$t^3$-Variational Autoencoder: Learning Heavy-tailed Data with Student's t and Power Divergence

arXiv:2312.01133v2 Announce Type: replace-cross
Abstract: The variational autoencoder (VAE) typically employs a standard normal prior as a regularizer for the probabilistic latent encoder. However, the Gaussian tail often decays too quickly to effectively accommodate the encoded points, failing to preserve crucial structures hidden in the data. In this paper, we explore the use of heavy-tailed models to combat over-regularization. Drawing upon insights from information geometry, we propose $t^3$VAE, a modified VAE framework that incorporates Student's t-distributions for the prior, encoder, …

abstract arxiv autoencoder cs.lg data divergence encoder hidden normal paper power prior standard stat.ml type vae