Feature Space Particle Inference for Neural Network Ensembles

TL;DR

This paper introduces a novel particle-based inference method in feature space for neural network ensembles, improving diversity and robustness while addressing over-parameterization and underfitting issues.

Contribution

It proposes optimizing particles in feature space to enhance ensemble diversity and performance, a novel approach compared to traditional weight or function space methods.

Findings

Outperforms Deep Ensembles on accuracy, calibration, and robustness

Addresses over-parameterization and underfitting issues effectively

Enhances ensemble diversity through feature space optimization

Abstract

Ensembles of deep neural networks demonstrate improved performance over single models. For enhancing the diversity of ensemble members while keeping their performance, particle-based inference methods offer a promising approach from a Bayesian perspective. However, the best way to apply these methods to neural networks is still unclear: seeking samples from the weight-space posterior suffers from inefficiency due to the over-parameterization issues, while seeking samples directly from the function-space posterior often results in serious underfitting. In this study, we propose optimizing particles in the feature space where the activation of a specific intermediate layer lies to address the above-mentioned difficulties. Our method encourages each member to capture distinct features, which is expected to improve ensemble prediction robustness. Extensive evaluation on real-world datasets shows that our model significantly outperforms the gold-standard Deep Ensembles on various metrics, including accuracy, calibration, and robustness. Code is available at https://github.com/DensoITLab/featurePI .