Applications of GA-based optimization of neural network connection topology

Abstract

A genetic algorithm (GA) is used to explore the connection space of an artificial neural network (ANN) with the objective of finding a sparsely connected network that yields the same accuracy as a fully connected network. Such sparsity is desired as it improves the generalization capabilities of the mapping. The ANN with the GA-chosen set of connections is then trained using a supervised mode of learning known as backpropagation error. Using this technique, three different applications are analyzed.