arXiv · 2611.21854
Kernel Tuning for the HSIC-Based Global Sensitivity Analysis of High-Dimensional Models
Abstract
We consider global sensitivity analysis of models with high-dimensional input parameters using HSIC-based indices. The numerical estimation of these kernel-based indices is known to be challenging for such models. Without proper kernel tuning, the standard estimator for the HSIC can become unstable and break down. Focusing on models with independent parameters, we introduce a dimension-aware bandwidth formula and provide theoretical analysis justifying the proposed approach. In particular, we show that unless kernel bandwidth parameters explicitly scale with the input dimension, the corresponding sensitivity indices return uninformative rankings of parameter influence. Further, we demonstrate the effectiveness of the proposed scaling strategy on linear and nonlinear test problems, including models with correlated inputs.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Troy Larsen, Alen Alexanderian. 2026-11-16. Kernel Tuning for the HSIC-Based Global Sensitivity Analysis of High-Dimensional Models. https://arxiv.science/abs/2611.21854
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.