Mixed Variational Flows for Discrete Variables

arXiv:2308.15613v3 Announce Type: replace-cross
Abstract: Variational flows allow practitioners to learn complex continuous distributions, but approximating discrete distributions remains a challenge. Current methodologies typically embed the discrete target in a continuous space - usually via continuous relaxation or dequantization - and then apply a continuous flow. These approaches involve a surrogate target that may not capture the original discrete target, might have biased or unstable gradients, and can create a difficult optimization problem. In this work, we develop a variational …

abstract apply arxiv challenge continuous cs.lg current embed flow learn mixed space stat.co stat.ml type variables via