SVMs与FANNs的关系及分类性能对比  

Relationship between SVMs and FANNs and performance comparison in classification

作　　者：冷强奎[1] 刘福德[2] 阎琦[1] Leng Qiang- kui Liu Fu -de Yan Qi(College of Information Science and Technology, Bohai University, Jinzhou 121013, China Research and Teaching Institute of College Basics, Bohai University, Jinzhou 121013, China)

机构地区：[1]渤海大学信息科学与技术学院,辽宁锦州121013 [2]渤海大学大学基础教研部,辽宁锦州121013

出　　处：《渤海大学学报（自然科学版）》2016年第3期254-260,共7页Journal of Bohai University:Natural Science Edition

基　　金：国家自然科学基金项目(No:61602056);辽宁省教育厅科学研究一般项目(No:L2014444);2014年度辽宁省社会科学规划基金项目(No:L14BTJ002)

摘　　要：研究了支持向量机(SVMs)与前馈人工神经网络(FANNs)的关系.首先说明了两者在网络结构上的相似性,即输入函数均可表示为基函数的线性组合形式.然后指出了两者之间存在的关键差异:一是优化目标不同,FANNs只需要达到经验风险最小化,而SVMs寻求结构风险最小化;二是隐含层含义不同,SVMs的隐含层节点表现为支持向量,而FANNs则需要预先设定;三是模型复杂性不同,FANNs的模型复杂性由隐含层神经元数量来控制,而SVMs的复杂性独立于维度.最后,给出了两者在UCI标准数据集上的对比实验,用以评估它们在分类中的性能.This paper presents the relationship between support vector machines （ SVMs） and feed-forward artificial neural networks （ FANNs） .It first illustrates the similarity in terms of network structure , i.e., their in-put function can be expressed as a linear combination of basis functions .Then, some key differences between SVMs and FANNs are pointed out .One is the difference of optimization goal .FANNs only needs to reach empiri-cal risk minimization , but for SVMs , it aims at minimizing the structural risk .Another is the difference of hidden layer.In SVMs, the nodes of hidden layer are exactly support vectors .While in FANNs, the nodes of hidden layer can be empirically determined in advance .Moreover , the difference of model complexity is also concerned . For FANNs, its complexity is controlled by the number of hidden nodes .But the complexity of SVMs is inde-pendent of data dimensionality .At the end, the comparative experiments on UCI benchmark datasets are provid-ed, for evaluating the performance of SVMs and FANNs in classification .

关 键 词：模式识别 支持向量机 前馈人工神经网络 深度学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]