基于支持向量机的并行学习方法研究  

On Parallel Learning Based on Support Vector Machines

作　　者：金潇明[1] 文益民[2] 

机构地区：[1]中南大学商学院 [2]湖南大学电气与信息工程学院,湖南长沙410082

出　　处：《湖南大学学报（自然科学版）》2008年第7期74-79,共6页Journal of Hunan University:Natural Sciences

基　　金：国家自然科学基金资助项目(70540014)

摘　　要：基于决策论中信息源合并规则,提出了基于支持向量机的并行学习方法(B-SVMs).在训练阶段,B-SVMs将一个大规模问题随机分解成若干规模较小的子问题,针对每个子问题同时训练支持向量机.在测试阶段,各个支持向量机并行给出测试结果,B-SVMs利用信息源合并规则将各支持向量机的输出组合而得到最终测试结果.在4个问题上与基于多数投票策略的组合支持向量机(MV-SVMs)和快速模块化支持向量机(FM-SVMs)相比,B-SVMs的泛化能力较高.与利用全部样本训练的标准支持向量机相比,B-SVMs极大地减少了训练时间和测试时间,而且具有与其相当的泛化能力.This paper has proposed a parallel combine SVMs on the basis of two information fusion rules （B-SVMs）. B-SVMs randomly decomposes a large-scale task into many smaller and simpler sub-tasks in the training phase and uses the two information fusion rules to make decisions for final classification in the test phase. B- SVMs has been compared with single SVMs that was trained on entire training data set, parallel SVMs combined by majority voting （MV-SVMs）, and one kind of fast modular SVMs （FM-SVMs）, Experiment results on four problems have shown that B-SVMs can get a higher accuracy than MV-SVMs and FM-SVMs. The proposed algorithm can significantly reduce training and test time. More importantly, it produces a test accuracy that is almost the same as single SVMs does.

关 键 词：并行处理系统 学习系统 支持向量机 分类器组合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]