摘要
针对随机模糊神经网络(SFNN)的网络结构没有明确的物理含义,仅仅是一种实现随机模糊逻辑系统的计算结构的问题,对其网络结构进行了改进,重新定义了每层的节点原型.改进后每层之间的物理含义明确且节点数目减少,从而计算量有所减少.对于SFNN的参数和结构,可以分别通过参数学习算法和结构学习算法来优化.将SFNN用于随机混沌时间序列预测,仿真结果表明:该系统由于引入了随机的概念,使网络能更有效地防止噪声的干扰,因而更适合于工程应用.
Focusing on the problem that the structure of the SFNN (stochastic fuzzy neural network) has no physical meaning and is just a computation network to realize the computation of stochastic fuzzy logic system, the structure of the SFNN is modified and the nodes in each layer of the SFNN are discussed for solving the problem. Each layer of the improved structure has exact physical meaning, and the number of the nodes and computation burden are decreased. The parameter and structure can be optimized by using learning algorithms and applied to predict the stochastic chaotic time series. The simulation results show that the SFNN is superior to the existing neural networks in modeling and controlling of stochastic system and more suitable for engineering applications.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2003年第10期991-994,共4页
Journal of Xi'an Jiaotong University
基金
国家高技术研究发展计划"八六三"重大专项资助项目 (2 0 0 3AA50 1 1 0 0 )
