VAE Learning via Stein Variational Gradient Descent

TL;DR

This paper introduces a novel VAE learning method using Stein variational gradient descent, eliminating the need for parametric encoder assumptions and improving performance with importance sampling, scalable to large datasets.

Contribution

The paper presents a non-parametric VAE training approach based on Stein variational gradient descent, enhancing flexibility and scalability over traditional methods.

Findings

Effective on multiple unsupervised and semi-supervised tasks

Demonstrates scalability to large datasets like ImageNet

Achieves improved performance with importance sampling

Abstract

A new method for learning variational autoencoders (VAEs) is developed, based on Stein variational gradient descent. A key advantage of this approach is that one need not make parametric assumptions about the form of the encoder distribution. Performance is further enhanced by integrating the proposed encoder with importance sampling. Excellent performance is demonstrated across multiple unsupervised and semi-supervised problems, including semi-supervised analysis of the ImageNet data, demonstrating the scalability of the model to large datasets.