model = Sequential()
model.add(
Dense(512,
activation='relu',
kernel_regularizer=l2(1e-5),
input_shape=(784, )))
model.add(Dense(512, activation='relu', kernel_regularizer=l2(1e-5)))
model.add(Dense(10, kernel_regularizer=l2(1e-5)))
model.add(Activation('softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer=RMSprop(),
metrics=['accuracy'])
if args.importance_training:
wrapped = ImportanceTraining(model, presample=5)
else:
wrapped = model
history = wrapped.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test))
score = model.evaluate(x_test, y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255
wrapped_model = ImportanceTraining(model,
k=0.5,
presample=64,
adaptive_smoothing=True,
smooth=0.5,
forward_batch_size=64)
if not data_augmentation:
print('Not using data augmentation.')
wrapped_model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(x_test, y_test),
shuffle=True)
else:
print('Using real-time data augmentation.')
# This will do preprocessing and realtime data augmentation:
datagen = ImageDataGenerator(
featurewise_center=False, # set input mean to 0 over the dataset
samplewise_center=False, # set each sample mean to 0
featurewise_std_normalization=
False, # divide inputs by std of the dataset
samplewise_std_normalization=False, # divide each input by its std
zca_whitening=False, # apply ZCA whitening
rotation_range=
0, # randomly rotate images in the range (degrees, 0 to 180)
width_shift_range=