IJPAM: Volume 82, No. 4 (2013)

THE NEWTON NEURAL NET:
A NEW APPROXIMATING NETWORK

A.M. Ugena
Department of Applied Mathematic
ETSI Telecomunication
Universidad Politécnica de Madrid
Avenida Complutense, 30 ``Ciudad Universitaria"
28040, Madrid, SPAIN


Abstract. In this paper we present the Newton neural network. We state several theorems that reveal this trained net as the best approximator to any curve or surface with the least number of training patterns. The low number of training samples is an advantage in this type of neural networks, as it affects substantially lowering the computational cost and training time.

Received: November 10, 2012

AMS Subject Classification: 68T05

Key Words and Phrases: backpropagation, artificial neural networks, adaptive learning, functional-link neural networks

Download paper from here.



DOI: 10.12732/ijpam.v82i4.13 How to cite this paper?
Source:
International Journal of Pure and Applied Mathematics
ISSN printed version: 1311-8080
ISSN on-line version: 1314-3395
Year: 2013
Volume: 82
Issue: 4