 predicted values across 2 classes for all 2,000 images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_preds = [pd.DataFrame([yhat[i] for yhat in yhats]) for i in range(len(test_images))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2000"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(all_preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10, 2)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_preds[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.971859</td>\n",
       "      <td>0.028141</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.962656</td>\n",
       "      <td>0.037343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.971436</td>\n",
       "      <td>0.028564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.962926</td>\n",
       "      <td>0.037074</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.714335</td>\n",
       "      <td>0.285665</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.955292</td>\n",
       "      <td>0.044708</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.803589</td>\n",
       "      <td>0.196411</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.943876</td>\n",
       "      <td>0.056124</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.821010</td>\n",
       "      <td>0.178990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.922986</td>\n",
       "      <td>0.077014</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0         1\n",
       "0  0.971859  0.028141\n",
       "1  0.962656  0.037343\n",
       "2  0.971436  0.028564\n",
       "3  0.962926  0.037074\n",
       "4  0.714335  0.285665\n",
       "5  0.955292  0.044708\n",
       "6  0.803589  0.196411\n",
       "7  0.943876  0.056124\n",
       "8  0.821010  0.178990\n",
       "9  0.922986  0.077014"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_preds[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.090418\n",
       "1    0.090418\n",
       "dtype: float64"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_preds[0].std()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1631"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "most_uncertain = np.argmax([pred.std().mean() for pred in all_preds])\n",
    "most_uncertain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.133390</td>\n",
       "      <td>0.866610</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.078883</td>\n",
       "      <td>0.921117</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.931562</td>\n",
       "      <td>0.068438</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.375245</td>\n",
       "      <td>0.624756</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.735670</td>\n",
       "      <td>0.264330</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.151736</td>\n",
       "      <td>0.848264</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.691403</td>\n",
       "      <td>0.308597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.194610</td>\n",
       "      <td>0.805390</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.879628</td>\n",
       "      <td>0.120372</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.732313</td>\n",
       "      <td>0.267687</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0         1\n",
       "0  0.133390  0.866610\n",
       "1  0.078883  0.921117\n",
       "2  0.931562  0.068438\n",
       "3  0.375245  0.624756\n",
       "4  0.735670  0.264330\n",
       "5  0.151736  0.848264\n",
       "6  0.691403  0.308597\n",
       "7  0.194610  0.805390\n",
       "8  0.879628  0.120372\n",
       "9  0.732313  0.267687"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_preds[most_uncertain]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/zach_anglin/.pyenv/versions/3.7.3/lib/python3.7/site-packages/numpy-1.17.2-py3.7-macosx-10.14-x86_64.egg/numpy/core/fromnumeric.py:61: FutureWarning: \n",
      "The current behaviour of 'Series.argmax' is deprecated, use 'idxmax'\n",
      "instead.\n",
      "The behavior of 'argmax' will be corrected to return the positional\n",
      "maximum in the future. For now, use 'series.values.argmax' or\n",
      "'np.argmax(np.array(values))' to get the position of the maximum\n",
      "row.\n",
      "  return bound(*args, **kwds)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x216 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = most_uncertain\n",
    "\n",
    "\n",
    "plt.figure(figsize=(6,3))\n",
    "plt.subplot(1,2,1)\n",
    "plot_image(i, all_preds[i].mean(), test_labels, test_images)\n",
    "plt.subplot(1,2,2)\n",
    "plot_value_array(i, all_preds[i].mean(),  test_labels)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To us, this image may look like an airplane, so it's surprising that this is where the model is least certain. Remember, however that:\n",
    "- We only replaced the first convolutional layer with one that's probabilistic in nature\n",
    "- We're measuring `epistemic` uncertainty, so uncertainty caused by a lack of clarity in how the model should calibrate its own weights (at least for the first layer).\n",
    "\n",
    "To interpret this, recognize that our image here may have low-level features that affect the activations from the first convolutional layer in the network, and there's enough variation in those weights to cause the ultimate output (several layers later) to swing wildly from one predicted category to another."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test sensitivity to adversarial data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's again pull in the full CIFAR-10 data set in order to give our model an image from outside the distribution it was trained on, in order to see if this model responds more appropriately than the one measuring aleatoric uncertainty."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()\n",
    "\n",
    "# Normalize pixel values to be between 0 and 1\n",
    "train_images, test_images = train_images / 255.0, test_images / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x15b131cc0>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.imshow(test_images[1000], cmap=plt.cm.binary)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll predict over this image 1,000 times in order to get a distribution of predictions to work with."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "dog_preds = [model(test_images[1000:1001]) for _ in range(1000)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "dog_samples = pd.DataFrame(dog_pred.numpy().squeeze() for dog_pred in dog_preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.313834</td>\n",
       "      <td>0.686166</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.285485</td>\n",
       "      <td>0.714515</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.209072</td>\n",
       "      <td>0.790928</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.187556</td>\n",
       "      <td>0.812444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.359614</td>\n",
       "      <td>0.640386</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0         1\n",
       "0  0.313834  0.686166\n",
       "1  0.285485  0.714515\n",
       "2  0.209072  0.790928\n",
       "3  0.187556  0.812444\n",
       "4  0.359614  0.640386"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dog_samples.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.300368\n",
       "1    0.699632\n",
       "dtype: float64"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dog_samples.mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ok, so not completely uncertain, but a lot better than our aleatoric model! Remember, this epistemic estimator is indicating uncertainty that it could in theory reduce with additional data -- so images far outside the training set would certainly qualify."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's take the standard deviation of our samples in order to get a measure of how variant the predictions are."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.08771\n",
       "1    0.08771\n",
       "dtype: float64"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dog_samples.std()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we've got an estimate of the uncertainty, let's figure out where that estimate ranks among all the predictions of birds and planes in the test set. What proportion of test set predictions have less variance than this one? In other words, what is the percentile rank of the dog image's epistemic uncertainty?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.532"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(np.array([pred.std()[0] for pred in all_preds]).squeeze() < dog_samples.std()[0]).sum() / len(all_preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Alright! So by introducing a single Bayesian layer to our network to indicate epistemic uncertainty, we are able to indicate uncertainty on an instance outside our training set at a far better rate than if we strictly measure aleatoric uncertainty (or no uncertainty at all). We went from being as confident as we could possibly being in the upper half of uncertainty across all images in the test set."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Takeaways\n",
    "- While aleatoric uncertainty is great for identifying border cases, epistemic uncertainty over a model's weights allows us to accurately indicate when a model input is far outside the distribution the model was trained on.\n",
    "- Leveraging Bayesian priors over network weights makes the network much less prone to overfitting, even in models as complex as CNNs.\n",
    "- TensorFlow Probability contains lots of sorts of layers that can be used to fit distributions over the network's weights, including convolutional layers that are drop-in replacements for their point-estimate counterparts in TensorFlow `keras`."
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}