Keras 빨리 훑어보기

conda create -n keras python=3.5
source activate keras
 
pip install tensorflow
conda install scipy
pip install keras
pip install h5py

(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 ...]]

model = Sequential()
nb_filter = 30
kernal_size = (5, 5)
input_shape = (img_rows, img_cols, 1)
 
model.add(Convolution2D(nb_filter, kernal_size[0], kernal_size[1], border_mode='valid', input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size(2,2)))
 
model.add(Flatten())
model.add(Dense(100))
model.add(Activation('relu'))
model.add(Dense(10))
model.add(Activation('softmax'))
 
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
model.fit(X_train, Y_train, batch_size=100, nb_epoch=10, verbose=1, validation_data=(X_test, Y_test))

from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation
 
model = Sequential()
model.add(Dense(512, input_shape=(784,))) # 입력 784, 출력 512
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(512)) # 입력 512(이전 레이어 입력), 출력 512
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(10)) # 입력 512, 출력 10
model.add(Activation('softmax'))

model = Sequential()
model.add(Dense(32, input_shape=(16,))) # 입력 16, 출력 32
 
# 첫 번째 레이어 이후로는 입력 개수 지정 필요 없음
# 앞 레이어의 출력 개수가 입력 개수가 됨
# 입력 32, 출력 64
model.add(Dense(64))

from keras.regularizers import l2
model.add(Dense(64, input_dim=64, init='he_normal', W_regularizer=l2(0.01)))

model.add(Activation('relu'))
...
model.add(Activation('softmax'))

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

his = model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, verbose=2, validation_data=(X_test, Y_test))

score = model.evaluate(X_test, Y_test, verbose=0)
print(model.metrics_names) # loss, acc
print('Test score', score[0]) # loss
print('Test accuracy:', score[1]) # acc

tensorcallback = TensorBoard(log_dir='./logs', histogram_freq=0, write_graph=True, write_images=False)
his = model.fit(X_train, Y_train, ..., callbacks=[tensorcallback])