Solving BDNK diffusion using physics-informed neural networks
The study solved relativistic BDNK diffusion by reformulating its equations into a flux-conservative form and applying two methods: a finite volume scheme and a self-adaptive PINN called SA-PINN-ACTO. The results showed the PINN closely matched the finite volume solution for smooth data, but struggled with discontinuities. This demonstrates that SA-PINN-ACTO is a flexible tool for solving complex physics problems.