Deep Learning with Physics Priors as Generalized Regularizers

Frank Liu, Agniva Chowdhury

Keywords: Structure Risk Minimization, Regularizers, Uncertainties

TL;DR: Leverage knowledge of a physical system as generalized regularizers to minimize structural risk

Abstract: In various scientific and engineering applications, there is typically an approximate model of the underlying complex system, even though it contains both aleatoric and epistemic uncertainties. In this paper, we present a principled method to incorporate these approximate models as physics priors in modeling, to prevent overfitting and enhancing the generalization capabilities of the trained models. Utilizing the structural risk minimization (SRM) inductive principle pioneered by Vapnik, this approach structures the physics priors into generalized regularizers. The experimental results demonstrate that our method achieves up to two orders of magnitude of improvement in testing accuracy.

Submission Track: Original Research

Submission Number: 167