摘要
针对 RBF网络的特点 ,提出一种递阶遗传算法 ,不仅可以同时确定网络参数 (连接权、隐节点中心和宽度 ) ,而且解决了网络拓扑结构的优化训练问题。算例仿真表明所提出的算法是很有效的。
A hierarchical genetic algorithm for RBF neural networks is proposed to train network parameters such as centers, widths and connection weights. In addition, the configuration of an RBF network is also determined at the same time during training. Training and test based on practical data sets are carried out respectively, and a good performance of the new algorithm is demonstrated.
出处
《控制与决策》
EI
CSCD
北大核心
2000年第2期165-168,共4页
Control and Decision
基金
国家自然科学基金项目!(796 70 0 6 4)
关键词
RBF网络
递阶遗传算法
参数训练
神经网络
RBF neural networks, hierarchical genetic algorithms, parameter training, structural optimization
