《计算机应用研究》|Application Research of Computers

多核学习中基于复合梯度映射的学习算法研究

Research on learning algorithm based on composite gradient mapping in multiple kernel learning

免费全文下载 (已被下载 次)  
获取PDF全文
作者 龙文光,刘益和
机构 内江师范学院 a.现代教育技术中心;b.计算机科学学院,四川 内江 641112
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2015)04-1019-05
DOI 10.3969/j.issn.1001-3695.2015.04.014
摘要 现有的多核学习算法大多假设训练样本分类完全正确,将其应用到受扰分类样本上时,由于分类存在差错,因此往往只能实现次优性能。为了解决这一问题,首先将受扰分类多核学习问题建模为随机规划问题,并得到一种极小极大表达式;然后提出基于复合梯度映射的一阶学习算法对问题进行求解。理论分析表明,该算法的收敛速度为O(1/T),大大快于传统算法的收敛速度O(1/T)。最后,基于五个UCI数据集的实验结果也验证了本文观点和优化算法的有效性。
关键词 多核学习;训练样本;随机规划;复合梯度映射;收敛速度;UCI数据集
基金项目
本文URL http://www.arocmag.com/article/01-2015-04-014.html
英文标题 Research on learning algorithm based on composite gradient mapping in multiple kernel learning
作者英文名 LONG Wen-guang, LIU Yi-he
机构英文名 a. Modern Educational Technology Center, b. College of Computer Science, Neijiang Normal University, Neijiang Sichuan 641112, China
英文摘要 The existing multiple kernel learning algorithms assume that the training sample classification entirely correct, when they used to the classification of disturbed samples, the suboptimal performance could only be achieved due to the incorrect class assignments. In order to solve this problem, firstly, it modeled the multiple kernel learning problems from noisy labels into a stochastic programming problem, and presenting a min-max formulation, and then proposed a first order learning algorithm based on composite gradient mapping to solve this problem. The theoretical analysis shows that, the convergence rate of O(1/T) for the proposed algorithm, significantly faster than the classical O(1/T) rate. Finally, the experimental results on five UCI data sets confirm the effectiveness and the efficiency of the proposed framework and the optimization algorithm.
英文关键词 multiple kernel learning(MKL); training sample; stochastic programming; composite gradient mapping; convergence rate; UCI data sets
参考文献 查看稿件参考文献
  [1] 武征鹏, 张学工. 弹性多核学习[J] . 自动化学报, 2011, 37(6):693-699.
[2] 汪洪桥, 孙富春, 蔡艳宁, 等. 多核学习方法[J] . 自动化学报, 2010, 36(8):1037-1050.
[3] YANG Tian-bao, JIN Rong, JAIN A K. Learning from noisy side information by generalized maximum entropy model[C] //Proc of the 27th International Conference on Machine Learning. 2010:1199-1206.
[4] LANCKRIET G R G, CRISTIANINI N, BARTLETT P, et al. Lear-ning the kernel matrix with semidefinite programming[J] . The Journal of Machine Learning Research, 2004, 11(5):27-72.
[5] CORTES C, MOHRI M, ROSTAMIZADEH A. Two-stage learning kernel algorithms[C] //Proc of the 27th International Conference on Machine Learning. 2010:239-246.
[6] BACH F R, LANCKRIET G R G, JORDAN M I. Multiple kernel learning, conic duality, and the SMO algorithm[C] //Proc of the 21st International Conference on Machine Learning. [S. l. ] :ACM Press, 2004:6-14.
[7] SONNENBURG S, RTSCH G, SCHFER C. A general and efficient multiple kernel learning algorithm[C] //Advances in Neural Information Processing Systems. 2005.
[8] RAKOTOMAMONJY A, BACH F, CANU S, et al. SimpleMKL[J] . Journal of Machine Learning Research, 2008, 34(9):2491-2521.
[9] XU Zeng-lin, JIN Rong, YANG Hai-qin, et al. Simple and efficient multiple kernel learning by group lasso[C] //Proc of the 27th International Conference on Machine Learning. 2010:1175-1182.
[10] 韩彦军, 王珏. 基于多核学习的双稀疏关系学习算法[J] . 计算机研究与发展, 2010, 10(8):1400-1406.
[11] 陶剑文, 王士同. 多核局部领域适应学习[J] . 软件学报, 2012, 23(9):2297-2310.
[12] RAMAKRISHNAN G, CHITRAPURA K P, KRISHNAPURAM R, et al. A model for handling approximate, noisy or incomplete labeling in text classification[C] //Proc of the 22nd International Conference on Machine Learning. New York:ACM Press, 2005:681-688.
[13] HUANG Kai-zhu, JIN Rong, XU Zeng-lin, et al. Robust metric learning by smooth optimization[C] //Proc of the 26th Conference on Uncertainty in Artificial Intelligence. 2012:227-236.
[14] BERTSIMAS D, BROWN D B, Caramanis C. Theory and applications of robust optimization[J] . SIAM Review, 2011, 53(3):464-501.
[15] NEMIROVSKI A. Prox-method with rate of convergence O(1/t) for variational inequalities with Lipschitz continuous monotone operators and smooth convex-concave saddle point problems[J] . SIAM Journal on Optimization, 2004, 15(1):229-251.
[16] MICCHELLI C A, PONTIL M. Learning the kernel function via regularization[J] . Journal of Machine Learning Research, 2005, 6:1099-1125.
[17] BOUCHERON S, LUGOSI G, BOUSQUET O. Concentration inequalities[C] //Advanced Lectures on Machine Learning. Berlin:Springer, 2004:208-240.
[18] GLER O. Duality theory and convex programming[C] //Proc of Foundations of Optimization. New York:Springer, 2010:275-312.
[19] KLOFT M, BREFELD U, SONNENBURG S, et al. Efficient and accurate lp-norm multiple kernel learning[C] //Proc of the 23rd An-nual Conference on Neural Information Processing Systems. 2009:997-1005.
收稿日期 2014/2/22
修回日期 2014/6/5
页码 1019-1023
中图分类号 TP181
文献标志码 A