An Improved ARPSO for Feedforward Neural Networks
Although particle swam optimization (PSO) algorithm is a good optimization tool for feedforward neural network s(FNN), it is easy to lose the diversity of the swarm and suffer from premature convergence. An improved PSO algorithm based on the attractive and repulsive PSO (ARPSO) is proposed to train FNN in this paper. In addition to the phases of repulsion and attraction, the third phase named as mixed phase is introduced in the improved PSO, in which the particles are attracting and compelling simultaneously to prevent premature convergence. Moreover, an improved mutation operation is taken to help particles jump out of local minima when the current global best position has not been changed for some predetermined iterations in the improved PSO. Since the improved PSO could improve the diversity of the swarm to avoid premature convergence, it has better convergence performance than traditional PSOs. Finally, the experimental results are given to show the effectiveness of the proposed algorithm on function approximation and iris classification problems.
Particle swarm optimization diversity premature convergence feedforward neural newtorks
Fei Han Jiansheng Zhu
School of Computer Science and Telecommunication Engineering Jiangsu University Zhenjiang, China
国际会议
2011 Seventh International Conference on Natural Computation(第七届自然计算国际会议 ICNC 2011)
上海
英文
1172-1176
2011-07-26（万方平台首次上网日期，不代表论文的发表时间）