标题：Clustering-based distributed Support Vector Machine in Wireless Sensor Networks
作者：Li, Ye ;Wang, Yongli ;He, Guoping
作者机构：[Li, Ye ;Wang, Yongli ;He, Guoping ] College of Information Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, Ch 更多
来源：Journal of Information and Computational Science
摘要：This paper develops a clustering-based algorithm to train a Support Vector Machine (SVM) in a distributed fashion. As a new classification tool, SVM has been widely used in the Wireless Sensor Networks (WSN). In this paper, we first propose a weight-based clustering algorithm, which takes into consideration the ideal node-degree, transmission power and energy consumption of nodes. Compared with the planar structure, the clustering structure has advantages in lower energy conservation and better scalability. Each clusterhead that is selected from the proposed clustering algorithm, collects hull vectors from its member nodes and exchanges information among its neighbor clusterheads to train the linear classifier. Finally, all the clusterheads can get the same optimal linear classifier, which is equal to the global optimal linear classifier trained by all the training data in the WSN. In the proposed algorithm, the clusterheads need not transfer data to the base station and communicate their training data with neighbor clusterheads, so energy conservation and privacy protection are achieved. Simulation experiments are conducted to evaluate the performance of the new algorithm. The results show that the new algorithm is efficient for large-scale WSN. Copyright © 2012 Binary Information Press.