A Locally Conservative Enriched Linearized Neural Network Approximation to Elliptic PDEs

License: See LICENSE file · Updated 2026-10-06
My rating

This paper develops a locally conservative enriched linearized neural network approximation for the Darcy flow model and elliptic partial differential equations. Its approximation uses a linearized shallow ReLU network with hidden-layer parameters fixed on a quasi-uniform spherical set, augmented by piecewise constant functions on an auxiliary subdivision that may contain general, possibly curved elements to ensure local mass conservation.

Tags: