基于相近关系的粗糙因子神经网络的模式识别方法

针对经典粗糙集理论中的不可分辨关系对连续属性值中噪声数据缺乏容错性的情况,提出一种基于个体属性值距离的相近关系,定义了相近关系下的粗糙集理论的基本概念.在相近关系的基础上,提出了衡量粗糙隶属度的方法,研究了该隶属度函数的性质,利用该函数作为粗糙因子设计了粗糙因子神经网络,可减小噪声污染的影响,并使网络的收敛速度得到提高.最后,通过对某型歼击机操纵面故障的模式识别的仿真研究验证了文中方法的正确性和有效性.

Model Identification in Rough Factor Neural Network Based on Nearness Relationship

In this paper, relation of nearness is proposed to replace that of indiscernibility for increased robustness of a decision system against noise. Preliminaries of nearness relationship rough sets are defined and the properties of nearness rough membership function are studied. An architecture of neural network based on nearness relationship rough sets and nearness rough membership functions is described. Neurons in such a network consist of conventional neurons and rough neurons, each neuron having a rough factor. Influence of noise on the network is reduced, and convergence is speeded. An example based on fault identification of an aircraft actuator is presented. Simulation results indicate effectiveness of the proposed method.