摘要
本文提出了一种基于自组织特征映射的前债式人工神经网络模型,介绍了其结构和算法.该模型基于自组织特征映射机理,用统计方法获得输入信号对不同模式类别的隶属程度,并由此进行模式分类判决计算.该神经网络模型还导出了“模式地形图”的概念,可以实现数据聚类分析的可视化.经计算机模拟验证,上述算法和概念是有效的.
This paper presents a novel feedforward artificial neural network (ANN)model based on Kohonen self-organizing feature mapping (SOM), whose structure , learning rules and adaptive algorithms are provided in detail here .The concept of membership function is combined with SOM, so that this ANN model can obtain the pobability distribution of the input information by statistical method, then make pattern classification decision. Besides, this paper presents a novel concept of 'Pattern Map', which is able to facilitate visual inspection of the data .The results of the computer simulations show that the new concepts and methods mentioned above are effective.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1998年第7期165-168,共4页
Acta Electronica Sinica
关键词
人工神经网络
模型
自组织特征映射
聚类分析
Artificial neulal network model, Sesf-organizing fearure mapping, Cluster analysis , Visualization