Regularized maximum mean discrepancy for variable importance measure
Statistics and Computing · 2026-08 · 36 (4)
DOI: 10.1007/s11222-026-10927-4 ↗Kurz gefasst
Fachartikel in der Zeitschrift Statistics and Computing, online erschienen am 10.07.2026. Verfasst von Junfeng Huo und 3 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: University of Melbourne.
Abstract (Originalsprache)
This paper introduced a novel variable importance measure within the framework of maximum mean discrepancy (MMD). The proposed method assigns weights to each variable and optimizes these weights through a regularized MMD criterion, allowing them to adapt to the signal strength of each variable. These optimized weights quantify the contribution of each variable to the discrepancy between two distributions, thereby serving as a measure of variable importance. We further develop an object-oriented variable selection approach, where the selected variables–determined by the optimized weights–are tailored to minimize a task-specific loss function. Focusing on two common scenarios–two-sample testing and classification–we demonstrate that the proposed method enhances the power of MMD-based tests and improves classification accuracy. We establish consistency of the estimated weights, and extensive simulations and real-data applications confirm the practical effectiveness of our approach.
Förderung
- University of Melbourne
Zitieren
APA
Huo, J., Huang, B., Liu, Y., & Peng, L. (2026). Regularized maximum mean discrepancy for variable importance measure. Statistics and Computing, 36(4). https://doi.org/10.1007/s11222-026-10927-4
Vancouver
Huo J, Huang B, Liu Y, Peng L. Regularized maximum mean discrepancy for variable importance measure. Statistics and Computing. 2026;36(4). doi:10.1007/s11222-026-10927-4
BibTeX
@article{Huo2026,
author = {Huo, Junfeng and Huang, Bingyao and Liu, Yanyan and Peng, Liuhua},
title = {Regularized maximum mean discrepancy for variable importance measure},
journal = {Statistics and Computing},
year = {2026},
volume = {36},
number = {4},
doi = {10.1007/s11222-026-10927-4}
}
EUVIMED