The performance of multi-layer feed-forward neural networks is closely related to the success of training algorithms in finding optimal weights in the network. Although conventional algorithms such as back-propagation are popular in this regard, they suffer from drawbacks such as a tendency to get stuck in local optima. In this paper, we propose an effective hybrid algorithm, BLPSO-GBS, for neural network training based on particle swarm optimisation (PSO), biogeography-based optimisation (BBO), and a global-best strategy. BLPSO-GBS updates each particle based on neighbouring particles and a biogeography-based learning strategy is used to generate the neighbouring particles using the migration operator in BBO. Our experiments on different benchmark datasets and comparison to various algorithms clearly show the competitive performance of BLPSO-GBS.
