摘要
ART2神经网络是按照自适应谐振理论建立的一种自组织、无监督的人工神经网络,可应用于连续的动态数据分类.但在实际应用中发现其理论存在同相位不可分"的缺点.针对此问题提出了一种改进的ART2神经网络模型结构及其算法,进行了仿真,给出了与常用ART2网络所做仿真的结果比较.
ART2 is a self-organized and unsupervised artificial neural network constructed from adaptive resonance theory which can be used to classify continuous active data. We have found that the theory is limited of the same phase data with different amplitudes and insensitivity to gradual change data during the simulation of data classified with ART2 neural network. Therefore,we propose a new neural network model based on adaptive resonance theory. We provide the model construction and relevant algorithm as well as the comparison with ART2.
出处
《机械与电子》
2005年第11期62-64,共3页
Machinery & Electronics
