A Group Search Optimizer for Neural Network Training

A novel optimization algorithm: Group Search Optimizer (GSO) [1] has been successfully developed, which is inspired by animal behavioural ecology. The algorithm is based on a Producer-Scrounger model of animal behaviour, which assumes group members search either for ‘finding’ (producer) or for ‘joining’ (scrounger) opportunities. Animal scanning mechanisms (

e.g.

, vision) are incorporated to develop the algorithm. In this paper, we apply the GSO to Artificial Neural Network (ANN) training to further investigate its applicability to real-world problems. The parameters of a 3-layer feed-forward ANN, including connection weights and bias are tuned by the GSO algorithm. Two real-world classification problems have been employed as benchmark problems trained by the ANN, to assess the performance of the GSO-trained ANN (GSOANN). In comparison with other sophisticated machine learning techniques proposed for ANN training in recent years, including some ANN ensembles, GSOANN has a better convergence and generalization performances on the two benchmark problems.