{"id":"arxiv/2611.07364","url":"https://arxiv.org/abs/2611.07364","tags":["math.NA","cs.LG"],"score":0,"title":"Asymptotic-Preserving Tensor Neural Networks for 3D High-Field Kinetic Equations with Applications to Uncertainty Quantification","domain":"math.NA","source":"arxiv","authors":["Shi Jin","Tian-ai Zhang"],"reviews":0,"subject":"math.NA","summary":"Kinetic equations in the high-field regime describe particle transport dominated by electric-field acceleration and collisional relaxation. These equations are widely used in simulations of plasmas and semiconductor devices. Typical models include the Vlasov--Poisson--Fokker--Planck (VPFP) and Boltzmann--Poisson systems. Their high dimensionality and multiscale nature pose two of the major challenges to accurate and efficient numerical solution. To address these challenges, we propose an asymptotic-preserving tensor neural network (APTNN) method. Specifically, to reduce high-dimensional…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-05","reactions":{}}