基于新型联想记忆神经网络的非线性系统辨识  

Research of Identification of Non-linear System Based on a New-type Associative Memory Neural Network

作　　者：吴学礼[1] 孟凡华[2] 王永骥[1] 李平[2] 贾辉然[2] 孟华[2] 

机构地区：[1]华中科技大学控制科学与工程系,湖北武汉430074 [2]河北科技大学电气信息学院,河北石家庄050054

出　　处：《地理与地理信息科学》2004年第4期110-112,共3页Geography and Geo-Information Science

基　　金：河北省科技攻关计划项目(01213407D)

摘　　要：Hopfield网络模型具有联想存储器功能,但对系统辨识不适用。具有动态记忆功能的Elman神经网络的泛化能力比较低。该文提出了一种新型联想记忆神经网络结构和学习算法,通过引入联想记忆衰减因子,提高了对非线性系统的辨识能力。通过与Elman动态神经网络辨识方法的仿真比较,说明联想记忆神经网络辨识方法具有很好的动态辨识能力和泛化能力。The Hopfield network model has function of memory, but it cannot be used in system identification. The neural network of Elman has dynamic memory ability, but its extension is low. This paper introduces a new kind of associative memory neural network and its algorithm. By introducing the factor of associative memory decay, the ability of identification of the non-linear system can be raised. In simulation and comparison with the neural network of Elman, the associative memory neural network has good ability of identification and extension.

关 键 词：联想记忆 神经网络 非线性系统辨识 泛化能力 生物神经系统 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]