{"id":"arxiv/2610.39845","url":"https://arxiv.org/abs/2610.39845","tags":["math.NA","math.AP"],"score":0,"title":"A Semigroup-Aware Kernel Neural Operator with Latent Exponential Integration for Forced PDEs","domain":"math.NA","source":"arxiv","authors":["Naseem Abbas","Vittorio Colao","Davide Macrì","Matthieu Barreau","William Spataro"],"reviews":0,"subject":"math.NA","summary":"We propose a semigroup-aware structured kernel neural operator for source-driven evolution PDEs. Unlike generic time-conditioned kernel models, the proposed formulation represents temporal dependence through a latent exponential propagator \\(e^{tA_θ}\\), where \\(A_θ\\) is a learned latent generator. This provides a structured transition from a non-ODE kernel representation to an ODE-inspired kernel operator whose latent temporal evolution satisfies the composition principle of evolution semigroups. For time-dependent forcing terms, a direct kernel-based evaluation requires repeated quadrature…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-29","reactions":{}}