分布式多分类支持向量机

来源期刊:中南大学学报(自然科学版)2011年第z1期

论文作者:郭一楠 程健 肖大伟 杨梅

文章页码:829 - 834

关键词:分布式支持向量机;多分类;遗传算法;超参数

Key words:distributed support vector machines; multi-classification; genetic algorithm; hyper-parameters

摘    要:实际中存在大量基于多分类的大样本数据集的问题。针对单一SVM解决该类问题时计算时间长、占用内存大等缺点,提出了一种分布式多分类支持向量机模型,并针对其中的子多分类SVM模型的参数选择问题,采用遗传算法来寻找具有最优性能的超参数组合,即通过最小输出编码方式,将解决多分类问题的多个二分类支持向量机整合形成单个多分类支持向量机,并把多分类支持向量机的分类正确率作为遗传算法的适应度函数,所得超参数对应的子多分类SVM模型具有较好的泛化性能。最后将分布式多分类支持向量机模型用于手写数字识别中,通过对仿真结果的对比和分析,表明该模型具有较高的精度,提高了模型的泛化能力。

Abstract: Large samples problems based on multi-classification exist in practical. Aiming at the problem of higher computational complexity and larger computer memory by single support vector machines when dealing with such problem, a distributed multi-classification support vector machines model was established and genetic algorithm was adopted in selection of multi-classification support vector machines models’ hyper-parameters. Some binary classification support vector machines were made as a multi-classification support vector machines through minimum output code method and its classification accuracy rate was used for fitness function of genetic algorithm. Support vector machines with optimal hyper-parameters can ensure generalization. Taking the dataset of Handwritten Digit Recognition as an example, comparison and analysis show that the distributed support vector machines has higher accuracy rate and better generalization.

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号