Structured Ensembles: an Approach to Reduce the Memory Footprint of Ensemble Methods

TL;DR

This paper introduces Structured Ensembles, a novel method to extract diverse sub-networks from a single untrained neural network, significantly reducing memory requirements while maintaining or improving accuracy and uncertainty calibration.

Contribution

The paper presents a new end-to-end optimization approach for creating memory-efficient neural network ensembles by extracting sub-structures, with applications to continual learning.

Findings

Achieves comparable or higher accuracy than existing methods.

Requires significantly less memory for ensemble storage.

Performs well in uncertainty estimation and continual learning scenarios.

Abstract

In this paper, we propose a novel ensembling technique for deep neural networks, which is able to drastically reduce the required memory compared to alternative approaches. In particular, we propose to extract multiple sub-networks from a single, untrained neural network by solving an end-to-end optimization task combining differentiable scaling over the original architecture, with multiple regularization terms favouring the diversity of the ensemble. Since our proposal aims to detect and extract sub-structures, we call it Structured Ensemble. On a large experimental evaluation, we show that our method can achieve higher or comparable accuracy to competing methods while requiring significantly less storage. In addition, we evaluate our ensembles in terms of predictive calibration and uncertainty, showing they compare favourably with the state-of-the-art. Finally, we draw a link with the continual learning literature, and we propose a modification of our framework to handle continuous streams of tasks with a sub-linear memory cost. We compare with a number of alternative strategies to mitigate catastrophic forgetting, highlighting advantages in terms of average accuracy and memory.