Clusterability as an Alternative to Anchor Points When Learning with Noisy Labels

Abstract

The label noise transition matrix, characterizing the probabilities of a training instance being wrongly annotated, is crucial to designing popular solutions to learning with noisy labels. Existing works heavily rely on finding "anchor points" or their approximates, defined as instances belonging to a particular class almost surely. Nonetheless, finding anchor points remains a non-trivial task, and the estimation accuracy is also often throttled by the number of available anchor points. In this paper, we propose an alternative option to the above task. Our main contribution is the discovery of an efficient estimation procedure based on a clusterability condition. We prove that with clusterable representations of features, using up to third-order consensuses of noisy labels among neighbor representations is sufficient to estimate a unique transition matrix. Compared with methods using anchor points, our approach uses substantially more instances and benefits from a much better sample complexity. We demonstrate the estimation accuracy and advantages of our estimates using both synthetic noisy labels (on CIFAR-10/100) and real human-level noisy labels (on Clothing1M and our self-collected human-annotated CIFAR-10). Our code and human-level noisy CIFAR-10 labels are available at https://github.com/UCSC-REAL/HOC.

arXiv commentICML 2021

References in corpus (9)

• Classification with Noisy Labels by Importance Reweighting
• Are Anchor Points Really Indispensable in Label-Noise Learning?
• Part-dependent Label Noise: Towards Instance-dependent Label Noise
• A Survey of Label-noise Representation Learning: Past, Present and Future
• Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels
• Provably End-to-end Label-Noise Learning without Anchor Points
• Extended T: Learning with Mixed Closed-set and Open-set Noisy Labels
• Learning Adaptive Loss for Robust Learning with Noisy Labels
• On the Resistance of Nearest Neighbor to Random Noisy Labels

Cited by in corpus (7)

• Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations
• Instance-dependent Label-noise Learning under a Structural Causal Model
• Learning with Instance-Dependent Label Noise: A Sample Sieve Approach
• A Second-Order Approach to Learning with Instance-Dependent Label Noise
• When Optimizing $f$-divergence is Robust with Label Noise
• Policy Learning Using Weak Supervision
• An Empirical Study and Analysis on Open-Set Semi-Supervised Learning