基于模拟双退火的模糊CMAC神经网络  

Fuzzy CMAC neural Network Based on Simulated double Annealing

作　　者：胡劲松[1] 李朝[1] 

机构地区：[1]华南理工大学计算机学院,广州510006

出　　处：《微计算机信息》2009年第24期139-140,148,共3页Control & Automation

摘　　要：本文首先给出了一种模拟双退火算法,增加第二退火温度,随着第二温度的降低,逐渐减少状态向量中可变元素的个数,直至只允许一个元素变化,由此可快速精确的逼近最优解的,其次利用该算法对模糊CMAC(小脑模型关节控制器)神经网络的权值进行优化,并构成了一种新型模糊CMAC神经网络。在模型未知的情况下,利用该神经网络对一个非线形时变模型进行控制仿真,效果很好。Firstly, the paper proposes a simulated double annealing algorithm. When the second annealing temperature that is introduced by us decreases, the number of state variables decreases until just one, so that the method can achieve optimal result rapidly and exactly. Secondly, a novel fuzzy CMAC neural network is constructed based on the simulated double annealing algorithm. The simulated result of a non-linear plant controlled by the neural network is good, although the model of the non-linear plant isn＇t known.

关 键 词：模糊神经网络 模拟双退火 非线性控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]