Classification with imperfect training labels

DOI10.1093/BIOMET/ASAA011MaRDI QIDQ5113013zbMATH OpenOpenAlexFDO

Authors Timothy I. Cannings, Yingying Fan, Richard Samworth

Publication date 9 June 2020

Published in Biometrika (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1805.11505

zbMATH Keywords

label noiselinear discriminant analysismisclassification errornearest neighbourstatistical learningsupport vector machine

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Learning and adaptive systems in artificial intelligence (68T05)

Abstract: We study the effect of imperfect training data labels on the performance of classification methods. In a general setting, where the probability that an observation in the training dataset is mislabelled may depend on both the feature vector and the true label, we bound the excess risk of an arbitrary classifier trained with imperfect labels in terms of its excess risk for predicting a noisy label. This reveals conditions under which a classifier trained with imperfect labels remains consistent for classifying uncorrupted test data points. Furthermore, under stronger conditions, we derive detailed asymptotic properties for the popular $k$-nearest neighbour ($k$nn), support vector machine (SVM) and linear discriminant analysis (LDA) classifiers. One consequence of these results is that the knn and SVM classifiers are robust to imperfect training labels, in the sense that the rate of convergence of the excess risks of these classifiers remains unchanged; in fact, our theoretical and empirical results even show that in some cases, imperfect labels may improve the performance of these methods. On the other hand, the LDA classifier is shown to be typically inconsistent in the presence of label noise unless the prior probabilities of each class are equal. Our theoretical results are supported by a simulation study.

Recommendations

• scientific article; zbMATH DE number 1759567
• Classification with asymmetric label noise: consistency and maximal denoising
• Bayesian classification using a noninformative prior and mislabeled training data
• Learning kernel logistic regression in the presence of class label noise
• scientific article; zbMATH DE number 6982911

Cited in

(19)

• Learning kernel logistic regression in the presence of class label noise
• Adaptive transfer learning
• Cautious classification based on belief functions theory and imprecise relabelling
• All-in-one robust estimator of the Gaussian mean
• Binary classification with corrupted labels
• Binary classification with ambiguous training data
• Classification with label noise: a Markov chain sampling framework
• Harmless label noise and informative soft-labels in supervised classification
• scientific article; zbMATH DE number 1759567 (Why is no real title available?)
• Domain generalization by marginal transfer learning
• scientific article; zbMATH DE number 7370646 (Why is no real title available?)
• Improving classifier performance by using fictitious training data? A case study
• Bayesian classification using a noninformative prior and mislabeled training data
• Asymmetric Error Control Under Imperfect Supervision: A Label-Noise-Adjusted Neyman–Pearson Umbrella Algorithm
• Nonparametric classification with missing data
• Measuring diagnostic test performance using imperfect reference tests: a partial identification approach
• An adaptive transfer learning perspective on classification in nonstationary environments
• Adaptive conformal classification with noisy labels
• Robust transfer learning with unreliable source data

This page was built for publication: Classification with imperfect training labels

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5113013)