Would it be helpful to have a model class for a De-noising Autoencoder?
Opened this issue · 2 comments
rfaulkner commented
Reference: http://jmlr.org/papers/volume11/vincent10a/vincent10a.pdf
I'm doing some work on this that builds on AutoEncoder.lua and would be willing to add this if it's useful.
koraykv commented
Sure, it would be nice. I have to say, I am not maintaining this repo actively, but if there is enough interest, I can start maintaining it and cleaning up the code.
rfaulkner commented
Sorry for the super late reply! Just checked back in on this recently since it had been mentioned to me that this repo was unsupported.
I wonder, would it be worth trying to merge some of these into nn? The only part of the interface that is different in unsup modules seems to be the normalize
method.