Title :
Refinements in training schemes for the Coulomb Energy network
Author :
Vassilopoulos, John F. ; Koutsougeras, Cris
Author_Institution :
Center for Bioenviron. Res., Tulane Univ., New Orleans, LA, USA
Abstract :
We discuss the interesting perspective offered by the Coulomb Energy network and we identify certain disadvantages with the existing approach to training it. We address these problems by constraining its architecture (topology) and offer a derivation of the new associated training algorithm. We study further refinements of this algorithm. Most notably, existing genetic algorithms are employed as initial search techniques and simulation results are provided.
Keywords :
feedforward neural nets; genetic algorithms; learning (artificial intelligence); multilayer perceptrons; neural net architecture; search problems; Coulomb Energy network; feedforward neural network; genetic algorithms; learning model; multilayer network; search techniques; simulation results; topology; training algorithm; training scheme refinement; Aggregates; Clustering algorithms; Curve fitting; Genetic algorithms; Intelligent networks; Network topology; Neural networks; Pattern recognition; Robustness; Space charge;
Conference_Titel :
Tools with Artificial Intelligence, 1996., Proceedings Eighth IEEE International Conference on
Print_ISBN :
0-8186-7686-7
DOI :
10.1109/TAI.1996.560451
