Learning from Binary Labels with Instance-Dependent Corruption

Abstract

Suppose we have a sample of instances paired with binary labels corrupted by arbitrary instance- and label-dependent noise. With sufficiently many such samples, can we optimally classify and rank instances with respect to the noise-free distribution? We provide a theoretical analysis of this question, with three main contributions. First, we prove that for instance-dependent noise, any algorithm that is consistent for classification on the noisy distribution is also consistent on the clean distribution. Second, we prove that for a broad class of instance- and label-dependent noise, a similar consistency result holds for the area under the ROC curve. Third, for the latter noise model, when the noise-free class-probability function belongs to the generalised linear model family, we show that the Isotron can efficiently and provably learn from the corrupted sample.

References in corpus (4)

• Noise Tolerance under Risk Minimization
• Surrogate Regret Bounds for Bipartite Ranking via Strongly Proper Losses
• Loss factorization, weakly supervised learning and label noise robustness
• Efficient Learning of Linear Separators under Bounded Noise

Cited by in corpus (9)

• Decoupling "when to update" from "how to update"
• Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noise
• Confidence Scores Make Instance-dependent Label-noise Learning Possible
• L_DMI: An Information-theoretic Noise-robust Loss Function
• Fairness Evaluation in Presence of Biased Noisy Labels
• Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization
• Importance Reweighting for Biquality Learning
• A Simple yet Effective Baseline for Robust Deep Learning with Noisy Labels
• Learning from Noisy Similar and Dissimilar Data