Alert button

Embedding-reparameterization procedure for manifold-valued latent variables in generative models

Add code
Bookmark button
Alert button
Dec 06, 2018
Eugene Golikov, Maksim Kretov

Figure 1 for Embedding-reparameterization procedure for manifold-valued latent variables in generative models
Figure 2 for Embedding-reparameterization procedure for manifold-valued latent variables in generative models
Figure 3 for Embedding-reparameterization procedure for manifold-valued latent variables in generative models
Figure 4 for Embedding-reparameterization procedure for manifold-valued latent variables in generative models

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon

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.

* Presented at Bayesian Deep Learning workshop (NeurIPS 2018)  
View paper onarxiv icon

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon