Robust deep neural network surrogate models with uncertainty quantification via adversarial training

Surrogate models have been used to emulate mathematical simulators of physical or biological processes for computational efficiency. High-speed simulation is crucial for conducting uncertainty quantification (UQ) when the simulation must repeat over many randomly sampled input points (aka the Monte Carlo method). A simulator can be so computationally intensive that UQ is only feasible with a surrogate model. Recently, deep neural network (DNN) surrogate models have gained popularity for their state-of-the-art emulation accuracy. However, it is well-known that DNN is prone to severe errors when input data are perturbed in particular ways, the very phenomenon which has inspired great interest in adversarial training. In the case of surrogate models, the concern is less about a deliberate attack exploiting the vulnerability of a DNN but more of the high sensitivity of its accuracy to input directions, an issue largely ignored by researchers using emulation models. In this paper, we show the severity of this issue through empirical studies and hypothesis testing. Furthermore, we adopt methods in adversarial training to enhance the robustness of DNN surrogate models. Experiments demonstrate that our approaches significantly improve the robustness of the surrogate models without compromising emulation accuracy.

This is the peer reviewed version of the following article: [Robust deep neural network surrogate models with uncertainty quantification via adversarial training. Statistical Analysis and Data Mining: The ASA Data Science Journal 16, 3 p295-304 (2023)], which has been published in final form at https://doi.org/10.1002/sam.11610. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions: https://authorservices.wiley.com/author-resources/Journal-Authors/licensing/self-archiving.html#3.

Files

Metadata

Work Title Robust deep neural network surrogate models with uncertainty quantification via adversarial training
Access
Open Access
Creators
  1. Lixiang Zhang
  2. Jia Li
License In Copyright (Rights Reserved)
Work Type Article
Publisher
  1. Statistical Analysis and Data Mining
Publication Date January 4, 2023
Publisher Identifier (DOI)
  1. https://doi.org/10.1002/sam.11610
Deposited March 11, 2024

Versions

Analytics

Collections

This resource is currently not in any collection.

Work History

Version 1
published

  • Created
  • Added 2211.09954.pdf
  • Added Creator Lixiang Zhang
  • Added Creator Jia Li
  • Published
  • Updated Publication Date Show Changes
    Publication Date
    • 2023-06-01
    • 2023-01-04
  • Updated