Incorporating Prototypes into Horizontal Collaborative Fuzzy C-Means
Collaborative fuzzy clustering concerns a process of revealing a structure of a given dataset, which is common or similar to a number of outer datasets, and the collaborative information is transmitted by some media (say partition matrix or prototypes) for the sake of the safety and personal privacy of data. In Horizontal Collaborative Fuzzy C-Means (HC-FCM), the dataset and the outer datasets are all about a same group of patterns but given in different feature spaces, and the tool for transmitting collaborative information is partition matrix. Thus, we can call it partition matrix-based HC-FCM. In this study, we introduce the prototypes into HC-FCM resulting in the partition matrix and prototype-based HC-FCM where the collaborative information is transmitted by both partition matrix and prototypes. Meanwhile we give the prototype-based HC-FCM algorithm where the collaborative information is transmitted by prototypes. We not only discuss the new versions of HC-FCM but also do the comparisons among them. The difference among these algorithms (FCM, partition matrix-based HC-FCM, prototype-based HC-FCM, partition matrix and prototype-based HC-FCM) were visualized in the experiments. Good performances of the new algorithm are revealed there.
FCM Horizontal collaborative clustering prototype-based HC-FCM
Shengli Yu Fusheng Yu
School of Mathematical Sciences, Beijing Normal University, Laboratory of Mathematics and Complex Systems, Ministry of Education, Beijing 100875, The Peoples Republic of China
国际会议
The 22nd China Control and Decision Conference(2010年中国控制与决策会议)
徐州
英文
3612-3616
2010-05-26（万方平台首次上网日期，不代表论文的发表时间）