IRNet: Iterative Refinement Network for Noisy Partial Label Learning

TL;DR

IRNet is a novel iterative framework that improves noisy partial label learning by detecting and correcting noisy samples, effectively reducing dataset noise and enhancing classification accuracy.

Contribution

This paper introduces IRNet, a new iterative refinement approach for noisy PLL that includes noise detection and label correction modules, with theoretical guarantees and practical effectiveness.

Findings

IRNet outperforms existing methods on benchmark datasets.

IRNet effectively reduces noise levels in datasets.

Theoretical analysis confirms IRNet's ability to approximate the Bayes optimal classifier.

Abstract

Partial label learning (PLL) is a typical weakly supervised learning, where each sample is associated with a set of candidate labels. Its basic assumption is that the ground-truth label must be in the candidate set, but this assumption may not be satisfied due to the unprofessional judgment of annotators. Therefore, we relax this assumption and focus on a more general task, noisy PLL, where the ground-truth label may not exist in the candidate set. To address this challenging task, we propose a novel framework called ``Iterative Refinement Network (IRNet)'', aiming to purify noisy samples through two key modules (i.e., noisy sample detection and label correction). To achieve better performance, we exploit smoothness constraints to reduce prediction errors in these modules. Through theoretical analysis, we prove that IRNet is able to reduce the noise level of the dataset and eventually approximate the Bayes optimal classifier. Meanwhile, IRNet is a plug-in strategy that can be integrated with existing PLL approaches. Experimental results on multiple benchmark datasets show that IRNet outperforms state-of-the-art approaches on noisy PLL. Our source code is available at: https://github.com/zeroQiaoba/IRNet.