An evolutionary algorithm for feed-forward neural networks optimization

We propose an evolutionary algorithm for optimizing both the topology and the synaptic weights of single hidden-layer feed-forward neural networks (SLFNs). We introduce new evolutionary operators of recombination and mutation we designed for evolving a population of SLFNs candidate solutions to a specific problem. The performance of the proposed algorithm in solving classification and prediction problems is experimentally tested using five real-world benchmark datasets. The experimental results are analyzed and compared to those produced by two other methods using two measures of performance.