cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import Sequential, layers\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense\n",
    "from tensorflow.keras.utils import to_categorical"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28) (60000,) (10000, 28, 28) (10000,) \n",
      "\n",
      "(10000, 28, 28) (10000,) (10000, 28, 28) (10000,)\n"
     ]
    }
   ],
   "source": [
    "(train_img, train_lb), (test_img, test_lb) = tf.keras.datasets.mnist.load_data()\n",
    "\n",
    "print(train_img.shape, train_lb.shape, test_img.shape, test_lb.shape, '\\n')\n",
    "\n",
    "# set random seed\n",
    "np.random.seed(42)\n",
    "trainsize = 10000  # number of training images we use, selected randomly\n",
    "randtrain = np.random.choice(train_img.shape[0], trainsize, replace=False)\n",
    "X_train, y_train = train_img[randtrain], train_lb[randtrain]\n",
    "\n",
    "np.random.seed(10)\n",
    "testsize = 10000  # number of testing images we use, selected randomly\n",
    "randtest = np.random.choice(test_img.shape[0], testsize, replace=False)\n",
    "X_test, y_test = test_img[randtest], test_lb[randtest]\n",
    "\n",
    "print(X_train.shape, y_train.shape, X_test.shape, y_test.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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": [
    "# demonstrate an image\n",
    "plt.figure()\n",
    "plt.imshow(X_train[0])\n",
    "plt.imshow(X_train[0], cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# build the cnn model\n",
    "model = Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)))\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.Flatten())\n",
    "model.add(layers.Dense(128))\n",
    "model.add(layers.Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_2 (Conv2D)            (None, 26, 26, 32)        320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 13, 13, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 11, 11, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 5, 5, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 1600)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               204928    \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 225,034\n",
      "Trainable params: 225,034\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n",
      "100/100 [==============================] - 4s 44ms/step - loss: 2.0562 - accuracy: 0.4442 - val_loss: 1.6129 - val_accuracy: 0.7036\n",
      "Epoch 2/15\n",
      "100/100 [==============================] - 5s 45ms/step - loss: 1.0470 - accuracy: 0.7767 - val_loss: 0.6146 - val_accuracy: 0.8504\n",
      "Epoch 3/15\n",
      "100/100 [==============================] - 6s 57ms/step - loss: 0.5198 - accuracy: 0.8592 - val_loss: 0.4059 - val_accuracy: 0.8856\n",
      "Epoch 4/15\n",
      "100/100 [==============================] - 6s 60ms/step - loss: 0.3868 - accuracy: 0.8879 - val_loss: 0.3282 - val_accuracy: 0.9055\n",
      "Epoch 5/15\n",
      "100/100 [==============================] - 6s 56ms/step - loss: 0.3221 - accuracy: 0.9051 - val_loss: 0.2808 - val_accuracy: 0.9206\n",
      "Epoch 6/15\n",
      "100/100 [==============================] - 5s 54ms/step - loss: 0.2820 - accuracy: 0.9197 - val_loss: 0.2549 - val_accuracy: 0.9273\n",
      "Epoch 7/15\n",
      "100/100 [==============================] - 6s 56ms/step - loss: 0.2518 - accuracy: 0.9275 - val_loss: 0.2278 - val_accuracy: 0.9335\n",
      "Epoch 8/15\n",
      "100/100 [==============================] - 6s 55ms/step - loss: 0.2278 - accuracy: 0.9345 - val_loss: 0.2125 - val_accuracy: 0.9382\n",
      "Epoch 9/15\n",
      "100/100 [==============================] - 6s 55ms/step - loss: 0.2090 - accuracy: 0.9415 - val_loss: 0.1959 - val_accuracy: 0.9448\n",
      "Epoch 10/15\n",
      "100/100 [==============================] - 6s 59ms/step - loss: 0.1899 - accuracy: 0.9465 - val_loss: 0.1832 - val_accuracy: 0.9476\n",
      "Epoch 11/15\n",
      "100/100 [==============================] - 6s 56ms/step - loss: 0.1760 - accuracy: 0.9506 - val_loss: 0.1670 - val_accuracy: 0.9534\n",
      "Epoch 12/15\n",
      "100/100 [==============================] - 5s 54ms/step - loss: 0.1640 - accuracy: 0.9540 - val_loss: 0.1634 - val_accuracy: 0.9544\n",
      "Epoch 13/15\n",
      "100/100 [==============================] - 5s 55ms/step - loss: 0.1528 - accuracy: 0.9577 - val_loss: 0.1505 - val_accuracy: 0.9567\n",
      "Epoch 14/15\n",
      "100/100 [==============================] - 6s 59ms/step - loss: 0.1424 - accuracy: 0.9603 - val_loss: 0.1443 - val_accuracy: 0.9565\n",
      "Epoch 15/15\n",
      "100/100 [==============================] - 6s 55ms/step - loss: 0.1352 - accuracy: 0.9624 - val_loss: 0.1358 - val_accuracy: 0.9597\n"
     ]
    }
   ],
   "source": [
    "train_images = X_train.reshape((trainsize, 28, 28, 1))\n",
    "train_images = (train_images.astype('float32') - 127.5) / 127.5\n",
    "\n",
    "test_images = X_test.reshape((testsize, 28, 28, 1))\n",
    "test_images = (test_images.astype('float32') - 127.5) / 127.5\n",
    "\n",
    "train_labels = to_categorical(y_train)\n",
    "test_labels = to_categorical(y_test)\n",
    "\n",
    "model.compile(optimizer='sgd',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_images, train_labels, batch_size=100, epochs=15,\n",
    "                    validation_data=(test_images, test_labels))\n",
    "\n",
    "# test_loss, test_acc = model.evaluate(test_images, test_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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": [
    "plt.figure()\n",
    "plt.plot(history.history['accuracy'], label='accuracy')\n",
    "plt.plot(history.history['val_accuracy'], label = 'val_accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.ylim([0.3, 1])\n",
    "plt.legend(loc='lower right')\n",
    "plt.show()\n",
    "# print('Test accuracy:', test_acc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that both training accuracy and testing are approximately 0.96 after 15 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[6.8568243e-06 6.9982873e-04 9.7350216e-01 2.5178740e-02 4.1780787e-08\n",
      "  2.3039405e-07 3.0347206e-05 8.8505876e-06 5.7245581e-04 5.4355633e-07]]\n",
      "prediction: 2\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": [
    "# demonstration: predict the ith test digit\n",
    "i = 15\n",
    "prediction = model.predict(test_images[i].reshape(1,28,28,1))\n",
    "\n",
    "# get probability distribution and classification of the test digit\n",
    "print(prediction)\n",
    "print('prediction:', np.argmax(prediction))\n",
    "\n",
    "# draw the barplot\n",
    "plt.figure()\n",
    "plt.bar(np.arange(0,10).astype('str'), prediction[0,:])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "actual label: 2\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": [
    "# show the actual ith digit\n",
    "print('actual label:', np.argmax(test_labels[i]))\n",
    "plt.figure()\n",
    "plt.imshow(test_images[i,:,:,0], cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_labels = model.predict(test_images).argmax(axis=1)\n",
    "true_labels = test_labels.argmax(axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create confusion matrix and generate heatmap. This part is modified from https://androidkt.com/keras-confusion-matrix-in-tensorboard/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 967,    0,    2,    0,    0,    1,    7,    1,    2,    0],\n",
       "       [   0, 1123,    3,    2,    0,    0,    4,    2,    1,    0],\n",
       "       [   8,    1,  964,   10,   10,    3,    4,   17,   12,    3],\n",
       "       [   1,    1,    7,  961,    0,   15,    1,   11,    8,    5],\n",
       "       [   1,    1,    4,    0,  950,    0,    8,    4,    3,   11],\n",
       "       [   5,    2,    0,    6,    0,  862,    8,    1,    3,    5],\n",
       "       [   9,    4,    1,    0,    6,    5,  931,    1,    1,    0],\n",
       "       [   0,    5,   20,    6,    5,    0,    0,  973,    3,   16],\n",
       "       [   7,    1,    2,   11,    7,    7,    5,   10,  910,   14],\n",
       "       [   7,    7,    0,    5,   10,    7,    1,   12,    4,  956]])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_mat = tf.math.confusion_matrix(labels=true_labels, predictions=pred_labels).numpy()\n",
    "confusion_mat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "\n",
    "# normalize\n",
    "confusion_mat_norm = np.around(confusion_mat.astype('float') / np.atleast_2d(confusion_mat.sum(axis=1)).T, decimals=2)\n",
    "\n",
    "classes = np.arange(0,10).astype('str')\n",
    "confusion_mat_df = pd.DataFrame(confusion_mat_norm,\n",
    "                                index = classes, \n",
    "                                columns = classes)\n",
    "\n",
    "# generate heatmap\n",
    "figure = plt.figure()\n",
    "sns.heatmap(confusion_mat_df, annot=True, cmap=plt.cm.Blues)\n",
    "plt.tight_layout()\n",
    "plt.ylabel('True label')\n",
    "plt.xlabel('Predicted label')\n",
    "plt.show()"
   ]
  },
  {
   "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
}