Physics-informed Neural Networks (PINNs) have been shown to be effective in solving partial differential equations by capturing the physics induced constraints as a part of the training loss function. This paper shows that a PINN can be sensitive to errors in training data and overfit itself in dynamically propagating these errors over the domain of the solution of the PDE. It also shows how physical regularizations based on continuity criteria and conservation laws fail to address this issue and rather introduce problems of their own causing the deep network to converge to a physics-obeying local minimum instead of the global minimum. We introduce Gaussian Process (GP) based smoothing that recovers the performance of a PINN and promises a robust architecture against noise/errors in measurements. Additionally, we illustrate an inexpensive method of quantifying the evolution of uncertainty based on the variance estimation of GPs on boundary data. Robust PINN performance is also shown to be achievable by choice of sparse sets of inducing points based on sparsely induced GPs. We demonstrate the performance of our proposed methods and compare the results from existing benchmark models in literature for time-dependent Schr\"odinger and Burgers' equations.

Twisted rebar, concrete, and splintered furniture lay scattered across the floor of this room. Our view through a jagged hole in the wall looks out on the city, showing steady civilian traffic crossing a bridge over a river below. This article has been reproduced in a new format and may be missing content or contain faulty links. Contact wiredlabs@wired.com to report an issue. An Army major beside me, Paul Tyrrell, scans the high-rises on the other side of the river through his laser rangefinder. He is the frontline eyes of the coalition, responsible for calling in air strikes. A platoon sergeant named Donald Prado tells Tyrrell that an office tower half a mile to the west is an enemy stronghold. Prado radios in for the Air Force to drop a smoke screen for cover.