BatchEnsemble: An Alternative Approach to Efficient Ensemble and Lifelong Learning

TL;DR

BatchEnsemble introduces a computationally efficient ensemble method using shared weights and rank-one matrices, achieving similar accuracy and uncertainty as traditional ensembles with significantly reduced training and inference costs.

Contribution

The paper presents BatchEnsemble, a novel ensemble approach that reduces computational and memory costs while maintaining competitive accuracy, and demonstrates its effectiveness in lifelong learning scenarios.

Findings

3X speedup at test time for ensembles of size 4

3X memory reduction compared to traditional ensembles

Achieves comparable accuracy and uncertainty in various tasks

Abstract

Ensembles, where multiple neural networks are trained individually and their predictions are averaged, have been shown to be widely successful for improving both the accuracy and predictive uncertainty of single neural networks. However, an ensemble's cost for both training and testing increases linearly with the number of networks, which quickly becomes untenable. In this paper, we propose BatchEnsemble, an ensemble method whose computational and memory costs are significantly lower than typical ensembles. BatchEnsemble achieves this by defining each weight matrix to be the Hadamard product of a shared weight among all ensemble members and a rank-one matrix per member. Unlike ensembles, BatchEnsemble is not only parallelizable across devices, where one device trains one member, but also parallelizable within a device, where multiple ensemble members are updated simultaneously for a given mini-batch. Across CIFAR-10, CIFAR-100, WMT14 EN-DE/EN-FR translation, and out-of-distribution tasks, BatchEnsemble yields competitive accuracy and uncertainties as typical ensembles; the speedup at test time is 3X and memory reduction is 3X at an ensemble of size 4. We also apply BatchEnsemble to lifelong learning, where on Split-CIFAR-100, BatchEnsemble yields comparable performance to progressive neural networks while having a much lower computational and memory costs. We further show that BatchEnsemble can easily scale up to lifelong learning on Split-ImageNet which involves 100 sequential learning tasks.