Kaggle Competition
"},{"metadata":{},"cell_type":"markdown","source":"Digit Recognizer
"},{"metadata":{},"cell_type":"markdown","source":"### Data Loading & Preprocessing"},{"metadata":{"trusted":true},"cell_type":"code","source":"# importing libraries\nfrom __future__ import print_function\nimport pandas as pd\nimport numpy as np\nimport keras\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\nfrom keras.layers.normalization import BatchNormalization\nfrom keras import backend as K\nimport matplotlib.pyplot as plt\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")","execution_count":16,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# importing train dataset\ntrain = pd.read_csv('../input/train.csv')\nprint('The dimensions of Train dataset are - ', train.shape)","execution_count":17,"outputs":[{"output_type":"stream","text":"The dimensions of Train dataset are - (42000, 785)\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# importing test dataset\ntest = pd.read_csv('../input/test.csv')\nprint('The dimensions of Test dataset are - ', test.shape)","execution_count":18,"outputs":[{"output_type":"stream","text":"The dimensions of Test dataset are - (28000, 784)\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# setting the model parameters\nbatch_size = 128\nnum_classes = 10\nepochs = 10\nimg_rows, img_cols = 28, 28","execution_count":19,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# dropping the label column from train dataset\nX_train = train.drop(['label'], axis=1)\nX_test = test\n\ny_train = train['label']\n# converting each value of y_train to binary vector of size=10\n# for y=2 => [0,0,1,0,0,0,0,0,0,0]\ny_train = keras.utils.to_categorical(y_train, num_classes)","execution_count":20,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# converting the pandas dataframe to numpy array\nX_train = np.array(X_train)\nX_test = np.array(X_test)","execution_count":21,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# resizing the train & test datasets\nif K.image_data_format() == 'channels_first':\n X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)\n X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)\n input_shape = (1, img_rows, img_cols)\nelse:\n X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)\n X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)\n input_shape = (img_rows, img_cols, 1)","execution_count":22,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print('The dimensions of the resized Train dataset is', X_train.shape)\nprint('The dimensions of the resized Test dataset is', X_train.shape)","execution_count":23,"outputs":[{"output_type":"stream","text":"The dimensions of the resized Train dataset is (42000, 28, 28, 1)\nThe dimensions of the resized Test dataset is (42000, 28, 28, 1)\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# this function is used to update the plots for each epoch and error\ndef plt_dynamic(x, ty, ax, colors=['b']):\n ax.plot(x, ty, 'r', label=\"Train Loss\")\n plt.legend()\n plt.grid()\n fig.canvas.draw()","execution_count":24,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### Applying CNN model"},{"metadata":{},"cell_type":"markdown","source":"**3 Convolution Layers + Kernel size=5 + Padding + Batch Norm. with Max Pooling**"},{"metadata":{"trusted":true},"cell_type":"code","source":"model = Sequential()\n# layer 1\nmodel.add(Conv2D(128, kernel_size=(5, 5),\n kernel_initializer='he_normal',\n activation='relu',\n padding='same',\n input_shape=input_shape))\nmodel.add(BatchNormalization())\n\n#layer 2\nmodel.add(Conv2D(64, kernel_size=(5, 5),\n kernel_initializer='he_normal',\n padding='same',\n activation='relu'\n ))\nmodel.add(BatchNormalization())\n\n#layer 3\nmodel.add(Conv2D(32, kernel_size=(5, 5),\n kernel_initializer='he_normal',\n padding='same',\n activation='relu'))\nmodel.add(BatchNormalization())\n\nmodel.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n\nmodel.add(Flatten())\n\nmodel.add(Dense(128, activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(rate=0.5))\n\nmodel.add(Dense(num_classes, activation='softmax'))\n\nmodel.compile(loss=keras.losses.categorical_crossentropy,\n optimizer=keras.optimizers.Adadelta(),\n metrics=['accuracy'])\n\nhistory = model.fit(X_train, y_train,\n batch_size=batch_size,\n epochs=epochs,\n verbose=1)","execution_count":25,"outputs":[{"output_type":"stream","text":"Epoch 1/10\n42000/42000 [==============================] - 8s 193us/step - loss: 0.1582 - acc: 0.9526\nEpoch 2/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0541 - acc: 0.9841\nEpoch 3/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0363 - acc: 0.9886\nEpoch 4/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0266 - acc: 0.9918\nEpoch 5/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0205 - acc: 0.9936\nEpoch 6/10\n42000/42000 [==============================] - 7s 167us/step - loss: 0.0159 - acc: 0.9954\nEpoch 7/10\n42000/42000 [==============================] - 7s 164us/step - loss: 0.0118 - acc: 0.9964\nEpoch 8/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0100 - acc: 0.9968\nEpoch 9/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0077 - acc: 0.9975\nEpoch 10/10\n42000/42000 [==============================] - 7s 165us/step - loss: 0.0068 - acc: 0.9982\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# epoch vs loss plot\nfig,ax = plt.subplots(1,1)\nax.set_xlabel('Epoch') \nax.set_ylabel('Categorical Crossentropy Loss')\n\n# list of epoch numbers\nx = list(range(1,epochs+1))\n\nty = history.history['loss']\nplt_dynamic(x, ty, ax)","execution_count":26,"outputs":[{"output_type":"display_data","data":{"text/plain":"