Genetic Search for Optimal Representations in Neural Networks

An approach to learning in feed-forward neural networks is put forward that combines gradual synaptic modification at the output layer with genetic adaptation in the lower layer(s). In this "GA-delta" technique, the alleles are threshold units (a set of weights and a threshold); a chromosome is a collection of such units, and hence defines a mapping from the input layer to a hidden layer. Genetic operators are defined on these chromosome-mapping to facilitate search for a mapping that renders the task solvable by a single layer of weights. The performance of GA-delta is presented on several tasks, and the effects of the various operators is studied