Subinterval sensitivity for high dimensional models

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

Abstract

This paper introduces an interval-based non-probabilistic sensitivity analysis method, named subinterval sensitivity. A powerful, reliable and rigorous sensitivity analysis method, which is best suited to quantify the importance of inputs purely with respect to their mathematical model. The method has only recently and partially appeared in the literature, while its scalability to high-dimensional models is claimed here for the first time. We apply subinterval sensitivity to quantify and rank the importance of the parameters of a trained neural network model while drawing comparisons with the established Sobol' sensitivity analysis method. Sensitivities on the parameters of a trained neural network can shed light on overparametrization and explainability of the neural network surrogate model.
Original languageEnglish
Title of host publicationProc. of the 35th European Safety and Reliability & 33rd Society for Risk Analysis Europe Conference
Place of PublicationSingapore
Number of pages8
Publication statusAccepted/In press - 15 Feb 2025
EventESREL & SRA-E 2025: ESREL SRA-E 2025 - University of Stavanger, Stavanger, Norway
Duration: 15 Jun 202519 Jun 2025
https://esrel2025.com/

Conference

ConferenceESREL & SRA-E 2025
Abbreviated titleESREL SRA-E 2025
Country/TerritoryNorway
CityStavanger
Period15/06/2519/06/25
Internet address

Funding

This research is funded by the University of Strathclyde’s StrathDRUMS centre for doctoral training.

Keywords

  • sensitivity analysis
  • interval computation
  • subinterval reconstitution
  • high-dimensional models

Fingerprint

Dive into the research topics of 'Subinterval sensitivity for high dimensional models'. Together they form a unique fingerprint.

Cite this