用于连接层的 LSTM

Question

我正在尝试通过 LSTM 处理 5 层，但遇到了错误。 下面是代码。

yhdistelma=layers.concatenate([input1,input2, 
                              input3, input4, 
                              input5, input6, 
                              input7, input8, input9])

first_output  = layers.Dense(30,name='output_1')(yhdistelma)
second_output = layers.Dense(30,name='output_2')(yhdistelma)
third_output  = layers.Dense(30,name='output_3')(yhdistelma)
fourth_output = layers.Dense(30,name='output_4')(yhdistelma)
fifth_output  = layers.Dense(30,name='output_5')(yhdistelma)

hhs= layers.concatenate([first_output, second_output, 
                         third_output, fourth_output, 
                         fifth_output])

xxs=layers.LSTM(5)(hhs)

model.compile 给出的错误是

 Input 0 of layer lstm_2 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: (None, 150)

Answer 1

LSTM 的输入形状应该是一个 3D 张量，例如[batch, timesteps, feature]。对于您的情况，您可以按如下方式扩展它的输入维度：

from tensorflow.keras import layers

yhdistelma    = layers.Input(20)
first_output  = layers.Dense(30, name='output_1')(yhdistelma)
second_output = layers.Dense(30, name='output_2')(yhdistelma)
third_output  = layers.Dense(30, name='output_3')(yhdistelma)
fourth_output = layers.Dense(30, name='output_4')(yhdistelma)
fifth_output  = layers.Dense(30, name='output_5')(yhdistelma)
hhs = layers.concatenate([first_output, second_output, 
                         third_output, fourth_output, 
                         fifth_output])
xxs = layers.LSTM(5)(tf.expand_dims(hhs, axis=-1))


print(hhs.shape)
print(tf.expand_dims(hhs, axis=-1).shape)
print(xxs.shape)
(None, 150)
(None, 150, 1)
(None, 5)