Gaussian linear approximation for the estimation of the Shapley effects - Archive ouverte HAL Access content directly
Journal Articles SIAM/ASA Journal on Uncertainty Quantification Year : 2021

Gaussian linear approximation for the estimation of the Shapley effects

Abstract

In this paper, we address the estimation of the sensitivity indices called "Shapley effects". These sensitivity indices enable to handle dependent input variables. The Shapley effects are generally difficult to estimate, but they are easily computable in the Gaussian linear framework. The aim of this work is to use the values of the Shapley effects in an approximated Gaussian linear framework as estimators of the true Shapley effects corresponding to a non-linear model. First, we assume that the input variables are Gaussian with small variances. We provide rates of convergence of the estimated Shapley effects to the true Shapley effects. Then, we focus on the case where the inputs are given by an non-Gaussian empirical mean. We prove that, under some mild assumptions, when the number of terms in the empirical mean increases, the difference between the true Shapley effects and the estimated Shapley effects given by the Gaussian linear approximation converges to 0. Our theoretical results are supported by numerical studies, showing that the Gaussian linear approximation is accurate and enables to decrease the computational time signifficantly.
Fichier principal
Vignette du fichier
main.pdf (571 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02737005 , version 1 (02-06-2020)

Identifiers

Cite

Baptiste Broto, François Bachoc, Marine Depecker, Jean-Marc Martinez. Gaussian linear approximation for the estimation of the Shapley effects. SIAM/ASA Journal on Uncertainty Quantification, 2021, 9, pp.1132-1151. ⟨10.1137/20M1342884⟩. ⟨hal-02737005⟩
60 View
94 Download

Altmetric

Share

Gmail Facebook X LinkedIn More