Title :
New learning factor and testing methods for conjugate gradient training algorithm
Author :
Kim, Tae ; Manry, Michael T. ; Maldonado, Javier
Author_Institution :
Dept. of Electr. Eng., Texas Univ., Arlington, TX, USA
Abstract :
The conjugate gradient method has advantages over backpropagation in the training of artificial neural networks. Unlike previous investigators who have obtained learning factors using computationally expensive iterative line searches, we obtain the optimal learning factor in one step. We validate the learning factor with several tests, and analyze the input bias problem. Examples confirm the usefulness of improved conjugate gradient.
Keywords :
backpropagation; conjugate gradient methods; learning (artificial intelligence); multilayer perceptrons; artificial neural networks training; backpropagation; conjugate gradient method; conjugate gradient training algorithm; input bias problem; iterative line searches; learning factors; multilayer perceptron; optimal learning factor; testing methods; Artificial neural networks; Character generation; Gradient methods; Image processing; Joining processes; Multilayer perceptrons; Optimization methods; Power system modeling; Predictive models; Testing;
Conference_Titel :
Neural Networks, 2003. Proceedings of the International Joint Conference on
Print_ISBN :
0-7803-7898-9
DOI :
10.1109/IJCNN.2003.1223716