English

A multivariate Riesz basis of ReLU neural networks

Information Theory 2023-03-02 v1 Functional Analysis math.IT

Abstract

We consider the trigonometric-like system of piecewise linear functions introduced recently by Daubechies, DeVore, Foucart, Hanin, and Petrova. We provide an alternative proof that this system forms a Riesz basis of L2([0,1])L_2([0,1]) based on the Gershgorin theorem. We also generalize this system to higher dimensions d>1d>1 by a construction, which avoids using (tensor) products. As a consequence, the functions from the new Riesz basis of L2([0,1]d)L_2([0,1]^d) can be easily represented by neural networks. Moreover, the Riesz constants of this system are independent of dd, making it an attractive building block regarding future multivariate analysis of neural networks.

Keywords

Cite

@article{arxiv.2303.00076,
  title  = {A multivariate Riesz basis of ReLU neural networks},
  author = {Cornelia Schneider and Jan Vybíral},
  journal= {arXiv preprint arXiv:2303.00076},
  year   = {2023}
}