一种新型ARTⅡ无监督分类算法

摘要

ART是一种典型的、无监督的、能够对复杂输入模式实现自稳定和自组织识别的神经网络。该文针对标准ARTⅡ算法的预处理信号畸变问题，提出了新的非线性变换函数和竞争学习算法，该新型ARTⅡ算法的输入域由原来的非负实数域扩大到整个实数域，且分类性能良好，以多种分类问题对该算法的性能进行验证，结果表明该算法性能优良，能自适应地识别未知故障模式，分类准确。%The ART is a representative,non-supervising neural network which can recognize complicated inputting patterns self-organically.This paper presents a new non-linear transfer function and competitive learning algorithm aim to the pretreatment signal distortion problem of the standard ART Ⅱ. The inputting field of the new ART Ⅱ algorithm can be expanded to real number field from nonnegative number field and its clustering performance is very well. The new algorithm is examed by several clustering problems,the result shows that the new algorithm can recognize unknown patterns adaptively and precisely.