基于种群小生境微粒群算法的前向神经网络设计

根据自然界中鱼鸟等所具有的种群运动特征，借鉴递阶编码的思想，构造出一种种群小生境微粒群算法，具有小生境内个体微粒自由运动特征分量和小生境种群运动特征分量分层递阶进化的特征，克服了标准微粒群算法或其改进算法在多蜂函数寻优时出现的微粒“早熟”现象，应用该算法进行三层前向神经网络连接权值和网络结构联合并行自适应设计，在混沌时间序列预测中显示了良好的性能。

Evolutionary Design of Feed-forward Neural Network Based on Species Niching Particle Swarm Optimizer

A species niching particle swarm optimizer(SNPSO) is proposed according to the moving behavior of bird flock or fish school. The new algorithm has the charateristics of the particle's free moving offset and the niching population's moving offset evolved in a hierachy model. Using these characterastics, our algorithm orercomes the "prematurity" in global optimization by the normal particle swarm optimization and its modified model. When using this algorithm in the feed forward structure and weights associated training, the neural network shows a good performance in chaotic time squence forecasting.