Improving Pre-trained Self-Supervised Embeddings Through Effective Entropy Maximization

TL;DR

This paper introduces a new entropy maximization criterion (E2MC) for self-supervised learning that improves downstream task performance by focusing on low-dimensional constraints, overcoming high-dimensional entropy estimation challenges.

Contribution

The paper proposes E2MC, an effective low-dimensional entropy maximization method that enhances pre-trained SSL models with minimal additional training.

Findings

E2MC improves downstream performance after few epochs.

Continued training with E2MC outperforms other criteria.

Performance gains are validated through ablation studies.

Abstract

A number of different architectures and loss functions have been applied to the problem of self-supervised learning (SSL), with the goal of developing embeddings that provide the best possible pre-training for as-yet-unknown, lightly supervised downstream tasks. One of these SSL criteria is to maximize the entropy of a set of embeddings in some compact space. But the goal of maximizing the embedding entropy often depends -- whether explicitly or implicitly -- upon high dimensional entropy estimates, which typically perform poorly in more than a few dimensions. In this paper, we motivate an effective entropy maximization criterion (E2MC), defined in terms of easy-to-estimate, low-dimensional constraints. We demonstrate that using it to continue training an already-trained SSL model for only a handful of epochs leads to a consistent and, in some cases, significant improvement in downstream performance. We perform careful ablation studies to show that the improved performance is due to the proposed add-on criterion. We also show that continued pre-training with alternative criteria does not lead to notable improvements, and in some cases, even degrades performance.