Embedding-reparameterization procedure for manifold-valued latent variables in generative models
This is an incremental improvement for researchers in generative modeling, focusing on enhancing VAE learning capacity through manifold-valued priors.
The authors tackled the problem of using Gaussian priors in Variational Auto-Encoders (VAEs) by proposing an embedding-reparameterization procedure (ER) to incorporate arbitrary manifold-valued latent variables, but the results are preliminary as it is a work in progress with no concrete numbers reported.
Conventional prior for Variational Auto-Encoder (VAE) is a Gaussian distribution. Recent works demonstrated that choice of prior distribution affects learning capacity of VAE models. We propose a general technique (embedding-reparameterization procedure, or ER) for introducing arbitrary manifold-valued variables in VAE model. We compare our technique with a conventional VAE on a toy benchmark problem. This is work in progress.