Comparison of Approximation Capabilities of Neural Networks and Linear Models

Authors: Věra Kůrková
Publication year: 2010
Publication type: Konferenční příspěvek (mezinárodní konf.)
Periodical: Informačné Technológie - Aplikácie a Teória
Projects: Centre for Applied Cybernetics
Bibliographic citation: KŮRKOVÁ, Věra. Comparison of Approximation Capabilities of Neural Networks and Linear Models. In: PARDUBSKÁ, D., ed. Informačné Technológie - Aplikácie a Teória. Seňa: Pont, 2010, S. 31-36. ISBN 978-80-970179-3-4.
ID: 348390

Abstract

Using method from nonlinear approximation and integral representations tailored to computational units, we describe some cases when neural networks outperform any linear approximator.