Practical Deep Learning with Bayesian Principles

TL;DR

This paper presents a practical approach to deep learning using Bayesian principles through natural-gradient variational inference, achieving competitive performance and improved uncertainty estimation on large datasets like ImageNet.

Contribution

It introduces a practical training method combining Bayesian principles with deep networks, utilizing techniques like batch normalization and distributed training.

Findings

Achieves performance comparable to Adam optimizer on ImageNet

Predictive probabilities are well-calibrated and uncertainties are improved

Enhances continual learning capabilities

Abstract

Bayesian methods promise to fix many shortcomings of deep learning, but they are impractical and rarely match the performance of standard methods, let alone improve them. In this paper, we demonstrate practical training of deep networks with natural-gradient variational inference. By applying techniques such as batch normalisation, data augmentation, and distributed training, we achieve similar performance in about the same number of epochs as the Adam optimiser, even on large datasets such as ImageNet. Importantly, the benefits of Bayesian principles are preserved: predictive probabilities are well-calibrated, uncertainties on out-of-distribution data are improved, and continual-learning performance is boosted. This work enables practical deep learning while preserving benefits of Bayesian principles. A PyTorch implementation is available as a plug-and-play optimiser.