Title :
The further discussions on constrained learning algorithms
Author :
Huang, De-Shuang
Author_Institution :
Inst. of Intelligent Machines, Chinese Acad. of Sci., Hefei, China
Abstract :
This paper revisits the constrained learning algorithm (CLA) proposed by Perantonis and Karras (1995), and makes further analyses and discussions on the parameters with the CLAs. Specifically, we investigate the effect of removing the constrained condition of the weight change on the CLAs. It is suggested that for those problems that do not need to do precise computation, the modified CLA is a better choice. Finally, some simulation results are presented to support our claims.
Keywords :
backpropagation; feedforward neural nets; function approximation; optimisation; Perantonis Karras algorithm; backpropagation; constrained learning algorithms; feedforward neural nets; optimisation; Acceleration; Algorithm design and analysis; Approximation algorithms; Computational modeling; Convergence; Cost function; Feedforward neural networks; Learning systems; Machine learning; Neural networks;
Conference_Titel :
Neural Networks, 2003. Proceedings of the International Joint Conference on
Print_ISBN :
0-7803-7898-9
DOI :
10.1109/IJCNN.2003.1223692
