一种基于极端学习机的半监督学习方法  

A semi-supervised learning method based on extreme learning machine

作　　者：唐晓亮[1] 韩敏[1] 

机构地区：[1]大连理工大学信息与通信工程学院,辽宁大连116024

出　　处：《大连理工大学学报》2010年第5期771-776,共6页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(60674073);"八六三"国家高技术研究发展计划资助项目(2007AA04Z158);"十一五"国家科技支撑计划资助项目(2006BAB14B05);"九七三"国家重点基础研究发展计划资助项目(2006CB403405)

摘　　要：针对半监督学习方法存在的学习速度缓慢、不确定性递增等问题,提出一种基于极端学习机的半监督学习方法.该方法将极端学习机从监督学习模式扩展到半监督学习模式,以输出阈值向量控制标记样本的扩充程度,利用"换位"策略评估扩充标记样本中不确定性的影响.仿真结果表明,所提方法能够显著提高半监督学习的速度并有效减小对标记样本的依赖程度.A semi-supervised learning method based on extreme learning machine（ELM）is proposed to solve two problems in semi-supervised learning,i.e.the slow learning speed and the increasing uncertainty.In the proposed method,firstly,ELM is extended from supervised learning to semisupervised learning.Secondly,the output threshold vector is utilized to control the extensibility degree of label samples.Finally,the alternate detection strategy is employed to evaluate the influence of uncertainty in extended samples.It is indicated by simulation results that the semi-supervised learning speed is significantly improved and the dependence for labeled samples is effectively reduced.

关 键 词：半监督 极端学习机 分类 神经网络 

分 类 号：TP138[自动化与计算机技术—控制理论与控制工程]