Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels

TL;DR

This paper introduces Rank Pruning, a fast and general method for robust binary classification with noisy labels, capable of accurately estimating noise rates and achieving state-of-the-art results on standard datasets.

Contribution

Rank Pruning is a novel, efficient approach that estimates noise rates and maintains optimal risk in noisy label scenarios, outperforming prior methods.

Findings

RP achieves state-of-the-art noise estimation accuracy.

RP with CNN predicts MNIST digits with less than 0.5% error under high noise.

RP performs well even with noise from a third distribution.

Abstract

Noisy PN learning is the problem of binary classification when training examples may be mislabeled (flipped) uniformly with noise rate rho1 for positive examples and rho0 for negative examples. We propose Rank Pruning (RP) to solve noisy PN learning and the open problem of estimating the noise rates, i.e. the fraction of wrong positive and negative labels. Unlike prior solutions, RP is time-efficient and general, requiring O(T) for any unrestricted choice of probabilistic classifier with T fitting time. We prove RP has consistent noise estimation and equivalent expected risk as learning with uncorrupted labels in ideal conditions, and derive closed-form solutions when conditions are non-ideal. RP achieves state-of-the-art noise estimation and F1, error, and AUC-PR for both MNIST and CIFAR datasets, regardless of the amount of noise and performs similarly impressively when a large portion of training examples are noise drawn from a third distribution. To highlight, RP with a CNN classifier can predict if an MNIST digit is a "one"or "not" with only 0.25% error, and 0.46 error across all digits, even when 50% of positive examples are mislabeled and 50% of observed positive labels are mislabeled negative examples.