Estimates on the generalization error of Physics Informed Neural Networks (PINNs) for approximating a class of inverse problems for PDEs
This work provides theoretical justification for PINNs in inverse PDE problems, which is important for researchers in computational science and engineering, though it is incremental as it builds on existing PINN frameworks with new error bounds.
The authors tackled the problem of quantifying the generalization error of Physics Informed Neural Networks (PINNs) for approximating inverse problems in PDEs, specifically data assimilation or unique continuation problems, by proving rigorous estimates and validating them with numerical experiments on four linear PDE examples.
Physics informed neural networks (PINNs) have recently been very successfully applied for efficiently approximating inverse problems for PDEs. We focus on a particular class of inverse problems, the so-called data assimilation or unique continuation problems, and prove rigorous estimates on the generalization error of PINNs approximating them. An abstract framework is presented and conditional stability estimates for the underlying inverse problem are employed to derive the estimate on the PINN generalization error, providing rigorous justification for the use of PINNs in this context. The abstract framework is illustrated with examples of four prototypical linear PDEs. Numerical experiments, validating the proposed theory, are also presented.