Rethinking Weak Supervision in Helping Contrastive Learning

TL;DR

This paper develops a theoretical framework to analyze how weak supervision, like semi-supervised and noisy labels, influences contrastive learning, revealing that semi-supervised labels improve performance while noisy labels have limited impact.

Contribution

It introduces a unified theoretical framework for contrastive learning with weak supervision and analyzes the effects of semi-supervised versus noisy labels on downstream performance.

Findings

Semi-supervised labels improve downstream error bounds.

Noisy labels have limited effects under the studied paradigm.

Theoretical insights challenge assumptions about noisy label utility.

Abstract

Contrastive learning has shown outstanding performances in both supervised and unsupervised learning, and has recently been introduced to solve weakly supervised learning problems such as semi-supervised learning and noisy label learning. Despite the empirical evidence showing that semi-supervised labels improve the representations of contrastive learning, it remains unknown if noisy supervised information can be directly used in training instead of after manual denoising. Therefore, to explore the mechanical differences between semi-supervised and noisy-labeled information in helping contrastive learning, we establish a unified theoretical framework of contrastive learning under weak supervision. Specifically, we investigate the most intuitive paradigm of jointly training supervised and unsupervised contrastive losses. By translating the weakly supervised information into a similarity graph under the framework of spectral clustering based on the posterior probability of weak labels, we establish the downstream classification error bound. We prove that semi-supervised labels improve the downstream error bound whereas noisy labels have limited effects under such a paradigm. Our theoretical findings here provide new insights for the community to rethink the role of weak supervision in helping contrastive learning.