{"id":"arxiv/2611.22099","url":"https://arxiv.org/abs/2611.22099","tags":["math.NA","cs.MS"],"score":0,"title":"Differentiable High-Order Mesh $r$-Adaptivity","domain":"math.NA","source":"arxiv","authors":["Ketan Mittal","Julian Andrej","John Camier","Veselin Dobrev","Tzanio Kolev","Leonardo Molinari","Vladimir Tomov"],"reviews":0,"subject":"math.NA","summary":"High-order mesh r-adaptivity typically entails minimizing a nonlinear functional using its gradients and, for Newton-based methods, its Hessian. These derivatives can be expensive to derive, must be updated with any changes to the functional, and errors in the derivatives directly impact the convergence of r-adaptivity. Additionally, derivatives of the r-adaptivity operator are required for consistent sensitivity analysis for inverse problems. This work develops a framework for differentiable r-adaptivity based on the Target-Matrix Optimization Paradigm. Given a pointwise functional defining…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-17","reactions":{}}