Interactive reweighting for mitigating label quality issues

Yang, Weikai, Guo, Yukai, Wu, Jing ORCID: https://orcid.org/0000-0001-5123-9861, Wang, Zheng, Guo, Lan-Zhe, Li, Yu-Feng and Liu, Shixia 2024. Interactive reweighting for mitigating label quality issues. IEEE Transactions on Visualization and Computer Graphics 30 (3) , pp. 1837-1852. 10.1109/TVCG.2023.3345340

Preview

PDF - Accepted Post-Print Version Download (5MB) | Preview

Abstract

Label quality issues, such as noisy labels and imbalanced class distributions, have negative effects on model performance. Automatic reweighting methods identify problematic samples with label quality issues by recognizing their negative effects on validation samples and assigning lower weights to them. However, these methods fail to achieve satisfactory performance when the validation samples are of low quality. To tackle this, we develop Reweighter, a visual analysis tool for sample reweighting. The reweighting relationships between validation samples and training samples are modeled as a bipartite graph. Based on this graph, a validation sample improvement method is developed to improve the quality of validation samples. Since the automatic improvement may not always be perfect, a co-cluster-based bipartite graph visualization is developed to illustrate the reweighting relationships and support the interactive adjustments to validation samples and reweighting results. The adjustments are converted into the constraints of the validation sample improvement method to further improve validation samples. We demonstrate the effectiveness of Reweighter in improving reweighting results through quantitative evaluation and two case studies.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Computer Science & Informatics
Publisher:	Institute of Electrical and Electronics Engineers
ISSN:	1077-2626
Date of First Compliant Deposit:	19 January 2024
Date of Acceptance:	18 December 2023
Last Modified:	26 Mar 2024 16:29
URI:	https://orca.cardiff.ac.uk/id/eprint/165688

Actions (repository staff only)

Edit Item