分级在线自组织学习的GD-FNN算法研究  

Research on Online Self-Organizational Learning GD-FNN Algorithm by Grading

作　　者：左军[1] 周灵[1] 孙亚民[2] 

机构地区：[1]佛山科学技术学院电子与信息工程学院,广东佛山528000 [2]南京理工大学计算机科学与技术学院,江苏南京210094

出　　处：《中山大学学报（自然科学版）》2015年第3期26-29,35,共5页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：广东省自然科学基金资助项目(S2011020002719)

摘　　要：提出了基于椭圆基函数(EBF)的广义动态模糊神经网络(GD-FNN)算法。算法提取的模糊规则具有很好可理解性,可以作为建模工具,也可以作为知识提取的工具。广义动态模糊神经网络由于基于模糊ε-完备性,同时提出了一种新颖的在线参数分配机制,从而缓解了初始化的随机选择,且与输入变量不同值域没有关系,因而更容易构造一个较好性能的模糊系统。开发了仿真程序,对具体案例进行仿真,取得了较为理想的结果。General dynamic fuzzy neural network （GD-FNN） algorithm is proposed based on the elliptic basis function （EBF）. Fuzzy rules generated from the algorithm are intelligibility. It can be used as a modeling tool. and a tool of knowledge extraction. Because of a novel on-line parameter allocation mecha- nism for allevialing the random selection in initialization without relation to different input variable range, the proposed GD-FNN based on fuzzy ε-completeness is more easy to construct a good fuzzy system in per- formance. The simulation program is also developed based on the GD-FNN algorithm and ideal results are achieved by simulation in specific design case.

关 键 词：广义动态模糊神经网络 动态模糊神经网络 径向基函数 椭圆基函数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]