计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (16): 46-50.

• 理论与研发 • 上一篇    下一篇

MSVR算法中系数矩阵条件数的估计研究

邹远洋,安聪沛   

  1. 暨南大学 信息科学技术学院,广州 510632
  • 出版日期:2016-08-15 发布日期:2016-08-12

Study on condition number estimation for coefficient matrix of MSVR algorithm

ZOU Yuanyang, AN Congpei   

  1. School of Information Science and Technology, Jinan University, Guangzhou 510632, China
  • Online:2016-08-15 Published:2016-08-12

摘要: 基于支持向量机的基本原理及线性方程组中条件数的估计理论,对多输出支持向量机回归(MSVR,Multi-output Support Regression)算法中线性方程组系数矩阵进行分析。给出了其条件数估值范围的定理。此结果为支持向量机中核函数的选取提供了有效的方法。通过实际数据做出的数值仿真实验验证了理论结果的正确性与实用性。

关键词: 多输出支持向量机回归(MSVR), 系数矩阵, 条件数, 估计

Abstract: Based on the fundamental principle of the Support Vector Machine(SVM) and the theory of condition number estimation of a linear system. The coefficient matrix of the Multi-output Support Regression(MSVR) algorithm is analyzed. Moreover, a condition number estimation for coefficient matrix of the linear system is presented. This result yields a helpful method to select kernel function of SVM. Finally, the correctness and practicability of theoretical result is supported by the numerical simulation from practical data.

Key words: Multi-output Support Regression(MSVR), coefficient matrix, condition number, estimation