Keras:在几层中重复使用权重



我正在尝试使用Keras函数API实现一个神经元网络,该API对多个层使用相同的权重。代码正在工作,但我不确定我创建的"共享层"是否符合我的需求。示例中的两个隐藏层是否使用相同的权重,或者我是否创建了一个层的两个不同实例,它们只有共同的结构?如果没有,有没有办法实现我想要的?

# create shared_layer
inputs = Input(shape=(784,))
outputs = layers.Dense(784, activation='relu')(inputs)
shared_layer = Model(inputs=inputs, outputs=outputs)
# create model
visible = Input(shape=(28, 28, 1))
flat = layers.Flatten()(visible)
hidden = shared_layer(flat)
hidden2 = shared_layer(hidden)
output = layers.Dense(10, activation='softmax')(hidden2)
new_model = Model(inputs=visible, outputs=output)

当我查看模型的摘要时,我得到这个:

Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_4 (InputLayer)            (None, 28, 28, 1)    0                                            
__________________________________________________________________________________________________
flatten_2 (Flatten)             (None, 784)          0           input_4[0][0]                    
__________________________________________________________________________________________________
model_3 (Model)                 (None, 784)          615440      flatten_2[0][0]                  
model_3[1][0]                    
__________________________________________________________________________________________________
dense_4 (Dense)                 (None, 10)           7850        model_3[2][0]                    
==================================================================================================

它是共享的,但你正在做不必要的事情。

您可以:

shared_layer = layers.Dense(784, activation='relu')
visible = Input(shape=(28, 28, 1))
flat = layers.Flatten()(visible)
hidden = shared_layer(flat)
hidden2 = shared_layer(hidden)
output = layers.Dense(10, activation='softmax')(hidden2)
new_model = Model(inputs=visible, outputs=output)

最新更新