English

Discouraging posterior collapse in hierarchical Variational Autoencoders using context

Machine Learning 2023-09-29 v2 Computer Vision and Pattern Recognition

Abstract

Hierarchical Variational Autoencoders (VAEs) are among the most popular likelihood-based generative models. There is a consensus that the top-down hierarchical VAEs allow effective learning of deep latent structures and avoid problems like posterior collapse. Here, we show that this is not necessarily the case, and the problem of collapsing posteriors remains. To discourage this issue, we propose a deep hierarchical VAE with a context on top. Specifically, we use a Discrete Cosine Transform to obtain the last latent variable. In a series of experiments, we observe that the proposed modification allows us to achieve better utilization of the latent space and does not harm the model's generative abilities.

Keywords

Cite

@article{arxiv.2302.09976,
  title  = {Discouraging posterior collapse in hierarchical Variational Autoencoders using context},
  author = {Anna Kuzina and Jakub M. Tomczak},
  journal= {arXiv preprint arXiv:2302.09976},
  year   = {2023}
}

Comments

Code: https://github.com/AKuzina/dct_vae

R2 v1 2026-06-28T08:44:32.118Z