Is it possible to add a trainable filter after an autoencoder?

2

Based on your previous post it seems your wiener filter implementation doesn’t contain any trainable parameters.
How and which parameters would you like to train?