Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning

TL;DR

This paper proposes the dual-T estimator, a novel method that improves transition matrix estimation in label-noise learning by factorizing it through an intermediate class, resulting in more accurate classifiers.

Contribution

The paper introduces the dual-T estimator, which reduces estimation error by avoiding direct noisy class posterior estimation through a divide-and-conquer approach.

Findings

The dual-T estimator achieves more accurate transition matrix estimation.

The method leads to improved classification performance.

Theoretical analysis supports the effectiveness of the approach.

Abstract

The transition matrix, denoting the transition relationship from clean labels to noisy labels, is essential to build statistically consistent classifiers in label-noise learning. Existing methods for estimating the transition matrix rely heavily on estimating the noisy class posterior. However, the estimation error for noisy class posterior could be large due to the randomness of label noise, which would lead the transition matrix to be poorly estimated. Therefore, in this paper, we aim to solve this problem by exploiting the divide-and-conquer paradigm. Specifically, we introduce an intermediate class to avoid directly estimating the noisy class posterior. By this intermediate class, the original transition matrix can then be factorized into the product of two easy-to-estimate transition matrices. We term the proposed method the dual-T estimator. Both theoretical analyses and empirical results illustrate the effectiveness of the dual-T estimator for estimating transition matrices, leading to better classification performances.