这里我有一个用于Keras的GoogleNet模型。是否有任何可能的方法来阻止网络中各个层的变化？我希望阻止预训练模型的前两层的变化。

回答：

当我说“阻止各个层的变化”时，我假设你不希望训练这些层，也就是说，你不希望修改在之前训练中可能已经学到的权重。

如果是这样的话，你可以将trainable=False传递给层，这样这些参数就不会用于训练更新规则。

示例：

from keras.models import Sequentialfrom keras.layers import Dense, Activationmodel = Sequential([    Dense(32, input_dim=100),    Dense(output_dim=10),    Activation('sigmoid'),])model.summary()model2 = Sequential([    Dense(32, input_dim=100,trainable=False),    Dense(output_dim=10),    Activation('sigmoid'),])model2.summary()

你可以在第二个模型的模型摘要中看到，这些参数被计为非可训练参数。

____________________________________________________________________________________________________Layer (type)                     Output Shape          Param #     Connected to                     ====================================================================================================dense_1 (Dense)                  (None, 32)            3232        dense_input_1[0][0]              ____________________________________________________________________________________________________dense_2 (Dense)                  (None, 10)            330         dense_1[0][0]                    ____________________________________________________________________________________________________activation_1 (Activation)        (None, 10)            0           dense_2[0][0]                    ====================================================================================================Total params: 3,562Trainable params: 3,562Non-trainable params: 0________________________________________________________________________________________________________________________________________________________________________________________________________Layer (type)                     Output Shape          Param #     Connected to                     ====================================================================================================dense_3 (Dense)                  (None, 32)            3232        dense_input_2[0][0]              ____________________________________________________________________________________________________dense_4 (Dense)                  (None, 10)            330         dense_3[0][0]                    ____________________________________________________________________________________________________activation_2 (Activation)        (None, 10)            0           dense_4[0][0]                    ====================================================================================================Total params: 3,562Trainable params: 330Non-trainable params: 3,232