Making decoder model for sequence to sequence autoencoder in Keras

Question

I'm working with this example but am confused on how to make the separate decoder model.

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)

I understand how to make the encoder, but how do we make a separate decoder? I can define all the layers and make the encoder and decoder separately but is there a simpler way to do it like we've done with the encoder model?

You could try just: decoder = Model(encoder.output,decoded) (Never tried this, but I'll add an answer with what I know that works) — Daniel Möller
– Daniel Möller, Commented May 7, 2017 at 16:34

Daniel Möller · Accepted Answer · 2017-05-07 16:37:17Z

1

Create the encoder:

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

Create the decoder:

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)

Joining models:

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

answered May 7, 2017 at 16:37

Daniel Möller

86.8k24 gold badges202 silver badges222 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Making decoder model for sequence to sequence autoencoder in Keras

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related