machine-learning - 在 Keras 中为序列到序列自动编码器制作解码器模型

Question

我正在处理这个示例，但对如何制作单独的解码器模型感到困惑。

from keras.layers import Input, LSTM, RepeatVector
from keras.models import Model

inputs = Input(shape=(timesteps, input_dim))
encoded = LSTM(latent_dim)(inputs)

decoded = RepeatVector(timesteps)(encoded)
decoded = LSTM(input_dim, return_sequences=True)(decoded)

sequence_autoencoder = Model(inputs, decoded)
encoder = Model(inputs, encoded)

我知道如何制作编码器，但是我们如何制作单独的解码器呢？我可以定义所有层并分别制作编码器和解码器，但是有没有更简单的方法可以像我们对编码器模型所做的那样？

Answer 1

创建编码器:

inputs = Input(shape=(timesteps, input_dim))
encoded = LSTM(latent_dim)(inputs)
encoder = Model(inputs, encoded)

创建解码器:

decInput = Input((the shape of the encoder's output))    
decoded = RepeatVector(timesteps)(decInput)
decoded = LSTM(input_dim, return_sequences=True)(decoded)
decoder = Model(decInput,decoded)

加入模型:

joinedInput = Input(shape=(timesteps, input_dim))
encoderOut = encoder(joinedInput)    
joinedOut = decoder(encoderOut)
sequence_autoencoder = Model(joinedInput,joinedOut)