An improved training algorithm for feedforward neural network learning based on terminal attractors

Search Swinburne Research Bank

Home

Please use this identifier to cite or link to this item: http://hdl.handle.net/1959.3/92864

Title: An improved training algorithm for feedforward neural network learning based on terminal attractors
Author(s): Yu, Xinghuo; Wang, Bin; Batbayar, Batsukh; Wang, Liuping; Man, Zhihong
Abstract: In this paper, an improved training algorithm based on the terminal attractor concept for feedforward neural network learning is proposed. A condition to avoid the singularity problem is proposed. The effectiveness of the proposed algorithm is evaluated by various simulation results for a function approximation problem and a stock market index prediction problem. It is shown that the terminal attractor based training algorithm performs consistently in comparison with other existing training algorithms.
Publication type: Journal article
Research centre: Swinburne University of Technology. Faculty of Engineering and Industrial Sciences
Source: Journal of Global Optimization, Vol. 51, no. 2 (Oct 2011), pp. 271-284
Publication year: 2011
FOR Code(s): 0102 Applied Mathematics; 0103 Numerical and Computational Mathematics
Keyword(s): Algorithms; Back-propagation; Feedforward neural networks; Learning; Optimisation; Terminal attractors; Training algorithms
Publisher: Springer
ISSN: 0925-5001
Publisher URL: http://dx.doi.org/10.1007/s10898-010-9597-6
Peer reviewed