Title :
Delta training strategies
Author :
Cloete, Ian ; Ludik, Jacques
Author_Institution :
Dept. of Comput. Sci., Stellenbosch Univ., South Africa
fDate :
27 Jun-2 Jul 1994
Abstract :
The training strategy used in connectionist learning has been overlooked as a method to improve learning time. We suggest several new strategies for backpropagation learning for both feedforward and recurrent architectures and show experimentally that they produce much faster convergence compared to conventional training using a fixed set. In particular, the Delta training strategy produced the best improvement of 76%
Keywords :
backpropagation; feedforward neural nets; recurrent neural nets; Delta training strategies; backpropagation learning; connectionist learning; feedforward architectures; fixed set; learning time; recurrent architectures; Africa; Artificial neural networks; Computer architecture; Computer science; Convergence; Euclidean distance; Merging; Neural networks; Sun;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374178
