We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Functional Analysis

Title: Approximation of Smoothness Classes by Deep Rectifier Networks

Abstract: We consider approximation rates of sparsely connected deep rectified linear unit (ReLU) and rectified power unit (RePU) neural networks for functions in Besov spaces $B^\alpha_{q}(L^p)$ in arbitrary dimension $d$, on general domains. We show that \alert{deep rectifier} networks with a fixed activation function attain optimal or near to optimal approximation rates for functions in the Besov space $B^\alpha_{\tau}(L^\tau)$ on the critical embedding line $1/\tau=\alpha/d+1/p$ for \emph{arbitrary} smoothness order $\alpha>0$. Using interpolation theory, this implies that the entire range of smoothness classes at or above the critical line is (near to) optimally approximated by deep ReLU/RePU networks.
Comments: To appear in SIAM Journal on Numerical Analysis
Subjects: Functional Analysis (math.FA); Machine Learning (cs.LG); Numerical Analysis (math.NA)
MSC classes: 41A65, 41A15 (primary), 68T05, 65D99 (secondary)
Journal reference: SIAM Journal on Numerical Analysis, Vol. 59, Iss. 6, 2021
DOI: 10.1137/20M1360657
Cite as: arXiv:2007.15645 [math.FA]
  (or arXiv:2007.15645v2 [math.FA] for this version)

Submission history

From: Anthony Nouy [view email]
[v1] Thu, 30 Jul 2020 17:58:05 GMT (41kb,D)
[v2] Fri, 29 Oct 2021 14:55:34 GMT (568kb,D)

Link back to: arXiv, form interface, contact.