Improved Contrastive Divergence Training of Energy Based Models

TL;DR

This paper enhances contrastive divergence training for energy-based models by including a crucial gradient term, improving stability and performance across various tasks like image generation and out-of-distribution detection.

Contribution

It introduces a novel adaptation that accounts for a previously neglected gradient term, leading to more stable training and better results.

Findings

Improved training stability in energy-based models.

Enhanced performance on image generation benchmarks.

Better out-of-distribution detection and compositional generation.

Abstract

Contrastive divergence is a popular method of training energy-based models, but is known to have difficulties with training stability. We propose an adaptation to improve contrastive divergence training by scrutinizing a gradient term that is difficult to calculate and is often left out for convenience. We show that this gradient term is numerically significant and in practice is important to avoid training instabilities, while being tractable to estimate. We further highlight how data augmentation and multi-scale processing can be used to improve model robustness and generation quality. Finally, we empirically evaluate stability of model architectures and show improved performance on a host of benchmarks and use cases,such as image generation, OOD detection, and compositional generation.