Radial-Angular Geometry for Reliable Update Diagnosis in Noisy-Label Learning

TL;DR

This paper introduces a novel geometric approach, RGC, for diagnosing label reliability in noisy-label learning by comparing observed-label gradients with reference gradients, improving robustness and accuracy.

Contribution

It proposes the Relative Geometric Conflict (RGC) method that leverages a backward-space measure to better distinguish between clean and corrupted label updates.

Findings

RGC improves accuracy on noisy-label benchmarks.

It better preserves clean hard samples compared to existing methods.

The approach effectively distinguishes conflicting updates caused by label noise.

Abstract

Noisy-label methods often estimate sample reliability from forward-space signals such as loss, confidence, or entropy. These signals indicate whether a sample is difficult to predict, but they do not directly test whether its observed label induces a reliable parameter update. This gap matters because hard clean samples and mislabeled samples can have similar loss while inducing different updates. We recast reliability estimation as diagnosis of the observed-label update. The sample-wise empirical Fisher trace gives a backward-space measure of update energy: for the classifier layer, it factorizes into a prediction-residual term and a feature-sensitivity term, so it captures information beyond scalar loss. Trace, however, is still a radial magnitude signal and cannot decide whether a large update is useful or harmful. We therefore propose Relative Geometric Conflict (RGC), which compares the observed-label gradient with a reference gradient induced by an EMA teacher. The conflict term helps distinguish large but aligned hard-clean updates from large conflicting updates caused by corrupted labels. Across synthetic and real-world noisy-label benchmarks, RGC improves hard-clean preservation and accuracy under our evaluation protocol.