Weak adversarial networks solve NS-Brinkman equations, outperforming PINNs

FrnkNlsn · x · 2026-08-20

A paper in Information Geometry proposes using Weak Adversarial Networks (WAN) to solve 2D incompressible Navier–Stokes-Brinkman equations. Leveraging weak solution theory, the method transforms PDE solving into a Min-Max optimization problem addressed by GANs. The study introduces a biharmonic formulation and extends the WAN framework (WAN2DNSB). Results demonstrate that the approach offers better stability and accuracy than traditional Physics-Informed Neural Networks (PINNs) and PINOs.

Original post →

More from Research

Research channel →