吉林大学学报(理学版)

• 数学 • 上一篇    下一篇

基于新的初始化策略计算MVEE的积极集算法

丛伟杰, 何磊   

  1. 西安邮电大学 理学院, 西安 710121
  • 收稿日期:2016-12-27 出版日期:2017-09-26 发布日期:2017-09-26
  • 通讯作者: 丛伟杰 E-mail:wjcong@xupt.edu.cn

ActiveSet Algorithm for Computing MVEEBased on New Initialization Strategy

CONG Weijie, HE Lei   

  1. School of Science, Xi’an University of Posts and Telecommunications, Xi’an 710121, China
  • Received:2016-12-27 Online:2017-09-26 Published:2017-09-26
  • Contact: CONG Weijie E-mail:wjcong@xupt.edu.cn

摘要: 针对计算最小体积闭包椭球(MVEE)的积极集算法中原初始化策略耗时较多的问题, 先给出一个基于样本协方差矩阵构造的新初始化策略, 然后将该初始化策略应用于秩2更新算法中, 并给出一个计算MVEE改进的积极集算法. 数值实验结果表明, 基于新的初始化策略的积极集算法能有效提高求解大规模数据集MVEE问题的计算效率.

关键词: 样本协方差矩阵, 最小体积闭包椭球, 积极集, 初始化策略, 大规模数据集

Abstract: Aiming at the problem that the original initialization strategy of the activeset algorithm took more time to compute minimum volume enclosing ellipsoid (MVEE). Firstly, we gave a new initialization strategy based on sample covariance matrix. Secondly, we applied the initialization strategy to the ranktwo update algorithm, and gave a modified activeset algorithm to compute MVEE. The results of numerical experiments show that activeset algorithm based on the new initialization strategy can effectively improve the computational efficiency of solving the MVEE problem of largescale date sets.

Key words: largescale date set, minimum volume enclosing ellipsoid, activeset, sample covariance matrix, initialization strategy

中图分类号: 

  • O221.2