2016-03-19 16:07:15 +00:00
|
|
|
'''Trains a simple deep NN on the MNIST dataset.
|
2015-12-09 02:49:14 +00:00
|
|
|
|
2016-03-19 16:07:15 +00:00
|
|
|
Gets to 98.40% test accuracy after 20 epochs
|
2015-12-09 02:49:14 +00:00
|
|
|
(there is *a lot* of margin for parameter tuning).
|
|
|
|
2 seconds per epoch on a K520 GPU.
|
|
|
|
'''
|
|
|
|
|
2015-04-30 00:17:22 +00:00
|
|
|
from __future__ import print_function
|
2015-06-27 00:41:13 +00:00
|
|
|
|
2017-02-28 22:41:30 +00:00
|
|
|
import keras
|
2015-04-30 00:17:22 +00:00
|
|
|
from keras.datasets import mnist
|
|
|
|
from keras.models import Sequential
|
2017-02-28 22:41:30 +00:00
|
|
|
from keras.layers import Dense, Dropout
|
2017-01-06 17:25:03 +00:00
|
|
|
from keras.optimizers import RMSprop
|
2015-04-30 00:17:22 +00:00
|
|
|
|
|
|
|
|
2015-06-19 19:52:43 +00:00
|
|
|
batch_size = 128
|
2017-02-11 02:16:11 +00:00
|
|
|
num_classes = 10
|
2017-02-15 00:08:30 +00:00
|
|
|
epochs = 20
|
2015-04-30 00:17:22 +00:00
|
|
|
|
2016-04-04 18:30:24 +00:00
|
|
|
# the data, shuffled and split between train and test sets
|
2017-02-11 02:16:11 +00:00
|
|
|
(x_train, y_train), (x_test, y_test) = mnist.load_data()
|
2015-05-04 17:31:03 +00:00
|
|
|
|
2017-02-11 02:16:11 +00:00
|
|
|
x_train = x_train.reshape(60000, 784)
|
|
|
|
x_test = x_test.reshape(10000, 784)
|
|
|
|
x_train = x_train.astype('float32')
|
|
|
|
x_test = x_test.astype('float32')
|
|
|
|
x_train /= 255
|
|
|
|
x_test /= 255
|
|
|
|
print(x_train.shape[0], 'train samples')
|
|
|
|
print(x_test.shape[0], 'test samples')
|
2015-04-30 00:17:22 +00:00
|
|
|
|
|
|
|
# convert class vectors to binary class matrices
|
2017-02-28 22:41:30 +00:00
|
|
|
y_train = keras.utils.to_categorical(y_train, num_classes)
|
|
|
|
y_test = keras.utils.to_categorical(y_test, num_classes)
|
2015-04-30 00:17:22 +00:00
|
|
|
|
|
|
|
model = Sequential()
|
2017-02-11 02:16:11 +00:00
|
|
|
model.add(Dense(512, activation='relu', input_shape=(784,)))
|
2015-05-04 17:31:03 +00:00
|
|
|
model.add(Dropout(0.2))
|
2017-02-11 02:16:11 +00:00
|
|
|
model.add(Dense(512, activation='relu'))
|
2015-05-04 17:31:03 +00:00
|
|
|
model.add(Dropout(0.2))
|
2017-02-11 02:16:11 +00:00
|
|
|
model.add(Dense(10, activation='softmax'))
|
2015-04-30 00:17:22 +00:00
|
|
|
|
2016-03-19 16:07:15 +00:00
|
|
|
model.summary()
|
2015-04-30 00:17:22 +00:00
|
|
|
|
2016-03-19 16:07:15 +00:00
|
|
|
model.compile(loss='categorical_crossentropy',
|
2016-04-10 12:38:00 +00:00
|
|
|
optimizer=RMSprop(),
|
2016-03-19 16:07:15 +00:00
|
|
|
metrics=['accuracy'])
|
|
|
|
|
2017-02-11 02:16:11 +00:00
|
|
|
history = model.fit(x_train, y_train,
|
2017-02-15 00:08:30 +00:00
|
|
|
batch_size=batch_size, epochs=epochs,
|
2017-02-11 02:16:11 +00:00
|
|
|
verbose=1, validation_data=(x_test, y_test))
|
|
|
|
score = model.evaluate(x_test, y_test, verbose=0)
|
|
|
|
print('Test loss:', score[0])
|
2015-05-04 17:31:03 +00:00
|
|
|
print('Test accuracy:', score[1])
|