Provably End-to-end Label-Noise Learning without Anchor Points

TL;DR

This paper introduces an end-to-end method for label-noise learning that does not rely on anchor points, using a novel optimization approach to identify the transition matrix under mild conditions, ensuring consistent classifiers.

Contribution

The paper presents a new anchor-point-free framework for estimating the transition matrix in label-noise learning, enabling statistically consistent classification.

Findings

Effective on benchmark datasets

Robust to various noise levels

Identifies transition matrix under mild assumptions

Abstract

In label-noise learning, the transition matrix plays a key role in building statistically consistent classifiers. Existing consistent estimators for the transition matrix have been developed by exploiting anchor points. However, the anchor-point assumption is not always satisfied in real scenarios. In this paper, we propose an end-to-end framework for solving label-noise learning without anchor points, in which we simultaneously optimize two objectives: the cross entropy loss between the noisy label and the predicted probability by the neural network, and the volume of the simplex formed by the columns of the transition matrix. Our proposed framework can identify the transition matrix if the clean class-posterior probabilities are sufficiently scattered. This is by far the mildest assumption under which the transition matrix is provably identifiable and the learned classifier is statistically consistent. Experimental results on benchmark datasets demonstrate the effectiveness and robustness of the proposed method.