| Abstract: |
| Machine learning is used as an accessible meshless approach for solving PDEs. For interface problem, the computational domain can be decomposed into several subdomains, and the solution on each subdomain can be accordingly represented by one network. In this talk, we will present, 1) a Physics-Informed Neural Network, interfaced neural network (INN) to solve interface problems with discontinuous coefficients as well as irregular interfaces, 2) an interfaced operator network (IONet) to solve parametric elliptic interface PDEs, where different coefficients, source terms, and boundary conditions are considered as input features. The convergence of the INN will be discussed as well. |
|