动态数据流分析的在线超限学习算法综述

Survey of Online Sequential Extreme Learning Algorithms for Dynamic Data Stream Analysis

动态数据流分析是一个具有广泛应用价值的研究课题,在线学习方法是其中的一种关键技术。在众多在线学习方法中,在线贯序超限学习机(Online Sequential Extreme Learning Machine,OSELM)是一种新颖且实用的在线学习算法,目前已在动态数据流分析中得到了成功应用。首先,介绍了OSELM的理论基础和算法执行过程;然后,以动态数据流分析为应用背景,对各种改进OSELM算法进行了分类综述,包括基于滑动窗口的OSELM、基于遗忘因子的OSELM、基于样本加权的OSELM以及其他方法,重点论述了各类算法的设计思路和实现策略,并对其优缺点进行了比较和分析;最后,探讨了值得进一步研究的问题。

Dynamic data stream analysis has become a research focus for its widespread application prospects,and online learning method is key to solve this problem.Among existing online learning methods,online sequential extreme lear-ning machine (OSELM) is a novel and practical online learning algorithm,and it has been successfully applied in the field of dynamic data stream analysis.Firstly,the theoretical foundation and the execution process of OSELM were reviewed.Then,regarding dynamic data flow analysis as the application background,this paper classified and summarized different kinds of improved OSELM algorithms,including the sliding window based OSELM,forgetting factor based OSELM,sample weighting based OSELM and other methods.This paper focused on the design ideas and implementation strategies of different kinds of algorithms, compared and analyzed the advantages and disadvantages of main algorithms.Finally,the possible works for future research were presented.