An Integral Upper Bound for Neural Network Approximation

Autoři: RNDr. Věra Kůrková, DSc.
Rok vydání: 2009
Typ publikace: Článek v odborném periodiku
Periodikum: Neural Computation
Projekty: Centrum aplikované kybernetiky
Bibliografická citace: KAINEN, P.C., KŮRKOVÁ, Věra. An Integral Upper Bound for Neural Network Approximation. Neural Computation. 2009, 21(10), 2970-2989. ISSN 0899-7667. E-ISSN 1530-888X
Dostupné z: doi:10.1162/neco.2009.04-08-745
DOI: 10.1162/neco.2009.04-08-745
ID: 328415

Abstrakt

For functions with suitable integral representations in the form of networks with infinitely many hidden units, upper bounds are derived on the speed of decrease of approximation error as the number of network units increases. These bounds are obtained using the framework of Bochner.