A neural-network based nonlinear non-intrusive reduced basis method with online adaptation for parametrized partial differential equations

Jingye Li, Alex Bespalov, Jinglai Li · 2025

Read A neural-network based nonlinear non-intrusive reduced basis method with online adaptation for parametrized partial differential equations by Jingye Li, Alex Bespalov, Jinglai Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-07684.txt.

Description

Jingye Li, Alex Bespalov, and Jinglai Li cut the cost of solving parametrized partial equations. Engineers rerun such equations for many settings, and each full solve is expensive. Classic reduced basis methods combine fixed snapshots linearly, which struggles when solutions bend in nonlinear ways as settings change.

Their Nemytskii neural operator replaces linear mixing with a structured pointwise map over learned feature functions. A hypernetwork turns settings into a small rebuild network, refined online by physics-based checks. That smoothness survives the neural design. Faster surrogates with physics guardrails could speed design loops in fluids, heat, and structures.

Catalog record

CREATORJingye Li, Alex Bespalov, Jinglai Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-07684

Browse the full library · About the Tower of Babel