Tackling covariate shift with node-based Bayesian neural networks

TL;DR

This paper enhances node-based Bayesian neural networks by increasing the entropy of latent variables during training, leading to better uncertainty estimation and robustness under covariate shift and noisy labels.

Contribution

It introduces a simple method to boost the entropy of latent variables in node-based BNNs, improving their performance under covariate shift and noisy data.

Findings

Improved uncertainty estimation under covariate shift.

Enhanced robustness to noisy training labels.

Better out-of-distribution classification performance.

Abstract

Bayesian neural networks (BNNs) promise improved generalization under covariate shift by providing principled probabilistic representations of epistemic uncertainty. However, weight-based BNNs often struggle with high computational complexity of large-scale architectures and datasets. Node-based BNNs have recently been introduced as scalable alternatives, which induce epistemic uncertainty by multiplying each hidden node with latent random variables, while learning a point-estimate of the weights. In this paper, we interpret these latent noise variables as implicit representations of simple and domain-agnostic data perturbations during training, producing BNNs that perform well under covariate shift due to input corruptions. We observe that the diversity of the implicit corruptions depends on the entropy of the latent variables, and propose a straightforward approach to increase the entropy of these variables during training. We evaluate the method on out-of-distribution image classification benchmarks, and show improved uncertainty estimation of node-based BNNs under covariate shift due to input perturbations. As a side effect, the method also provides robustness against noisy training labels.