(X_train, Y_train), (X_test, Y_test) = load_mnist(normalize=True, one_hot_label=True)

model = Sequential()
model.add(Dense(100, input_shape(784,)))
model.add(Activation('sigmoid'))
model.add(Dense(10))
model.add(Activation('softmax'))

sgd = SGD(lr = 0.1)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])

batch_size = 100
nb_epoch = 16 # X_train: (60000, 784), 1 epoch = 100*600
model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, verbose=0, validation_data=(X_test, Y_test))

# 모델 저장
model.save('mlp.h5')

# 모델을 사용해서 예측
from keras.models import load_model

model = load_model('mlp.h5')

pc = model.predict_classes(X_test[0:100, :], 100)
print(pc) # [7 2 1 0 4 1 ...]

pb= model.predict_proba(X_test[0:100, :], 100)
print(pb) # [[1.44900128e-04 2.8030103e-06 ...]]