Extending Momentum Contrast with Cross Similarity Consistency Regularization

TL;DR

This paper introduces XMoCo, a self-supervised learning method that enhances contrastive learning by regularizing cross-similarity consistency among both positive and negative pairs, leading to improved representation quality.

Contribution

The paper proposes a novel cross similarity regularization for contrastive learning, extending the momentum contrast framework to better utilize negative pairs and improve downstream task performance.

Findings

Achieves competitive results on ImageNet-1K classification.

Improves downstream task performance with the proposed regularization.

Plug-and-play extension compatible with existing self-supervised methods.

Abstract

Contrastive self-supervised representation learning methods maximize the similarity between the positive pairs, and at the same time tend to minimize the similarity between the negative pairs. However, in general the interplay between the negative pairs is ignored as they do not put in place special mechanisms to treat negative pairs differently according to their specific differences and similarities. In this paper, we present Extended Momentum Contrast (XMoCo), a self-supervised representation learning method founded upon the legacy of the momentum-encoder unit proposed in the MoCo family configurations. To this end, we introduce a cross consistency regularization loss, with which we extend the transformation consistency to dissimilar images (negative pairs). Under the cross consistency regularization rule, we argue that semantic representations associated with any pair of images (positive or negative) should preserve their cross-similarity under pretext transformations. Moreover, we further regularize the training loss by enforcing a uniform distribution of similarity over the negative pairs across a batch. The proposed regularization can easily be added to existing self-supervised learning algorithms in a plug-and-play fashion. Empirically, we report a competitive performance on the standard Imagenet-1K linear head classification benchmark. In addition, by transferring the learned representations to common downstream tasks, we show that using XMoCo with the prevalently utilized augmentations can lead to improvements in the performance of such tasks. We hope the findings of this paper serve as a motivation for researchers to take into consideration the important interplay among the negative examples in self-supervised learning.