Abstract:Global sensitivity analysis is the main quantitative technique for identifying the most influential input variables in a numerical simulation model. In particular when the inputs are independent, Sobol' sensitivity indices attribute a portion of the output of interest variance to each input and all possible interactions in the model, thanks to a functional ANOVA decomposition. On the other hand, moment-independent sensitivity indices focus on the impact of input variables on the whole output distribution instead of the variance only, thus providing complementary insight on the inputs / output relationship. Unfortunately they do not enjoy the nice decomposition property of Sobol' indices and are consequently harder to analyze. In this paper, we introduce two moment-independent indices based on kernel-embeddings of probability distributions and show that the RKHS framework used for their definition makes it possible to exhibit a kernel-based ANOVA decomposition. This is the first time such a desirable property is proved for sensitivity indices apart from Sobol' ones. When the inputs are dependent, we also use these new sensitivity indices as building blocks to design kernel-embedding Shapley effects which generalize the traditional variance-based ones used in sensitivity analysis. Several estimation procedures are discussed and illustrated on test cases with various output types such as categorical variables and probability distributions. All these examples show their potential for enhancing traditional sensitivity analysis with a kernel point of view.

High Order Properties of Kernel Functions and Their Application in Sensitivity Analysis

Design Method and Application with Kernel Function for Pattern Analysis

Frequency-domain Sensitivities with Application to Power System Modeling

A Novel Analytical Second‐order Sensitivity Calculation Approach Using the Finite Element Method for Chemical Engineering Problems

Second-order Reliability Alogorithm and Its Applications Using Response Surface Method

Stability Analysis of Discrete-Time Systems with Arbitrary Delay Kernels Based on Kernel-Related Summation Inequality and Model Transformation

Novel Computational Methods for High-Dimensional Stochastic Sensitivity Analysis

Sensitivity Analysis of Random and Interval Uncertain Variables Based on Polynomial Chaos Expansion Method

Reliability sensitivity analysis for structural systems in interval probability form

Vine copula-based parametric sensitivity analysis of failure probability-based importance measure in the presence of multidimensional dependencies

Inverse Sensitivity Method for Asymmetric Multivariable Nondefective Systems

Higher-order Eigensensitivity Analysis of a Defective Matrix

A Novel Bayesian-inference-based Method for Global Sensitivity Analysis of System Reliability with Multiple Failure Modes

Uncertainty analysis method based on a combination of the maximum entropy principle and the point estimation method | Metoda analizy niepewności oparta na poła{ogonek}czeniu zasady maksymalnej entropii i metody oceny punktowej

Kernel-based ANOVA decomposition and Shapley effects -- Application to global sensitivity analysis

Non-parametric Kernel Estimation for the ANOVA Decomposition and Sensitivity Analysis.

UNCERTAINTY ANALYSIS METHOD BASED ON A COMBINATION OF THE MAXIMUM ENTROPY PRINCIPLE AND THE POINT ESTIMATION METHOD

An improved high order moment-based saddlepoint approximation method for reliability analysis

A points estimation and series approximation method for uncertainty analysis

Sensitivity Analysis of High-Dimensional Models with Correlated Inputs

An efficient hybrid reliability analysis method for structures involving random and interval variables