Classification with asymmetric label noise: consistency and maximal denoising

DOI10.1214/16-EJS1193zbMATH Open1347.62106arXiv1303.1208OpenAlexW2963314381MaRDI QIDQ315419FDOQ315419

Authors: Gilles Blanchard, Marek Flaska, Clayton Scott, Gregory Handy, Sara Pozzi 

Publication date: 21 September 2016

Published in: Electronic Journal of Statistics (Search for Journal in Brave)

Abstract: In many real-world classification problems, the labels of training examples are randomly corrupted. Most previous theoretical work on classification with label noise assumes that the two classes are separable, that the label noise is independent of the true class label, or that the noise proportions for each class are known. In this work, we give conditions that are necessary and sufficient for the true class-conditional distributions to be identifiable. These conditions are weaker than those analyzed previously, and allow for the classes to be nonseparable and the noise levels to be asymmetric and unknown. The conditions essentially state that a majority of the observed labels are correct and that the true class-conditional distributions are "mutually irreducible," a concept we introduce that limits the similarity of the two distributions. For any label noise problem, there is a unique pair of true class-conditional distributions satisfying the proposed conditions, and we argue that this pair corresponds in a certain sense to maximal denoising of the observed distributions. Our results are facilitated by a connection to "mixture proportion estimation," which is the problem of estimating the maximal proportion of one distribution that is present in another. We establish a novel rate of convergence result for mixture proportion estimation, and apply this to obtain consistency of a discrimination rule based on surrogate loss minimization. Experimental results on benchmark data and a nuclear particle classification problem demonstrate the efficacy of our approach.

Full work available at URL: https://arxiv.org/abs/1303.1208

Recommendations

• Corrigendum to: ``Classification with asymmetric label noise: consistency and maximal denoising
• scientific article; zbMATH DE number 7306869
• How to handle noisy labels for robust learning from uncertainty
• Classification with label noise: a Markov chain sampling framework
• On classifier behavior in the presence of mislabeling noise
• Asymmetric Error Control Under Imperfect Supervision: A Label-Noise-Adjusted Neyman–Pearson Umbrella Algorithm
• Learning from binary labels with instance-dependent noise
• scientific article; zbMATH DE number 6982911
• Margin-based generalization for classifications with input noise

zbMATH Keywords

classificationconsistencylabel noisemixture proportion estimationsurrogate loss

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Pattern recognition, speech recognition (68T10)

Cites Work

• Support Vector Machines
• Title not available (Why is that?)
• Title not available (Why is that?)
• Convexity, Classification, and Risk Bounds
• Strong uniform times and finite random walks
• Efficient noise-tolerant learning from statistical queries
• Foundations of machine learning
• Title not available (Why is that?)
• Calibrated asymmetric surrogate losses
• Random classification noise defeats all convex potential boosters
• Robust supervised classification with mixture models: learning from data with uncertain labels
• Composite binary losses
• Semi-supervised novelty detection
• Tutorial on practical prediction theory for classification
• Classification with asymmetric label noise: consistency and maximal denoising
• Noise-tolerant distribution-free learning of general geometric concepts
• Title not available (Why is that?)
• Multi-instance learning with any hypothesis class
• Boosting in the presence of noise

Cited In (20)

• Boosting in the presence of outliers: adaptive classification with nonconvex loss functions
• Adaptive transfer learning
• Classification trees with mismeasured responses
• A theory of learning with corrupted labels
• Harmless label noise and informative soft-labels in supervised classification
• Binary classification with corrupted labels
• Classification with imperfect training labels
• Title not available (Why is that?)
• Classification with asymmetric label noise: consistency and maximal denoising
• Principled analytic classifier for positive-unlabeled learning via weighted integral probability metric
• Asymmetric Error Control Under Imperfect Supervision: A Label-Noise-Adjusted Neyman–Pearson Umbrella Algorithm
• Estimating the class prior for positive and unlabelled data via logistic regression
• Decontamination of mutual contamination models
• Learning from binary labels with instance-dependent noise
• Corrigendum to: ``Classification with asymmetric label noise: consistency and maximal denoising
• Title not available (Why is that?)
• Positive-unlabeled classification under class-prior shift: a prior-invariant approach based on density ratio estimation
• Classification with label noise: a Markov chain sampling framework
• On the noise estimation statistics
• Learning from positive and unlabeled data: a survey

This page was built for publication: Classification with asymmetric label noise: consistency and maximal denoising

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q315419)