Fuzzy Control of Back-Propagation Training
Michael Negnevitsky and Martin J. Ringrose
School of Engineering, University of Tasmania,
GPO Box 252-65 Hobart, Tasmania 7001 Australia
A fuzzy logic controller for updating training parameters in the error back-propagation algorithm is presented. The controller is based on heuristic rules for speeding up the convergence of training process, incorporating both learning rate and momentum constant changes.