A Convex Hull Based SVM Active Learning Algorithm

摘要：

Active learning and its preferable performance in the training process of support vector machines (SVM) have attracted the attentions of researchers in the area of machine learning. How to reduce the annotation cost of unlabelled samples in the process of SVM active learning is studied in this paper, the convex hull vectors of training samples are used in the SVM active learning, and based on the uncertainty based sampling and the integrated stopping strategy, an algorithm of SVM active learning by using convex hull is given, experimental results indicate the effectiveness of the algorithm.

关键词： Active Learning Convex Hull Uncertainly Based Sampling

作者: Xiaodan Wang Hailong Xu Dongying Bai Hongda Zhang

作者单位: Department of Computer Engineering, Air Force Engineering University

会议类型: 国际会议

会议名称: 2010 International Conference on Information Security and Artificial Intelligence(2010年信息安全与人工智能国际会议 ISAI 2010)

会议地点: 成都

会议语种:英文

页码: 297-300

在线出版日期: 2010-12-17（万方平台首次上网日期，不代表论文的发表时间）