{"id":"arxiv/2604.01719","url":"https://arxiv.org/abs/2604.01719","tags":["math.NA"],"score":0,"title":"Mixed Consistent PINNs for Elliptic Obstacle Problems with Stability Analysis","domain":"math.NA","source":"arxiv","authors":["Arbaz Khan","Kent-Andre Mardal","Shiv Mishra"],"reviews":0,"subject":"math.NA","summary":"We propose a consistent physics-informed neural networks (CPINNs) framework for elliptic obstacle problems formulated as variational inequalities. The method is based on a mixed loss functional that is rigorously aligned with the stability structure of the underlying problem and incorporates obstacle constraints through a consistent treatment of the associated Lagrange multiplier. Relying on optimal recovery theory under Besov regularity assumptions, we establish near-optimal convergence rates for the simultaneous reconstruction of the solution and the multiplier from pointwise interior and…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-04-02","reactions":{}}