Multi-Symmetry Ensembles: Improving Diversity and Generalization via Opposing Symmetries

TL;DR

Multi-Symmetry Ensembles (MSE) enhance model diversity by leveraging symmetry axes to explore hypothesis space beyond stochastic methods, improving performance and generalization in large datasets.

Contribution

The paper introduces MSE, a novel framework using symmetry-based hypotheses and contrastive learning to diversify ensembles beyond stochastic weight perturbations.

Findings

MSE improves classification accuracy on ImageNet.

MSE enhances uncertainty quantification.

MSE boosts transfer learning performance.

Abstract

Deep ensembles (DE) have been successful in improving model performance by learning diverse members via the stochasticity of random initialization. While recent works have attempted to promote further diversity in DE via hyperparameters or regularizing loss functions, these methods primarily still rely on a stochastic approach to explore the hypothesis space. In this work, we present Multi-Symmetry Ensembles (MSE), a framework for constructing diverse ensembles by capturing the multiplicity of hypotheses along symmetry axes, which explore the hypothesis space beyond stochastic perturbations of model weights and hyperparameters. We leverage recent advances in contrastive representation learning to create models that separately capture opposing hypotheses of invariant and equivariant functional classes and present a simple ensembling approach to efficiently combine appropriate hypotheses for a given task. We show that MSE effectively captures the multiplicity of conflicting hypotheses that is often required in large, diverse datasets like ImageNet. As a result of their inherent diversity, MSE improves classification performance, uncertainty quantification, and generalization across a series of transfer tasks.