使用Keras合并两个不同网络的特征表示

from keras.models import Model
from keras.layers import Input, Dense, Concatenate

# 定义模型A
input_a = Input(shape=(input_shape_a,))
hidden_a = Dense(64, activation='relu')(input_a)
output_a = Dense(10, activation='softmax')(hidden_a)
model_a = Model(inputs=input_a, outputs=output_a)

# 定义模型B
input_b = Input(shape=(input_shape_b,))
hidden_b = Dense(64, activation='relu')(input_b)
output_b = Dense(10, activation='softmax')(hidden_b)
model_b = Model(inputs=input_b, outputs=output_b)

# 合并模型A和模型B的特征表示
merged = Concatenate()([model_a.output, model_b.output])

# 定义新的模型
output = Dense(10, activation='softmax')(merged)
model = Model(inputs=[model_a.input, model_b.input], outputs=output)

# 编译和训练模型
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
model.fit([input_data_a, input_data_b], labels, epochs=10, batch_size=32)