SVM+模型中可用信息用作特权信息

doi:10.12066/j.issn.1007-2861.1675

上海大学学报(自然科学版) ›› 2017, Vol. 23 ›› Issue (4): 524-534.doi: 10.12066/j.issn.1007-2861.1675

SVM+模型中可用信息用作特权信息

董勇, 孙广玲, 刘志

上海大学通信与信息工程学院, 上海 200444

收稿日期:2015-09-10 出版日期:2017-08-30 发布日期:2017-08-30
通讯作者: 孙广玲(1973—), 女, 副教授, 博士, 研究方向为模式识别、机器学习等. E-mail: sunguangling@shu.edu.cn
作者简介:孙广玲(1973—), 女, 副教授, 博士, 研究方向为模式识别、机器学习等. E-mail: sunguangling@shu.edu.cn
基金资助:
教育部科学技术研究重点资助项目(212053); 上海市自然科学基金资助项目(16ZR1411100)

Using available information as privileged information in SVM+

DONG Yong, SUN Guangling, LIU Zhi

School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China

Received:2015-09-10 Online:2017-08-30 Published:2017-08-30

摘要/Abstract

摘要：

在机器学习中, 当测试阶段无法得到训练阶段拥有的特权信息时, 特权学习(learning using privileged information, LUPI) 是一个有效的解决框架. 由于获取特权信息需要特殊的条件, 或由于其他原因, 往往不能获得全部训练样本的特权信息, 因此提出了一种直观却有效的方法. 对于缺失特权信息的这部分训练样本, 将它们的可用信息同时用作特权信息, 并将其纳入到支持向量机(support vector machine+, SVM+) 的模型中, 引入了一种新的扩展SVM+(extended SVM+, eSVM+) 模型. 进一步地, 对于不涉及特权信息的常规有监督学习问题, 也将训练样本的特征(可用信息)同时用作特权信息, 引出一种新的扩展SVM 模型(eSVM), eSVM 也可认为是SVM+ 的特例. 在两个公开的人脸表情数据库BU-3DFE 和Bosphorus 上进行了实验, 结果证实了将可用信息用作特权信息策略的有效性.

关键词: 可用信息, 特权信息, 支持向量机, 特权学习

Abstract:

In machine learning, some information is only available during learning phase. Learning using privileged information (LUPI) can provide an effective solution to the problem. Such information is called privileged information. To deal with the issue that only partial privileged information of training data is available, this paper presents an intuitive but effective strategy called extended support vector machine+ (eSVM+). Specifically, for data without privileged information, available information is used as privileged information simultaneously and further cooperate it into the SVM+ formulation. In addition, for a regular supervised learning paradigm, a similar idea is adopted that all training data are both available and privileged. Naturally, it is extended SVM (eSVM), and also a special aspect of SVM+. Experimental results show that the proposed strategy can boost generalization performance of the classifier on two benchmark expression databases, BU-3DFE and Bosphorus.

Key words: available information, privileged information, support vector machine+ (SVM+), learning using privilegend information (LUPI)

董勇, 孙广玲, 刘志. SVM+模型中可用信息用作特权信息[J]. 上海大学学报(自然科学版), 2017, 23(4): 524-534.

DONG Yong, SUN Guangling, LIU Zhi. Using available information as privileged information in SVM+[J]. Journal of Shanghai University（Natural Science Edition）, 2017, 23(4): 524-534.

参考文献

[1] Vapnik V, Vashist A, Pavlovitch N. Learning using hidden information (learning with teacher) [C]//International Joint Conference on Neural Networks. 2009.
[2] Vapnik V, Vashist A. A new learning paradigm: learning using privileged information [J]. Neural Networks, 2009, 22(5/6): 544-557.
[3] Christoph L. Learning with asymmetric information [EB/OL]. [2015-09-01]. http://pub.ist.ac.at/~chl/talks/lampert-us2014.pdf.
[4] Jan F, Uwe A. Privileged information for data clustering [J]. Information Sciences, 2012, 194: 4-23.
[5] Ribeiro B, Silva C, Vieira A, et al. Financial distress model prediction using SVM+ [C]//International Joint Conference on Neural Networks. 2010: 1-7.
[6] Pechyony D, Vapnik V. On the theory of learning with privileged information [C]//NIPS. 2010: 1894-1902.
[7] Xu X X, Li W, Xu D. Distance metric learning using privileged information for face verification and person re-identification [J]. IEEE Transactions on Neural Networks and Learning Systems,2015, 26(12): 3150-3162.
[8] Hernandez-Lobato D, Sharmanska V, Kersting K, et al. Mind the nuisance: Gaussian process classification using privileged noise [J]. Advances in Neural Information Processing Systems, 2014(1): 837-845.
[9] Lapin M, Hein M, Schiele B. Learning using privileged information: SVM+ and weighted SVM [J]. Neural Networks, 2014, 53: 95-108.
[10] Feyereisl J, Kwak S, Son J, et al. Object localization based on structural SVM using privileged information [J]. Advances in Neural Information Processing Systems, 2014(1): 208-216.
[11] Yang H, Patras I. Privileged information based conditional structured output regression forest for facial point detection [J]. IEEE Transactions on Circuits and Systems for Video Technology, 2015, 25(9): 1507-1520.
[12] Viktoriia S, Novi Q, Christoph H, et al. Learning to rank using privileged information [C]//ICCV. 2013: 825-832.
[13] Wang S Z, Tao D C, Yang J. Relative attribute SVM+ learning for age estimation [J]. IEEE Transactions on Cybernetics, 2016, 46(3): 827-839.
[14] Pechyony D, Vapnik V. Fast optimization algorithms for solving SVM+ [EB/OL]. [2015-09-05]. http://www.cs.technion.ac.il/pechyony/book chapter.pdf.
[15] Viktoriia S, Novi Q, Christoph L. Learning to transfer privileged information [J]. Computer Science, 2014, arXiv: 1410.0389v1.
[16] Valstar M, Girard J, Almaev T, et al. Fera 2015-second facial expression recognition and analysis challenge [C]// IEEE International Conference on Automatic Face and Gesture Recognition. 2015: 1-18.
[17] Yin L, Wei X, Wang J, et al. A 3D facial expression database for facial behavior research [C]//IEEE International Conference on Automatic Face and Gesture Recognition. 2006: 211-216.
[18] Savran A. Bosphorus database for 3D face analysis [C]//Proc COST Workshop Biometrics Identity Manag. 2008: 47-56.
[19] Ahonen T, Hadid A, Pietikainen M. Face description with local binary patterns: application to face recognition [J]. IEEE T-PAMI, 2006, 28(12): 2037-2041.

SVM+模型中可用信息用作特权信息

Using available information as privileged information in SVM+

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 10

编辑推荐

Metrics

本文评价

[1]	楼志挺, 李春祥. 基于空间多点输入的 LSSVM 非高斯风压预测[J]. 上海大学学报(自然科学版), 2019, 25(6): 1013-1022.
[2]	胡嘉成, 王向阳, 刘晗. 基于深度学习的连铸坯表面缺陷检测[J]. 上海大学学报(自然科学版), 2019, 25(4): 445-452.
[3]	董惠忠, 毕艳玖, 赵晓华, 葛炯, 沙云菲. 基于致香成分的上部烟叶和中部烟叶分类判别[J]. 上海大学学报(自然科学版), 2019, 25(2): 309-316.
[4]	涂伟平, 李春祥. 基于混合智能算法优化LSSVM的短期风压预测[J]. 上海大学学报(自然科学版), 2019, 25(2): 347-356.
[5]	徐言沁, 李春祥. 基于优化组合核最小二乘支持向量机的脉动风速预测[J]. 上海大学学报(自然科学版), 2018, 24(4): 627-633.
[6]	钟旺, 李春祥. 基于极限学习机的非平稳下击暴流风速预测[J]. 上海大学学报(自然科学版), 2018, 24(3): 446-455.
[7]	钱权,陈孟,桂林开. 特征约简在入侵检测上的应用[J]. 上海大学学报(自然科学版), 2010, 16(1): 86-90.
[8]	高峻峻1，谭崇利1，刘悦2，尹亚锋2. 一类产品需求预测的支持向量机方法[J]. 上海大学学报(自然科学版), 2009, 15(1): 71-76.
[9]	雷咏梅;王雄;郭恒明;金亨科. 一种基于支持向量机的并行训练策略[J]. 上海大学学报(自然科学版), 2007, 13(5): 545-549 .
[10]	李国平;路长厚;李健美. 基于最小二乘支持向量机的压印字符识别方法[J]. 上海大学学报(自然科学版), 2007, 13(2): 125-129 .