Hamoye Stage D - Amit Purswani.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hamoye Stage D - Amit Purswani"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\krane\\AppData\\Roaming\\Python\\Python38\\site-packages\\tensorflow\\python\\compat\\v2_compat.py:101: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "non-resource variables are not supported in the long term\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow.compat.v1 as tf\n",
    "tf.disable_v2_behavior()\n",
    "from tensorflow import keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training data: (60000, 28, 28), (60000,)\n",
      "Test data: (10000, 28, 28), (10000,)\n",
      "There are 10 classes in the dataset. They are: [0 1 2 3 4 5 6 7 8 9]\n"
     ]
    }
   ],
   "source": [
    "mnist = keras.datasets.mnist\n",
    "(train_images, train_labels), (test_images, test_labels) = mnist.load_data()\n",
    "\n",
    "print(\"Training data: {}, {}\".format(train_images.shape, train_labels.shape))\n",
    "# prints Training data: (60000, 28, 28), (60000,)\n",
    "print(\"Test data: {}, {}\".format(test_images.shape, test_labels.shape))\n",
    "# prints Test data: (10000, 28, 28), (10000,)\n",
    "class_labels = np.unique(train_labels)\n",
    "print(\"There are {} classes in the dataset. They are: {}\".format(len(class_labels), class_labels))\n",
    "# prints There are 10 classes in the dataset. They are: [0 1 2 3 4 5 6 7 8 9]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 5))\n",
    "\n",
    "plt.subplot(1,3,1)\n",
    "plt.imshow(train_images[0])\n",
    "plt.title(\"Label: {}\".format(train_labels[0]))\n",
    "\n",
    "plt.subplot(1,3,2)\n",
    "plt.imshow(train_images[2500])\n",
    "plt.title(\"Label: {}\".format(train_labels[2500]))\n",
    "\n",
    "plt.subplot(1,3,3)\n",
    "plt.imshow(test_images[12])\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_images = train_images / 255.0\n",
    "test_images = test_images / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train: (50000, 28, 28)\n",
      "x_val: (10000, 28, 28)\n",
      "y_train: (50000,)\n",
      "y_val: (10000,)\n"
     ]
    }
   ],
   "source": [
    "x_train = train_images[0:50000]\n",
    "x_val = train_images[50000:]\n",
    "y_train = train_labels[0:50000]\n",
    "y_val = train_labels[50000:]\n",
    "\n",
    "print(\"x_train: {}\".format(x_train.shape)) # prints x_train: (50000, 28, 28)\n",
    "print(\"x_val: {}\".format(x_val.shape)) # prints x_val: (10000, 28, 28)\n",
    "print(\"y_train: {}\".format(y_train.shape)) #prints y_train: (50000,)\n",
    "print(\"y_val: {}\".format(y_val.shape)) #prints y_val: (10000,)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train: (50000, 784)\n",
      "x_val: (10000, 784)\n",
      "test_images: (10000, 784)\n"
     ]
    }
   ],
   "source": [
    "new_dimension = np.prod(train_images.shape[1:])\n",
    "x_train = x_train.reshape(x_train.shape[0], new_dimension)\n",
    "x_val = x_val.reshape(x_val.shape[0], new_dimension)\n",
    "test_images = test_images.reshape(test_images.shape[0], new_dimension)\n",
    "\n",
    "print(\"x_train: {}\".format(x_train.shape)) #prints  x_train: (50000, 784)\n",
    "print(\"x_val: {}\".format(x_val.shape)) #prints x_val: (10000, 784)\n",
    "print(\"test_images: {}\".format(test_images.shape)) #prints test_images: (10000, 784)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.utils import to_categorical\n",
    "no_labels = 10\n",
    "y_train = to_categorical(y_train, no_labels)\n",
    "y_val = to_categorical(y_val, no_labels)\n",
    "y_test = to_categorical(test_labels, no_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# inputs\n",
    "#placeholders are not enabled in tf 2\n",
    "X = tf.placeholder(tf.float32, [None, new_dimension])\n",
    "Y = tf.placeholder(tf.float32, [None, no_labels])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create model architecture\n",
    "def multilayer_perceptron(x, no_classes, first_layer_neurons=256, second_layer_neurons=128):\n",
    "  # first layer\n",
    "  first_weight = tf.Variable(tf.random_uniform([new_dimension, first_layer_neurons]))\n",
    "  first_bias = tf.Variable(tf.zeros([first_layer_neurons]))\n",
    "  first_layer_output = tf.nn.relu(tf.add(tf.matmul(x, first_weight), first_bias))\n",
    "\n",
    "  # second layer\n",
    "  second_weight = tf.Variable(tf.random_uniform([first_layer_neurons,second_layer_neurons]))\n",
    "  second_bias = tf.Variable(tf.zeros([second_layer_neurons]))\n",
    "  second_layer_output = tf.nn.relu(tf.add(tf.matmul(first_layer_output, second_weight), \n",
    "                                          second_bias))\n",
    "\n",
    "  # output layer\n",
    "  final_weight = tf.Variable(tf.random_uniform([second_layer_neurons, no_classes]))\n",
    "  final_bias = tf.Variable(tf.zeros([no_classes]))\n",
    "  logits = tf.add(tf.matmul(second_layer_output, final_weight), final_bias)\n",
    "\n",
    "  return logits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\krane\\AppData\\Roaming\\Python\\Python38\\site-packages\\tensorflow\\python\\util\\dispatch.py:206: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "\n",
      "Future major versions of TensorFlow will allow gradients to flow\n",
      "into the labels input on backprop by default.\n",
      "\n",
      "See `tf.nn.softmax_cross_entropy_with_logits_v2`.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "logits = multilayer_perceptron(X, no_labels)\n",
    "\n",
    "learning_rate = 0.01\n",
    "#we define the loss and optimiser for the network\n",
    "loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=Y))\n",
    "optimiser = tf.train.AdamOptimizer(learning_rate=learning_rate)\n",
    "train_op = optimiser.minimize(loss_op)\n",
    "\n",
    "#initialise the variables\n",
    "init = tf.global_variables_initializer()\n",
    "\n",
    "epochs = 20\n",
    "batch_size = 1000\n",
    "iteration = len(x_train) // batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch========0\n",
      "Epoch========1\n",
      "Epoch========2\n",
      "Epoch========3\n",
      "Epoch========4\n",
      "Epoch========5\n",
      "Epoch========6\n",
      "Epoch========7\n",
      "Epoch========8\n",
      "Epoch========9\n",
      "Epoch========10\n",
      "Epoch========11\n",
      "Epoch========12\n",
      "Epoch========13\n",
      "Epoch========14\n",
      "Epoch========15\n",
      "Epoch========16\n",
      "Epoch========17\n",
      "Epoch========18\n",
      "Epoch========19\n",
      "Accuracy: 0.9169999957084656\n"
     ]
    }
   ],
   "source": [
    "#train model\n",
    "with tf.Session() as session:\n",
    "  session.run(init)\n",
    "  for epoch in range(epochs):\n",
    "    average_cost = 0\n",
    "    start, end = 0, batch_size\n",
    "\n",
    "    for i in range(iteration):\n",
    "      batch_x, batch_y = x_train[start: end], y_train[start: end]\n",
    "      _, loss = session.run([train_op, loss_op], feed_dict={X: batch_x, Y: batch_y})\n",
    "      start += batch_size\n",
    "      end += batch_size \n",
    "      #average loss\n",
    "      average_cost += loss/iteration\n",
    "    print(\"Epoch========{}\".format(epoch))\n",
    "    \n",
    "  #evaluate model\n",
    "  prediction = tf.nn.softmax(logits)\n",
    "  ground_truth = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))\n",
    "  accuracy = tf.reduce_mean(tf.cast(ground_truth, \"float\"))\n",
    "  print(\"Accuracy: {}\".format(accuracy.eval({X: test_images, Y: y_test})))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Building a Sequential Feed Forward Network in Keras\n",
    "\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.models import Sequential\n",
    "#Build the model object\n",
    "model = Sequential()\n",
    "# Build the input and the hidden layers\n",
    "model.add(Dense(256, activation='relu', input_shape=(new_dimension,)))\n",
    "model.add(Dense(128, activation='relu'))\n",
    "# Build the output layer\n",
    "model.add(Dense(no_labels, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss=tf.keras.losses.categorical_crossentropy, \n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "49000/50000 [============================>.] - ETA: 0s - loss: 0.7042 - acc: 0.8109"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\krane\\AppData\\Roaming\\Python\\Python38\\site-packages\\keras\\engine\\training.py:2470: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.\n",
      "  warnings.warn('`Model.state_updates` will be removed in a future version. '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "50000/50000 [==============================] - 3s 61us/sample - loss: 0.6956 - acc: 0.8131 - val_loss: 0.2625 - val_acc: 0.9241\n",
      "Epoch 2/20\n",
      "50000/50000 [==============================] - 2s 31us/sample - loss: 0.2378 - acc: 0.9322 - val_loss: 0.1868 - val_acc: 0.9477\n",
      "Epoch 3/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.1731 - acc: 0.9503 - val_loss: 0.1510 - val_acc: 0.9573\n",
      "Epoch 4/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.1352 - acc: 0.9614 - val_loss: 0.1286 - val_acc: 0.9633\n",
      "Epoch 5/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.1096 - acc: 0.9684 - val_loss: 0.1141 - val_acc: 0.9675\n",
      "Epoch 6/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0917 - acc: 0.9743 - val_loss: 0.1038 - val_acc: 0.9682\n",
      "Epoch 7/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.0770 - acc: 0.9785 - val_loss: 0.0941 - val_acc: 0.9713\n",
      "Epoch 8/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.0658 - acc: 0.9815 - val_loss: 0.0881 - val_acc: 0.9736\n",
      "Epoch 9/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0546 - acc: 0.9853 - val_loss: 0.0883 - val_acc: 0.9738\n",
      "Epoch 10/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0477 - acc: 0.9874 - val_loss: 0.0886 - val_acc: 0.9731\n",
      "Epoch 11/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.0421 - acc: 0.9889 - val_loss: 0.0804 - val_acc: 0.9761\n",
      "Epoch 12/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0370 - acc: 0.9903 - val_loss: 0.0828 - val_acc: 0.9754\n",
      "Epoch 13/20\n",
      "50000/50000 [==============================] - 1s 22us/sample - loss: 0.0302 - acc: 0.9927 - val_loss: 0.0810 - val_acc: 0.9759\n",
      "Epoch 14/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.0264 - acc: 0.9938 - val_loss: 0.0786 - val_acc: 0.9772\n",
      "Epoch 15/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.0223 - acc: 0.9949 - val_loss: 0.0784 - val_acc: 0.9776\n",
      "Epoch 16/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0189 - acc: 0.9960 - val_loss: 0.0769 - val_acc: 0.9770\n",
      "Epoch 17/20\n",
      "50000/50000 [==============================] - 1s 21us/sample - loss: 0.0168 - acc: 0.9966 - val_loss: 0.0783 - val_acc: 0.9772\n",
      "Epoch 18/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0150 - acc: 0.9972 - val_loss: 0.0800 - val_acc: 0.9770\n",
      "Epoch 19/20\n",
      "50000/50000 [==============================] - 1s 20us/sample - loss: 0.0122 - acc: 0.9980 - val_loss: 0.0819 - val_acc: 0.9765\n",
      "Epoch 20/20\n",
      "50000/50000 [==============================] - 1s 19us/sample - loss: 0.0102 - acc: 0.9985 - val_loss: 0.0785 - val_acc: 0.9783\n",
      "Test loss: 0.0743894427414576\n",
      "Test accuracy: 0.9790999889373779\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(x_train, y_train, validation_data=(x_val, y_val), epochs=20, batch_size=1000)\n",
    "\n",
    "test_loss, test_accuracy = model.evaluate(test_images, y_test)\n",
    "print('Test loss: {}'.format(test_loss))\n",
    "print('Test accuracy: {}'.format(test_accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Loss Curves - before regularisation')"
      ]
     },
     "execution_count": 21,
     "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": [
    "\n",
    "#Check for Overfitting\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(history.history['loss'], 'blue')\n",
    "plt.plot(history.history['val_loss'], 'red')\n",
    "plt.legend(['Training loss', 'Validation Loss'])\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Loss Curves - before regularisation')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.9551 - acc: 0.6950 - val_loss: 0.2894 - val_acc: 0.9192\n",
      "Epoch 2/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.3811 - acc: 0.8866 - val_loss: 0.2060 - val_acc: 0.9410\n",
      "Epoch 3/20\n",
      "50000/50000 [==============================] - 1s 29us/sample - loss: 0.2856 - acc: 0.9159 - val_loss: 0.1658 - val_acc: 0.9526\n",
      "Epoch 4/20\n",
      "50000/50000 [==============================] - 1s 23us/sample - loss: 0.2358 - acc: 0.9309 - val_loss: 0.1402 - val_acc: 0.9588\n",
      "Epoch 5/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.1994 - acc: 0.9408 - val_loss: 0.1229 - val_acc: 0.9645\n",
      "Epoch 6/20\n",
      "50000/50000 [==============================] - 1s 24us/sample - loss: 0.1784 - acc: 0.9466 - val_loss: 0.1100 - val_acc: 0.9685\n",
      "Epoch 7/20\n",
      "50000/50000 [==============================] - 1s 24us/sample - loss: 0.1542 - acc: 0.9546 - val_loss: 0.1026 - val_acc: 0.9700\n",
      "Epoch 8/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.1409 - acc: 0.9586 - val_loss: 0.0951 - val_acc: 0.9726\n",
      "Epoch 9/20\n",
      "50000/50000 [==============================] - 1s 27us/sample - loss: 0.1316 - acc: 0.9604 - val_loss: 0.0916 - val_acc: 0.9728\n",
      "Epoch 10/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.1190 - acc: 0.9653 - val_loss: 0.0868 - val_acc: 0.9744\n",
      "Epoch 11/20\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.1110 - acc: 0.9664 - val_loss: 0.0805 - val_acc: 0.9755\n",
      "Epoch 12/20\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.1036 - acc: 0.9680 - val_loss: 0.0814 - val_acc: 0.9753\n",
      "Epoch 13/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.0988 - acc: 0.9703 - val_loss: 0.0784 - val_acc: 0.9768\n",
      "Epoch 14/20\n",
      "50000/50000 [==============================] - 1s 28us/sample - loss: 0.0902 - acc: 0.9720 - val_loss: 0.0761 - val_acc: 0.9778\n",
      "Epoch 15/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.0839 - acc: 0.9748 - val_loss: 0.0737 - val_acc: 0.9785\n",
      "Epoch 16/20\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.0793 - acc: 0.9757 - val_loss: 0.0719 - val_acc: 0.9787\n",
      "Epoch 17/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.0782 - acc: 0.9758 - val_loss: 0.0733 - val_acc: 0.9784\n",
      "Epoch 18/20\n",
      "50000/50000 [==============================] - 1s 25us/sample - loss: 0.0709 - acc: 0.9779 - val_loss: 0.0708 - val_acc: 0.9776\n",
      "Epoch 19/20\n",
      "50000/50000 [==============================] - 1s 26us/sample - loss: 0.0697 - acc: 0.9780 - val_loss: 0.0748 - val_acc: 0.9798\n",
      "Epoch 20/20\n",
      "50000/50000 [==============================] - 1s 27us/sample - loss: 0.0676 - acc: 0.9795 - val_loss: 0.0720 - val_acc: 0.9783\n",
      "Test loss: 0.07113996025538072\n",
      "Test accuracy: 0.9799000024795532\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.layers import Dropout\n",
    "reg_model = Sequential()\n",
    "reg_model.add(Dense(256, activation='relu', input_shape=(new_dimension,)))\n",
    "reg_model.add(Dropout(0.4))\n",
    "reg_model.add(Dense(128, activation='relu'))\n",
    "reg_model.add(Dropout(0.4))\n",
    "reg_model.add(Dense(no_labels, activation='softmax'))\n",
    "\n",
    "reg_model.compile(optimizer='adam', loss=tf.keras.losses.categorical_crossentropy, \n",
    "              metrics=['accuracy'])\n",
    "\n",
    "reg_history = reg_model.fit(x_train, y_train, validation_data=(x_val, y_val), \n",
    "                            epochs=20, batch_size=1000)\n",
    "test_loss, test_accuracy = reg_model.evaluate(test_images, y_test)\n",
    "print('Test loss: {}'.format(test_loss))\n",
    "print('Test accuracy: {}'.format(test_accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.07113996025538072\n",
      "Test accuracy: 0.9799000024795532\n"
     ]
    }
   ],
   "source": [
    "\n",
    "test_loss, test_accuracy = reg_model.evaluate(test_images, y_test)\n",
    "print('Test loss: {}'.format(test_loss))\n",
    "print('Test accuracy: {}'.format(test_accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Loss Curves - after regularisation')"
      ]
     },
     "execution_count": 24,
     "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": [
    "\n",
    "plt.figure()\n",
    "plt.plot(reg_history.history['loss'], 'blue')\n",
    "plt.plot(reg_history.history['val_loss'], 'red')\n",
    "plt.legend(['Training loss', 'Validation Loss'])\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Loss Curves - after regularisation')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9\n"
     ]
    }
   ],
   "source": [
    "predictions = model.predict([test_images]) # Make prediction\n",
    "print(np.argmax(predictions[1000])) # Print out 9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Let's see whether the image would turn out to be 9\n",
    "# Remember that the input data to the network were flattened for the purpose of learning to a vector with a length of 784. Before displaying it, we need to convert it back to the size of 28 x 28 pixels.\n",
    "plt.figure(figsize=(8, 5))\n",
    "\n",
    "plt.subplot(1,3,1)\n",
    "plt.imshow(test_images[1000].reshape(28,28), cmap=\"gray\") # Import the image\n",
    "plt.show() # Show the image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    }
   ],
   "source": [
    "# You can try out others and play around with them\n",
    "predictions = model.predict([test_images]) # Make prediction\n",
    "print(np.argmax(predictions[6700])) # Print out the number"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 5))\n",
    "\n",
    "plt.subplot(1,3,1)\n",
    "plt.imshow(test_images[6700].reshape(28,28), cmap=\"gray\") # Import the image\n",
    "plt.show() # Show the image"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# End of Notebook"
   ]
  },
  {
   "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}