Pith. sign in

REVIEW 3 cited by

Asymptotic-Preserving Neural Networks for Multiscale Kinetic Equations

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2306.15381 v4 pith:6DTYUTAV submitted 2023-06-27 math.NA cs.NA

Asymptotic-Preserving Neural Networks for Multiscale Kinetic Equations

classification math.NA cs.NA
keywords apnnequationneuralkineticmultiscaleasymptotic-preservingboundaryconditions
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

In this paper, we present two novel Asymptotic-Preserving Neural Networks (APNNs) for tackling multiscale time-dependent kinetic problems, encompassing the linear transport equation and Bhatnagar-Gross-Krook (BGK) equation with diffusive scaling. Our primary objective is to devise efficient and accurate APNN approaches for resolving multiscale kinetic equations. We have established a neural network based on even-odd decomposition and concluded that enforcing the initial condition for the linear transport equation with inflow boundary conditions is crucial. This APNN method based on even-odd parity relaxes the stringent conservation prerequisites while concurrently introducing an auxiliary deep neural network. Additionally, we have incorporated the conservation laws of mass, momentum, and energy for the Boltzmann-BGK equation into the APNN framework by enforcing exact boundary conditions. This is our second contribution. The most notable finding of this study is that approximating the zeroth, first and second moments of the particle density distribution is simpler than the distribution itself. Furthermore, a compelling phenomenon in the training process is that the convergence of density is swifter than that of momentum and energy. Finally, we investigate several benchmark problems to demonstrate the efficacy of our proposed APNN methods.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Tail observability and fourth-order closure recovery in physics-informed neural networks for Bhatnagar-Gross-Krook normal shocks

    physics.flu-dyn 2026-05 unverdicted novelty 5.0

    PINN study of BGK shocks identifies anisotropic tail-weighted observability failure in fourth-order closure R_xx^cl and shows a shock-local correction reduces its relative error to 0.112 using DVM validation.

  2. A Simple but Efficient Transformer-Based Physics-Informed Neural Network for Incompressible Navier--Stokes Equations

    physics.flu-dyn 2026-01 unverdicted novelty 4.0

    PhysicsFormer applies a lightweight Transformer PINN with pseudo-sequential representations to convection, Burgers, lid-driven cavity, and inverse Navier-Stokes problems, reporting near-zero error in parameter identif...

  3. Bridging Scales: Asymptotic Analysis and AI-Assisted Formalization

    math.NA 2026-06 unverdicted novelty 2.0

    An expository review of common structures in asymptotic analysis across scales and their potential as inputs for AI-assisted formalization tools.