上海交通大学学报(自然版) ›› 2011, Vol. 45 ›› Issue (08): 1140-1145.

• 自动化技术、计算机技术 • 上一篇    下一篇

极限学习机的快速留一交叉验证算法

刘学艺a,李平a,b,郜传厚c   

  1. (浙江大学 a.航空航天学院; b.工业控制研究所; c.数学系,杭州 310027)
  • 收稿日期:2011-03-15 出版日期:2011-08-30 发布日期:2011-08-30
  • 基金资助:

    国家高技术研究发展计划(863)项目(2006AA04Z184),国家自然科学基金资助项目(10901139,60911130510,60874029)

Fast Leave-One-Out Cross-Validation Algorithm for Extreme Learning Machine

 LIU  Xue-Yi-a, LI  Ping-a, b , GAO  Chuan-Hou-c   

  1. (a. School of Aeronautics and Astronautics; b. Institute of Industrial Process Control;c. Department of Mathematics, Zhejiang University, Hangzhou 310027, China)
  • Received:2011-03-15 Online:2011-08-30 Published:2011-08-30

摘要: 针对回归和分类问题,提出一种极限学习机(Extreme Learning Machine, ELM)的快速留一交叉验证算法,并从理论和数值仿真两方面说明其有效性.结果表明,该算法避免了以训练样本数量N次的ELM模型的显式训练,其计算复杂度与N仅呈线性趋势增长,即O(N).即使在处理大型数据集建模问题时,该算法仍然可以快速地进行ELM模型的选择和评价.通过人工和实际数据集上的仿真实验,验证了该快速留一交叉验证算法的有效性.

关键词: 极限学习机, 留一法, 交叉验证, 计算复杂性

Abstract:  Leaveoneout crossvalidation has proved to be near capable of giving the unbiased estimation of the generalization performance of statistical models, and thus can provide a reliable criterion for model selection and comparison. For this reason, the current paper presented a fast leaveoneout crossvalidation algorithm in the framework of extreme learning machines (ELMs) with respect to both regression and classification problems, which can avoid training explicitly and just has the complexity of O(N) for a data set with N points. The validity of the algorithm is also strictly proved. The simulations conducted on the artificial and realworld problems show the effectiveness and efficiency of the proposed algorithm.

Key words: extreme learning machine (ELM), leaveoneout, crossvalidation, computational complexity

中图分类号: