Exploiting the Potential Supervision Information of Clean Samples in Partial Label Learning

TL;DR

This paper introduces CleanSE, a calibration strategy leveraging clean samples in partial label learning to improve disambiguation and label confidence, significantly boosting performance across multiple datasets.

Contribution

The paper proposes a novel calibration method called CleanSE that utilizes clean samples to guide disambiguation and enhance existing partial label learning algorithms.

Findings

CleanSE improves performance of state-of-the-art PLL methods.

Effective in both synthetic and real-world datasets.

Enhances label confidence and sample distribution characterization.

Abstract

Diminishing the impact of false-positive labels is critical for conducting disambiguation in partial label learning. However, the existing disambiguation strategies mainly focus on exploiting the characteristics of individual partial label instances while neglecting the strong supervision information of clean samples randomly lying in the datasets. In this work, we show that clean samples can be collected to offer guidance and enhance the confidence of the most possible candidates. Motivated by the manner of the differentiable count loss strat- egy and the K-Nearest-Neighbor algorithm, we proposed a new calibration strategy called CleanSE. Specifically, we attribute the most reliable candidates with higher significance under the assumption that for each clean sample, if its label is one of the candidates of its nearest neighbor in the representation space, it is more likely to be the ground truth of its neighbor. Moreover, clean samples offer help in characterizing the sample distributions by restricting the label counts of each label to a specific interval. Extensive experiments on 3 synthetic benchmarks and 5 real-world PLL datasets showed this calibration strategy can be applied to most of the state-of-the-art PLL methods as well as enhance their performance.