Linearly Constrained Neural Networks
Machine Learning
2021-04-29 v4 Machine Learning
Computational Physics
Abstract
We present a novel approach to modelling and learning vector fields from physical systems using neural networks that explicitly satisfy known linear operator constraints. To achieve this, the target function is modelled as a linear transformation of an underlying potential field, which is in turn modelled by a neural network. This transformation is chosen such that any prediction of the target function is guaranteed to satisfy the constraints. The approach is demonstrated on both simulated and real data examples.
Cite
@article{arxiv.2002.01600,
title = {Linearly Constrained Neural Networks},
author = {Johannes Hendriks and Carl Jidling and Adrian Wills and Thomas Schön},
journal= {arXiv preprint arXiv:2002.01600},
year = {2021}
}