{"id":"arxiv/2609.27548","url":"https://arxiv.org/abs/2609.27548","tags":["math.NA"],"score":0,"title":"Stable time-stepping via residual minimization: finite-element and neural-network approximations for transient parabolic problems","domain":"math.NA","source":"arxiv","authors":["Judit Muñoz-Matute","Sergio Rojas"],"reviews":0,"subject":"math.NA","summary":"We propose a time-stepping minimum-residual (MinRes) framework for transient coercive variational problems, applicable to finite-element and neural-network trial approximations. For the Backward Euler scheme, we measure the residual at each time level in the dual norm induced by the steady test-space norm, scaled by the time step. This choice yields stability estimates controlling a discrete parabolic energy quantity, up to the time-discretization defect. We develop conforming and broken-test formulations and specialize the analysis to diffusion-advection-reaction problems. For…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-23","reactions":{}}