A Fast Heuristic Global Learning Algorithm for Multilayer Neural Networks

作者: Siu-yeung Cho , Tommy W.S. Chow

DOI: 10.1023/A:1018685627113

关键词:

摘要: This paper presents a novel Heuristic Global Learning (HER-GBL) algorithm for multilayer neural networks. The is based upon the least squares method to maintain fast convergence speed, and penalized optimization solve problem of local minima. penalty term, defined as Gaussian-type function weight, provide an uphill force escape from As result, training performance dramatically improved. proposed HER-GBL yields excellent results in terms avoidance minima quality solution.

参考文章(11)
J.R. McDonnell, D. Waagen, Determining neural network connectivity using evolutionary programming asilomar conference on signals, systems and computers. pp. 786- 790 ,(1992) , 10.1109/ACSSC.1992.269165
Yi Shang, B.W. Wah, Global optimization for neural network training computational science and engineering. ,vol. 29, pp. 45- 54 ,(1996) , 10.1109/2.485892
M. Gori, A. Tesi, On the problem of local minima in backpropagation IEEE Transactions on Pattern Analysis and Machine Intelligence. ,vol. 14, pp. 76- 86 ,(1992) , 10.1109/34.107014
Scott E. Fahlman, Christian Lebiere, The Cascade-Correlation Learning Architecture neural information processing systems. ,vol. 2, pp. 524- 532 ,(1989)
S. S. Rao, R. C. Desai, Optimization Theory and Applications IEEE Transactions on Systems, Man, and Cybernetics. ,vol. 10, pp. 280- 280 ,(1980) , 10.1109/TSMC.1980.4308490
R. Battiti, G. Tecchiolli, Training neural nets with the reactive tabu search IEEE Transactions on Neural Networks. ,vol. 6, pp. 1185- 1200 ,(1995) , 10.1109/72.410361
C.R. Jones, C.P. Tsang, On the convergence of feedforward neural networks incoporating terminal attractors IEEE International Conference on Neural Networks. pp. 929- 935 ,(1993) , 10.1109/ICNN.1993.298682
Aimo Torn, Antanas Zilinskas, Global optimization ,(1989)