Improving Variational Autoencoder Estimation from Incomplete Data with Mixture Variational Families

TL;DR

This paper introduces mixture variational families to improve variational autoencoder estimation when training data is incomplete, addressing increased posterior complexity and enhancing model accuracy.

Contribution

The paper proposes two novel strategies using mixture variational families to better handle incomplete data in VAEs, improving estimation accuracy.

Findings

Mixture variational families improve VAE estimation from incomplete data.

Proposed methods outperform standard VAEs in experiments.

Mixture approaches effectively address posterior complexity issues.

Abstract

We consider the task of estimating variational autoencoders (VAEs) when the training data is incomplete. We show that missing data increases the complexity of the model's posterior distribution over the latent variables compared to the fully-observed case. The increased complexity may adversely affect the fit of the model due to a mismatch between the variational and model posterior distributions. We introduce two strategies based on (i) finite variational-mixture and (ii) imputation-based variational-mixture distributions to address the increased posterior complexity. Through a comprehensive evaluation of the proposed approaches, we show that variational mixtures are effective at improving the accuracy of VAE estimation from incomplete data.