Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
Queen's University Belfast
A new Jacobian matrix for optimal learning of single-layer neural networks
This EPSRC (GR/S85191/01) funded research introduces for the first time a new Jacobian matrix which yields a more accurate approximation of the true cost function for the training of single-layer forward neural networks (SLFNs) using second-order learning algorithms. This enables the development of a novel analytic framework which helps to speed up the convergence of network learning and to improve the network generalization performance, confirmed by its application to benchmark problems. This research resulted in a number of invited research seminars given in leading Chinese universities through a RCUK funded UK-China Science Bridge project (EP/G042594/1).