摘要
根据神经网络与模糊逻辑系统的各自特性,在分析了扩展RBF网络与T-S型模糊逻辑系统功能等价性的基础之上,提出了扩展RBF网络与T-S型模糊逻辑系统相互融合的一种新的自适应模糊逻辑系统。考察了系统现有的结构和参数优化方法,提出采用基于山峰函数的减法聚类算法与由梯度下降算法和最小二乘算法所组成的混合学习算法来对模糊推理系统进行训练,从而优化自适应模糊推理系统的结构和参数。最后,针对一非线性函数逼近问题进行了验证,仿真取得了很好的结果,系统的逼近精度和收敛速度获得了明显的提高,从而表明本文提出的算法是有效性和可行性。
According to the characteristics of neural networks and fuzzy logic system, this article firstly analyses the equivalence between extended RBF network and T-S fuzzy inference system, and advises a new adaptive fuzzy logic system. After reviewing some existed methods that are used to optimize the fuzzy logic system, this article brings forward the subtraction clustering method based on the mountain function to identify the structure of fuzzy system. At the same time, the mixed learning algorithms made up of the grads descend algorithm and least squares algorithm are adopted to train the parameters. Lastly, the proposed method has been evaluated by a nonlinear function, system抯 precision and the rate of convergence are enhanced remarkably. Those results demonstrate the efficiency and feasibility of the optimization algorithm.
出处
《系统仿真学报》
CAS
CSCD
2002年第4期501-503,共3页
Journal of System Simulation
