The Beta-VAE model offers the regularization over the -$\beta$ KL(q(z|x)||p(z)) term of the ELBO term in VAE formulation. Scaling this term up forces some latents to be distributed as unit Gaussians (given that p(z) is isotropic Gaussian). I wonder, is it possible to acquire the same regularization by only tuning the prior such as having a Gaussian with smaller std such as 0.01?

Beta VAE paper: https://openreview.net/forum?id=Sy2fzU9gl


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.