A PSO Algorithm with the Improved Diversity for Feedforward Neural Networks
Tong-Yue Gu, Shiguang Ju, Fei Han · 2009
In this paper, an improved particle swarm optimization (PSO) with the improved diversity is proposed to train feedforward neural networks (FNN). In this algorithm, first, the PSO algorithm is used to train the FNN. Second, when the particle swarm is trapped into local minima or loses its diversity, each particle in the swarm and its best position (Pb) are interrupted by a random function in order to improve the diversity of population, and the best position (Pg) of all particles remains unchanged. Third, the PSO algorithm with the new population is used to search the better global optimum. The proposed algorithm improves the diversity of the swarm and has good convergence performance. Finally, the experimental results are given to verify the efficiency and effectiveness of our proposed learning algorithm.