train_with_keras_mlp.py

train_with_keras_mlp.py

from keras.models import Sequential
from keras.layers.core import Dense, Activation
from keras.optimizers import SGD
from keras.losses import categorical_crossentropy
from keras.initializers import he_normal

n_features = X.shape[1]
n_classes = Y.shape[1]

batch_size = int(len(X_train)*0.2)  # ミニバッチサイズ
epochs = 100  # エポック数
mu = 0.05 # 学習率

random_state = 1
np.random.seed(random_state)

model = Sequential()
model.add(Dense(10, activation='relu', kernel_initializer=he_normal(seed=random_state), bias_initializer='zeros', input_shape=(n_features,)))
model.add(Dense(10, activation='relu', kernel_initializer=he_normal(seed=random_state), bias_initializer='zeros'))
model.add(Dense(10, activation='relu', kernel_initializer=he_normal(seed=random_state), bias_initializer='zeros'))
model.add(Dense(n_classes, activation='softmax', kernel_initializer=he_normal(seed=random_state), bias_initializer='zeros'))
#model.summary()

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

history = model.fit(X_train, Y_train,       # トレーニングデータ
                    batch_size=batch_size,  # バッチサイズ
                    epochs=epochs,          # 総エポック数
                    validation_data=(X_test, Y_test),   # テストデータ
                    shuffle=False,          # エポック毎にデータをシャッフルしない
                    verbose=0)              # 学習履歴をprintしない

print("acc_train: "+ str(history.history['acc'][-1]) + "   acc_test: "+ str(history.history['val_acc'][-1]))