Label Noise: Ignorance Is Bliss

TL;DR

This paper introduces a theoretical framework for learning with multi-class, instance-dependent label noise, framing it as domain adaptation under posterior drift, and demonstrates the effectiveness of a noise-ignorant approach in practice.

Contribution

It develops a novel theoretical framework using relative signal strength to analyze label noise and validates the noise-ignorant ERM approach with state-of-the-art results.

Findings

Nearly matching bounds on excess risk using RSS

Support for noise-ignorant ERM in noisy label settings

State-of-the-art performance on CIFAR-N challenge

Abstract

We establish a new theoretical framework for learning under multi-class, instance-dependent label noise. This framework casts learning with label noise as a form of domain adaptation, in particular, domain adaptation under posterior drift. We introduce the concept of \emph{relative signal strength} (RSS), a pointwise measure that quantifies the transferability from noisy to clean posterior. Using RSS, we establish nearly matching upper and lower bounds on the excess risk. Our theoretical findings support the simple \emph{Noise Ignorant Empirical Risk Minimization (NI-ERM)} principle, which minimizes empirical risk while ignoring label noise. Finally, we translate this theoretical insight into practice: by using NI-ERM to fit a linear classifier on top of a self-supervised feature extractor, we achieve state-of-the-art performance on the CIFAR-N data challenge.