基于规则产生准则与修剪策略的D-FNN算法研究  

The Rule Criteria and Pruning Strategy Based on D-FNN Algorithm Research

作　　者：左军[1] 周灵[1] 李晓东[1] 

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

出　　处：《中山大学学报（自然科学版）》2015年第5期43-48,共6页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：广东省自然科学基金资助项目(S2011020002719)

摘　　要：提出了一种D-FNN结构及其学习算法,该D-FNN的结构基于径向基神经网络。模糊规则的产生由输出误差或可容纳边界的有效半径决定。修剪技术的应用,使得网络结构能够保持紧凑,学习速度快,确保系统的泛化能力。对所提算法作了详细探讨,并与相关算法作比较,从而发现了D-FNN的独特思想。编写了D-FNN的仿真程序,对具体案例进行了仿真。结果表明,D-FNN具有紧凑的结构和优秀的性能。A new structure for D-FNN and its learning algorithm are put forward. The structure of this DFNN is based on RBF neural network. In the new algorithm and structure,generation of fuzzy rule is determined by the output error and the effective radius of the accommodate boundary. At the same time,the application of pruning technology makes a simple network structure,fast learning speed and generalization ability for system. The new algorithm is discussed in detail and compared with correlated algorithms. By these technology methods,the unique advantage of D-FNN is found. At last,simulation program for DFNN are wrote and the concrete cases are run in the program. Simulation results show that the new DFNN has a compact structure and excellent performance.

关 键 词：动态模糊神经网络 径向基函数 模糊规则 修剪策略 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]