A new hybrid evolutionary mechanism based on unsupervised learning for Connectionist Systems

Recent studies have confirmed that the modulation of synaptic efficacy affects emergent behaviour of brain cells assemblies. We report the first results of adding up the behaviour of particular brain circuits to Artificial Neural Networks. A new hybrid learning method has emerged. In order to find the best solution to a given problem, this method combines the use of Genetic Algorithms with particular changes to connection weights based on this behaviour. We show this combination in feed-forward multilayer architectures initially created to solve classification problems and we illustrate the benefits obtained with this new method.