摘要
指出了传统的ART-2神经网络对渐变过程不敏感的局限性,提出了一种新的改进算法。并对ART-2网络进行建模,通过与其它建模方法的对比,详尽讨论了ART-2的建模方法及特点。最后通过应用改进算法解决了原先模型中的“模式漂移”现象,使模型性能得到了明显的改善。
The paper indicates the limitation of the insensitivity for gradual change process of ART-2Neural Network,and brings forward a new refinement algorithm.The ART-2Network is rebuilt to model actual systems.The paper dis-cusses the modeling method and characteristic of the ART-2network by comparing with other modeling methods.Final-ly,the problem of'pattern drifting'is successfully resolved by applying the refinement algorithm to the ART-2.The per-formance of the new model is obviously improved.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第14期25-27,42,共4页
Computer Engineering and Applications
基金
中国科学院优秀青年学者奖基金资助
关键词
ART-2神经网络
建模
模式漂移
串并联模型
pattern recognition,neural network,system modeling,ART-2,pattern drifting,series and parallel model