[1]文益民,廖洪元,周立华.一种可减少训练时间的分层并行支持向量机方法[J].南京师范大学学报(工程技术版),2005,05(01):008-11.
　WEN Yimin,LIAO Hongyuan,ZHOU Lihua.A Hierarchical and Parallel Support Vector Machines Algorithm for Reducing the Training Time[J].Journal of Nanjing Normal University(Engineering and Technology),2005,05(01):008-11.
点击复制

一种可减少训练时间的分层并行支持向量机方法
分享到：

南京师范大学学报（工程技术版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
05卷
期数:
2005年01期
页码:
008-11
栏目:
出版日期:
2005-03-30

文章信息/Info

Title:
A Hierarchical and Parallel Support Vector Machines Algorithm for Reducing the Training Time
作者:
文益民; 廖洪元; 周立华
湖南工业职业技术学院信息工程系, 湖南长沙410007
Author(s):
WEN Yimin;  LIAO Hongyuan;  ZHOU Lihua
Department of Information Engineering, Hunan Industry Polytechnic, Hunan Changsha 410007, China
关键词:
分层筛选;  支持向量机;  交叉合并
Keywords:
h ie rarch ica l filtering;  support vectorm ach ines;  cross- comb in ing
分类号:
TP181
摘要:
基于支持向量的本质和并行计算方法,提出了一种新的分层并行的机器学习方法以加速支持向量机的训练过 程.该方法首先按照分而治之的思想将原分类问题分成若干子问题,然后将支持向量机的训练过程分解成级联的两个层次, 在每层采用并行的方法训练各个子支持向量机.各层训练集中的非支持向量被逐步筛选掉,交叉合并的规则保证问题的一 致性.仿真结果表明该方法在保证分类器推广能力的同时,缩短了训练支持向量机的时间.
Abstract:
Based on the essence of suppo rt vec to rs and pa ra lle l a lgor ithm, the paper proposes a nove l strategy of filtering the tra in ing sam ples in a hierarchica l and para llel w ay to speed up the tra ining o f support vector m ach ines ( SVM s). Dur ing the training pro cess, the entire c lassifica tion problem is d iv ided into several sm a ll sub- problem s that can be handled in a para llel w ay. H av ing h ie rarch ically filtered out the non- support-vec to r data, w e can obta in the final training data se t, wh ich is used to tra in a SVM tha t w ill be used as the final pattern c lassifie r. In order to keep the consistency, the cross- comb ining pr inciple is introduced. The simu la tion resu lts illustrate that ourm ethod speeds up tra in ing wh ilem a inta in ing the genera lization accuracy o f SVM s

参考文献/References:

[ 1] V lad im ir N. Vapnik. S tatistical Learn ing Theo ry [M ]. NewYork: Spr inger-Ver lag, 1998.
[ 2] 许建华, 张学工, 李衍达. 支持向量机的新发展[ J]. 控制与决策, 2004, 19: 481-484.
[ 3] Edga r Osuna, Robert Freund, Federico G iros.i An improved training algor ithm fo r support vector m ach ines [ A ]. Proceed ing s of IEEE [ C ] . NNSP, 1997. 276- 285.
[ 4 ] Thorsten Joach im s. M aking large-sca le SVM learning pratica l[ A]. Advances in Kerne lM ethods-Support Vecto r Lea rning, Cambr idg e [ C ] . M IT Press, 2000. 169- 184.
[ 5] Lu Baoliang, W angKa ian, U tiyam aM, et al. A part-versus- part m ethod for m ass ive ly pa ra lle l tra in ing o f support vecto r machines [ A ]. Proceedings o f IEEE / INNS Int. Joint Con.f on Neural Netw orks ( IJCNN2004 ) [ C ]. H ungary: Budapest, 2004. 735-740.
[ 6] Anton Schwa ighofe r, V loker Tresp. The bayesian comm i-t tee support vecto r m ach ine [ J] . Lecture No tes in Compute r Science, 2001, 2130: 411-417.
[ 7] N adeem Ahm ed Syed, H uan Liu, Kah Kay Sung. Increm enta l learn ing w ith suppo rt vector m ach ines[ A] . Proceedings o f theWo rkshop on Support V ectorM ach ines at the In ternational Jo int Confe rence on Artificia l In tell-i gence[ C]. Sw eden: S to ckho lm, 1999.
[ 8] B lake C L, M erz C J. UCI( ftp: / / ftp. ics. uc.i edu /pub / m ach ines- lea rning-database).

相似文献/References:

[1]程学云,吉根林,凌霄汉.基于SVM的多类代价敏感学习及其应用[J].南京师范大学学报(工程技术版),2006,06(04):079.
　CHENG Xueyun,J I Ge n lin,et al.SVM-Based Multiclass Cost-Sensitive Learning and its Application[J].Journal of Nanjing Normal University(Engineering and Technology),2006,06(01):079.

备注/Memo

备注/Memo:
基金项目: 湖南省青年骨干教师资助项目(湘教通[ 2001] 204号) .
作者简介: 文益民( 1969 - ) , 上海交通大学博士研究生, 副教授, 主要从事统计理论、生物信息学等方面的教学与研究.E-m ail:y im in-w en@ 163. com
更新日期/Last Update: 2013-04-29