Title :
A new adaptive scheme for the backpropagation algorithm
Author_Institution :
Dipt. de Autom. y Computacion, Univ. Publica de Navarra, Spain
Abstract :
An adaptive algorithm to calculate the learning rate of the backpropagation algorithm is proposed. It is based on the estimation of the maximum value of that parameter that guarantees stability in each training cycle. This scheme is computationally efficient and it is shown to be faster and to avoid local minima easier than other previously proposed adaptive algorithms
Keywords :
backpropagation; neural nets; parameter estimation; adaptive algorithm; adaptive scheme; learning rate; training cycle; Adaptive algorithm; Backpropagation algorithms; Computer networks; Error correction; Graphics; Heuristic algorithms; Minimization methods; Neural networks; Robustness; Stability;
Conference_Titel :
Neural Networks, 1999. IJCNN '99. International Joint Conference on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-5529-6
DOI :
10.1109/IJCNN.1999.832643
