Building Trust in PINNs: Error Estimation through Finite Difference Methods

TL;DR

This paper introduces a simple, post-hoc finite difference-based method to estimate pointwise errors in PINN predictions for linear PDEs, enhancing trust and interpretability without needing the true solution.

Contribution

It presents a novel, lightweight approach for error estimation in PINNs that does not require prior knowledge of the true solution, improving model validation.

Findings

Accurate error maps produced at low computational cost

Method effectively identifies where PINNs predictions deviate from true solutions

Applicable to several benchmark PDEs

Abstract

Physics-informed neural networks (PINNs) constitute a flexible deep learning approach for solving partial differential equations (PDEs), which model phenomena ranging from heat conduction to quantum mechanical systems. Despite their flexibility, PINNs offer limited insight into how their predictions deviate from the true solution, hindering trust in their prediction quality. We propose a lightweight post-hoc method that addresses this gap by producing pointwise error estimates for PINN predictions, which offer a natural form of explanation for such models, identifying not just whether a prediction is wrong, but where and by how much. For linear partial differential equations, the error between a PINN approximation and the true solution satisfies the same differential operator as the original problem, but driven by the PINN's PDE residual as its source term. We solve this error equation numerically using finite difference methods requiring no knowledge of the true solution. Evaluated on several benchmark PDEs, our method yields accurate error maps at low computational cost, enabling targeted and interpretable validation of PINNs.