在线序列主动学习方法  

Online Sequential Active Learning Approach

作　　者：翟俊海[1] 臧立光 张素芳[3] ZHAI Jun-hai ZANG Li-guang ZHANG Su-fang(Key Laboratory of Machine Learning and Computational Intelligence, College of Mathematics and Information Science, Hebei University,Baoding 071002,China College of Computer Science and Technology, Hebei University, Baoding 071002, China Hebei Branch of China Meteorological Administration Training Center, China Meteorological Administration, Baoding 071000, China)

机构地区：[1]河北大学数学与信息科学学院河北省机器学习与计算智能重点实验室,保定071002 [2]河北大学计算机科学与技术学院,保定071002 [3]中国气象局气象干部培训学院河北分院,保定071000

出　　处：《计算机科学》2017年第1期37-41,70,共6页Computer Science

基　　金：国家自然科学基金项目(71371063);河北省自然科学基金项目(F2013201220);河北省高等学校科学技术研究重点项目(ZD20131028);河北省高等学校科学技术研究项目(QN20131153)资助

摘　　要：现实世界中存在着大量无类标的数据,如各种医疗图像数据、网页数据等。在大数据时代,这种情况更加突出。标注这些无类标的数据需要付出巨大的代价。主动学习是解决这一问题的有效手段,也是近几年机器学习和数据挖掘领域中的一个研究热点。提出了一种基于在线序列极限学习机的主动学习算法,该算法利用在线序列极限学习机增量学习的特点,可显著提高学习系统的效率。另外,该算法用样例熵作为启发式度量无类标样例的重要性,用K-近邻分类器作为Oracle标注选出的无类标样例的类别。实验结果显示,提出的算法具有学习速度快、标注准确的特点。In the real world,there are a lot of unlabelled data, such as various medical images and web data, etc. In the era of big data, this situation is more prominent. It is expensive to label large amount of unlabelled data. Active learning is an effective method to solve this problem, and it is one of the hot research topics in the field of machine learning and data mining. Based on online sequential extreme learning machine, an active learning algorithm was proposed in this pa- per. Due to the nature of incremental learning embedded in online sequential extreme learning machine, the proposed al- gorithm can significantly improve the efficiency of learning system. Furthermore, the proposed algorithm uses instance entropy as heuristic to measure the importance of the unlabeled instances, and uses K-nearest neighbor classifier as Ora- cle to label the selected instances. The experimental results show that the proposed algorithm has fast learning speed with exact labeling.

关 键 词：主动学习 极限学习机 在线序列学习 样例熵 K-近邻 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]