AMLab-Amsterdam / DIVA

Implementation of 'DIVA: Domain Invariant Variational Autoencoders'
MIT License
98 stars 19 forks source link

about Domain Invariant #5

Open DCruss opened 2 years ago

DCruss commented 2 years ago

In your article,u have mentioned"Since our main goal is a domain invariant classifier, during inference we only use the encoder qφy (zy|x) and the auxiliary classifier qωy (y|zy)."But in your code or your final loss,u regard auxiliary multiplier y and auxiliary multiplier d as the same one.In the other word,where u reflect the Domain Invariant?