{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"from keras.models import Sequential\nfrom keras.utils import np_utils\nfrom keras.layers.core import Dense, Activation, Dropout, Flatten\nfrom keras.layers.convolutional import Convolution2D, MaxPooling2D\n\nimport pandas as pd\nimport numpy as np\n\n# Read data\ntrain = pd.read_csv('../input/digit-recognizer/train.csv')\nlabels = train.iloc[:,0].values.astype('int32')\nX_train = (train.iloc[:,1:].values).astype('float32')\nX_test = (pd.read_csv('../input/digit-recognizer/test.csv').values).astype('float32')\n\n# convert list of labels to binary class matrix\ny_train = np_utils.to_categorical(labels) \n\n# pre-processing: divide by max and substract mean\nscale = np.max(X_train)\nX_train /= scale\nX_test /= scale\n\nmean = np.std(X_train)\nX_train -= mean\nX_test -= mean\n\n# get input (784) and output (10) dimensions\ninput_dim = X_train.shape[1]\nnb_classes = y_train.shape[1]\n\n# 3 layers: 128 x 128 x 10\ndef get_our_simple_model():\n    model = Sequential()\n    model.add(Dense(128, input_dim=input_dim))\n    model.add(Activation('relu'))\n    model.add(Dropout(0.15))\n    model.add(Dense(128))\n    model.add(Activation('relu'))\n    model.add(Dropout(0.15))\n    model.add(Dense(nb_classes))\n    model.add(Activation('softmax'))\n    return model\n\ndef get_convolutional_model(input_shape):\n    model = Sequential()\n    model.add(Convolution2D(32, 5, 5, padding='same', input_shape=input_shape))\n    model.add(Activation('relu'))\n    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'))\n    model.add(Convolution2D(64, 5, 5, padding='same', input_shape=input_shape))\n    model.add(Activation('relu'))\n    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'))\n    model.add(Flatten())\n    model.add(Dense(1024))\n    model.add(Activation('relu'))\n    model.add(Dropout(0.5))\n    model.add(Dense(nb_classes))\n    model.add(Activation('softmax'))\n    return model\n\n# simple model\n#model = get_our_simple_model()\n#batch_size = 16\n\n# complex model\nbatch_size = 64\nimg_rows, img_cols = 28, 28\ninput_shape = (img_rows, img_cols, 1)\nX_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)\nX_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)\nmodel = get_convolutional_model(input_shape)\n\n\n# we'll use categorical xent for the loss, and RMSprop as the optimizer\nmodel.compile(loss='categorical_crossentropy', optimizer='adam')\n\nprint(\"Training...\")\nmodel.fit(X_train, y_train, epochs=10, batch_size=batch_size, validation_split=0.1, verbose=2)\n\nprint(\"Generating test predictions...\")\npreds = np.argmax(model.predict(X_test), axis=-1)\n\ndef write_preds(preds, fname):\n    pd.DataFrame({\"ImageId\": list(range(1,len(preds)+1)), \"Label\": preds}).to_csv(fname, index=False, header=True)\n\nwrite_preds(preds, \"test_prediction.csv\")\nprint(\"Done!\")","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-07-22T12:44:24.924566Z","iopub.execute_input":"2021-07-22T12:44:24.924925Z","iopub.status.idle":"2021-07-22T12:45:10.216722Z","shell.execute_reply.started":"2021-07-22T12:44:24.924836Z","shell.execute_reply":"2021-07-22T12:45:10.215623Z"},"trusted":true},"execution_count":null,"outputs":[]}]}