discretely-trained physics-informed neural networks using meshless radial basis function-finite differences - 42Papers