计算机工程 ›› 2013, Vol. 39 ›› Issue (1): 191-194.doi: 10.3969/j.issn.1000-3428.2013.01.041

• 人工智能及识别技术 • 上一篇    下一篇

双支持向量回归的牛顿算法

郑逢德,张鸿宾   

  1. (北京工业大学计算机学院,北京 100124)
  • 收稿日期:2011-10-14 修回日期:2011-12-20 出版日期:2013-01-15 发布日期:2013-01-13
  • 作者简介:郑逢德(1980-),男,博士研究生,主研方向:核方法,机器学习;张鸿宾,教授、博士生导师
  • 基金项目:

    国家自然科学基金资助项目(60775011)

Newton Algorithm of Twin Support Vector Regression

ZHENG Feng-de, ZHANG Hong-bin   

  1. (College of Computer Science, Beijing University of Technology, Beijing 100124, China)
  • Received:2011-10-14 Revised:2011-12-20 Online:2013-01-15 Published:2013-01-13

摘要: 为提高支持向量回归的运算速度,提出一种双支持向量回归的牛顿算法。求解2个只带一组约束的支持向量问题,以减少运算量,将2个约束优化问题转化为无约束最优化问题,并采用牛顿迭代算法求解。实验结果表明,在保证与支持向量回归和双支持向量回归拟合能力相当的同时,该算法能减少训练时间。

关键词: 机器学习, 模式识别, 支持向量回归, 双支持向量回归, 无约束优化, 牛顿算法

Abstract: For improving the learning speed of Support Vector Regression(SVR), this paper proposes a Newton algorithm for Twin Support Vector Regression(TSVR) that tries to find a pair of nonparallel planes by solving two related SVR-type problems and converts the classical Quadratic Programming Problem(QPP) to two small unconstrained optimization problems. Each of the unconstrained optimization problems is solved by Newton algorithm. Experimental results show that the proposed algorithm has good fitting ability as SVR and TSVR, and can reduce the training time.

Key words: machine learning, pattern recognition, Support Vector Regression(SVR), Twin Support Vector Regression(TSVR), unconstrained optimization, Newton algorithm

中图分类号: