Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems