Data-Enabled Poisson Equation Solver Using Multiple Input Artificial Neural Networks (ANNs)

摘要

We explore the possibility of using multiple input artificial neural networks for the solution of Poisson Equation. Results are shown for 1D case considering a boundary value problem with Dirichlet boundary conditions. As opposed to previous work, we consider the cases where input to the ANN is provided at multiple input nodes. A typical variation is considered containing sinc and cosine type functions and output validation is shown for different discretization steps. Time and error performances of the algorithm are presented.