KANs for Free-Boundary PDEs
What it is
This paper proposes using Kolmogorov-Arnold Networks (KANs) to solve free-boundary partial differential equations within a physics-informed framework. It incorporates obstacle constraints, PDE inequalities, and complementarity conditions via residual-based loss functions to handle problems like the linear elliptic obstacle problem and the Stefan problem.
The problem it solves
Standard physics-informed neural networks (PINNs) may struggle to accurately resolve contact regions and moving interfaces in free-boundary problems due to representation limitations.
What you could use it for
- Solving linear elliptic obstacle problems
- Resolving nonlinear p-Laplacian obstacle problems
- Simulating time-dependent one-phase Stefan problems
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.