摘要
自组织神经网络的主要目的是将任意维数的输入信号模式转变成为一维或二维的离散映射,并且以拓扑有序的方式自适应地实现这个过程。学习过程中,对邻域宽度函数和学习率函数参数是根据经验选择的,没有一定的规则或方法,因此,邻域保持映射的获得往往先于参数的学习过程。将线性Kalman滤波器和基于无先导变换的Kalman滤波器分别用于学习率函数和邻域宽度函数的预测,可以提高自组织神经网络的学习能力。改进后的算法可以根据输入数据自适应地调整邻域宽度函数和学习率函数。
The principal goal of the self-organizing feature map is to transform an incoming signal pattern oi arbitrary dimension into a one- or two-dimensional discrete map,and to perform this transformation adaptively in a topologically ordered fashion.The learning process is controlled by learning coefficient and the width of neighborhood function,which have to be chosen empirically because there aren't exist rules or a method for their calculation.To improve the learning ability of the self-organizing maps,a method is presented,which the learning coefficient and the width of neighborhood function is predicted by linear Kalman ritler and the Kalman filter based on the unscented transform respectively.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第31期22-24,共3页
Computer Engineering and Applications
基金
国家自然科学基金No.60671061~~