摘 要:
作为一种模拟嗅觉神经网络的仿生模型,KⅢ模型具有许多不同于传统人工神经网络的特点.KⅢ模型的结构更接近真实的生物神经网络结构,并具有良好的模式识别能力.分析了该模型进行模式识别时的动力学特性.研究结果表明,对应不同的刺激持续时间,KⅠ模型呈现出不同的输出响应;将时变信号作为刺激直接输入时,KⅢ模型也可以形成特定的吸引子组区别不同的模式,这也为KⅢ模型直接识别时变信号提供了依据.[著者文摘]
文章出处:
《湖南大学学报:自然科学版》-2008年35卷3期 -76-79页
栏目信息:
分 类 号:
文献标识码:
A
文章编号:
1000-2472(2008)03-0076-04
Research on the Dynamic Characteristics of KⅢ Model
ZHNNG jin , LI Si-duo, WANG Ru-long, LI Guang (1. Software School, Hunan Univ, Changsha, Hunan 410082, China; 2. National Laboratory of Industrial Control Technology, Zhejiang Univ, Hangzhou, Zhejiang 310027, China)
Abstract:
This paper analyzed the dynamic characteristics of K Ⅲ model when it was used to recognize different patterns. Research results have shown that K Ⅰ model has different responses to different durative time of stimulus. When the time-varying signals were used as the stimulus and input to KⅢ model directly, KⅢ model could also form the unique attractors group to distinguish different patterns, which provides the basis for K Ⅲ model to recognize time-varying signals directly.[著者文摘]
Key words:
artificial neural networks; olfactory neural system; dynamics; phase diagrams
基金资助:
国家973计划资助项目(2004CB720302);国家创新研究群体科学基金资助项目(60421002);湖南省教育厅科学研究资助项目(07D046)

学术















cqvip.com