Particle Swarm Optimisation in Feedforward Neural Network

Abstract

The paper describes a new evolutionary system for evolving artificial neural networks (ANN's) called PSONN, which is based on the particle swarm optimisation (PSO) algorithm. The PSO algorithm is used to evolve both the architecture and weights of ANN's, this means that an ANN's architecture is adaptively adjusted by PSO algorithm, then the nodes of this ANN's are also evolved by PSO algorithm to evaluate the quality of this network architecture. This process is repeated until the best ANN's is accepted or the maximum number of generations has been reached. In PSONN, a strategy of evolving added nodes and a partial training algorithm are used to maintain a close behavioural link between the parents and their offspring, which improves the efficiency of evolving ANN's. PSONN has been tested on two real problems in the medical domain. The results show that ANN's evolved by PSONN have good accuracy and generalisation ability.