Classification with Imperfect Training Labels

• Activity Number: 634 - Recent Advancements in Distance and Kernel-Based Metrics and Related Learning Methods
• Type: Invited
• Date/Time: Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Timothy I. Cannings and Yingying Fan and Richard Samworth*
• Companies: University of Edinburgh and University of Southern California and University of Cambridge
• Keywords: Classification; k-nearest neighbour classifier; Support Vector Machine; Linear disriminant analysis; Noisy labels

Abstract

We study the effect of imperfect training data labels on the performance of classification methods. When 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. Under stronger conditions, we derive detailed asymptotic properties for the popular $k$-nearest neighbour (knn), support vector machine (SVM) and linear discriminant analysis (LDA) classifiers. One consequence 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, 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.