Weighted Ensemble Self-Supervised Learning

TL;DR

This paper introduces a data-dependent weighted ensemble framework for self-supervised learning that enhances performance and robustness without additional computational costs, demonstrated on state-of-the-art SSL methods on ImageNet-1K.

Contribution

It proposes a novel ensemble method that improves SSL performance by weighting ensemble heads based on data, without changing model architecture or increasing training overhead.

Findings

Outperforms individual SSL methods like DINO and MSN on ImageNet-1K.

Improves few-shot learning performance, e.g., 3.9 percentage points in 1-shot MSN ViT-B/16.

Increasing diversity among ensemble heads enhances downstream results.

Abstract

Ensembling has proven to be a powerful technique for boosting model performance, uncertainty estimation, and robustness in supervised learning. Advances in self-supervised learning (SSL) enable leveraging large unlabeled corpora for state-of-the-art few-shot and supervised learning performance. In this paper, we explore how ensemble methods can improve recent SSL techniques by developing a framework that permits data-dependent weighted cross-entropy losses. We refrain from ensembling the representation backbone; this choice yields an efficient ensemble method that incurs a small training cost and requires no architectural changes or computational overhead to downstream evaluation. The effectiveness of our method is demonstrated with two state-of-the-art SSL methods, DINO (Caron et al., 2021) and MSN (Assran et al., 2022). Our method outperforms both in multiple evaluation metrics on ImageNet-1K, particularly in the few-shot setting. We explore several weighting schemes and find that those which increase the diversity of ensemble heads lead to better downstream evaluation results. Thorough experiments yield improved prior art baselines which our method still surpasses; e.g., our overall improvement with MSN ViT-B/16 is 3.9 p.p. for 1-shot learning.