A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks.