Improving Explorability in Variational Inference with Annealed Variational Objectives

TL;DR

This paper introduces Annealed Variational Objectives (AVO), a novel approach that enhances exploration in variational inference by incorporating energy tempering, leading to more robust and expressive approximate posteriors.

Contribution

The paper proposes AVO, a new training method for hierarchical variational models that improves exploration and reduces bias towards unimodal posteriors by integrating annealing techniques.

Findings

AVO improves robustness to warm-up procedures

Encourages exploration in latent space

Enhances the expressiveness of variational approximations

Abstract

Despite the advances in the representational capacity of approximate distributions for variational inference, the optimization process can still limit the density that is ultimately learned. We demonstrate the drawbacks of biasing the true posterior to be unimodal, and introduce Annealed Variational Objectives (AVO) into the training of hierarchical variational methods. Inspired by Annealed Importance Sampling, the proposed method facilitates learning by incorporating energy tempering into the optimization objective. In our experiments, we demonstrate our method's robustness to deterministic warm up, and the benefits of encouraging exploration in the latent space.