English

Neural networks: deep, shallow, or in between?

Machine Learning 2023-10-12 v1 Machine Learning Numerical Analysis Numerical Analysis

Abstract

We give estimates from below for the error of approximation of a compact subset from a Banach space by the outputs of feed-forward neural networks with width W, depth l and Lipschitz activation functions. We show that, modulo logarithmic factors, rates better that entropy numbers' rates are possibly attainable only for neural networks for which the depth l goes to infinity, and that there is no gain if we fix the depth and let the width W go to infinity.

Keywords

Cite

@article{arxiv.2310.07190,
  title  = {Neural networks: deep, shallow, or in between?},
  author = {Guergana Petrova and Przemyslaw Wojtaszczyk},
  journal= {arXiv preprint arXiv:2310.07190},
  year   = {2023}
}
R2 v1 2026-06-28T12:46:54.323Z