您的位置：山东大学 -> 科技期刊社 -> 《山东大学学报（理学版）》

《山东大学学报(理学版)》 ›› 2022, Vol. 57 ›› Issue (4): 12-20.doi: 10.6040/j.issn.1671-9352.7.2021.149

• • 上一篇    

基于约束分层加权的多度量学习算法

韩露1,郭鑫垚1,魏巍1,2,梁吉业1,2*   

  1. 1.山西大学计算机与信息技术学院, 山西 太原 030006;2.山西大学计算智能与中文信息处理教育部重点实验室, 山西 太原 030006
  • 发布日期:2022-03-29
  • 作者简介:韩露(1997— ),女,硕士研究生,研究方向为机器学习.E-mail:3247415932@qq.com*通信作者简介:梁吉业(1962— ),博士,教授,博士生导师,研究方向为数据挖掘与机器学习.E-mail:ljy@sxu.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(61976184,62006147);山西省重点研发计划资助项目(201903D121162)

Multi-metric learning algorithm based on constraint hierarchical weighting

HAN Lu1, GUO Xin-yao1, WEI Wei1,2, LIANG Ji-ye1,2*   

  1. 1. School of Computer and Information Technology, Shanxi University, Taiyuan 030006, Shanxi, China;
    2. Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education, Shanxi University, Taiyuan 030006, Shanxi, China
  • Published:2022-03-29

PDF (PC)

317

摘要： 为解决经典的多度量学习算法从预先获得的样本划分中学习度量时,样本划分不精确所导致局部度量拟合能力不足的问题,基于约束分层加权的思想,提出了为约束逐层分配度量并使其度量尽可能满足当前所有约束的优化模型,同时添加正则项使得不同度量对应的约束应该尽可能不同。由于单个样本所形成的不同约束可能对应不同的局部度量,相比于传统的多度量学习方法,提出的算法能够获得更精细的局部度量且更具有灵活性,使得度量的拟合能力更强。实验结果表明,提出的算法在真实数据集上对比代表性的单度量学习算法和多度量学习算法具有明显的优势。

关键词: 度量学习, 约束加权, 马氏距离, 三元约束, 多度量学习

Abstract: In order to solve the problem of insufficient local metric fitting ability caused by inaccurate sample partitioning when the classic multi-metric learning algorithm learns metrics from sample partitions obtained in advance, based on the idea of constraint stratification weighting, this paper proposes to assign metrics to constraints layer by layer and makes the measurement as far as possible to meet the optimization model of all constraints, while adding regular terms to make the constraints corresponding to different metrics should be as different as possible. Since different constraints formed by a single sample may correspond to different local metrics, compared with traditional multi-metric learning methods, the proposed algorithm can obtain finer local metrics and is more flexible, making the metric's fitting ability stronger. Experimental results show that the proposed algorithm has obvious advantages compared with representative single-metric learning algorithms and multi-metric learning algorithms on real data sets.

Key words: metric learning, constraint weighting, Mahalanobis distance, triplet constraint, multi-metric learning

中图分类号: 

  • TP181
[1] XU Yaowen, WU Lifang, JIAN Meng, et al. Identity-constrained noise modeling with metric learning for face anti-spoofing[J]. Neurocomputing, 2021, 434:149-164.
[2] YU Jian, HU Changhui, JING Xiaoyuan, et al. Deep metric learning with dynamic margin hard sampling loss for face verification[J]. Signal Image and Video Processing, 2020, 14(4):791-798.
[3] HU Junlin, LU Jiwen, YUAN Junsong, et al. Large margin multi-metric learning for face and kinship verification in the wild[C] //Asian Conference on Computer Vision. Berlin: Springer, 2014: 252-267.
[4] CAO Rui, ZHANG Qian, ZHU Jiasong, et al. Enhancing remote sensing image retrieval using a triplet deep metric learning network[J]. International Journal of Remote Sensing, 2020, 41(2):740-751.
[5] 杜宇宁, 艾海舟. 基于二次相似度函数学习的行人再识别[J]. 计算机学报, 2016, 39(8):1639-1651. DU Yuning, AI Haizhou. Learning quadratic similarity function for pedestrian re-identification[J]. Chinese Journal of Computers, 2016, 39(8):1639-1651.
[6] JIN Yi, LI Chenning, LI Yidong, et al. Model latent views with multi-center metric learning for vehicle re-identification[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(3):1-13.
[7] SHEN C, WELSH A, WANG L. PSDBoost: matrix-generation linear programming for positive semidefinite matrices learning[C] //Advances in Neural Information Processing Systems 21. Cambridge, MA: MIT Press, 2008: 1473-1480.
[8] SHEN C, KIM J, WANG L, et al. Positive semidefinite metric learning with boosting[C] //Advances in Neural Information Processing Systems 22. Cambridge, MA: MIT Press, 2009: 1651-1659.
[9] BI Jinbo, WU Dijia, LU Le, et al. Adaboost on low-rank psd matrices for metric learning[C] //IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2011: 2617-2624.
[10] 齐忍, 朱鹏飞, 梁建青. 混杂数据的多核几何平均度量学习[J]. 软件学报, 2017, 28(11):2992-3001. QI Ren, ZHU Pengfei, LIANG Jianqing. Multiple kernel geometric mean metric learning for heterogeneous data[J]. Journal of Software, 2017, 28(11):2992-3001.
[11] ZHU Pengfei, CHENG Hao, HU Qinghua, et al. Towards generalized and efficient metric learning on Riemannian manifold[C] //International Joint Conference on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2018: 3235-3241.
[12] PAABEN B, GALLICCHIO C, MICHELI A, et al. Tree edit distance learning via adaptive symbol embeddings[C] //International Conference on Machine Learning. New York: ACM, 2018: 3976-3985.
[13] XING E P, NG A Y, JORDAN M I, et al. Distance metric learning with application to clustering with side-information[C] //Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press, 2002: 521-528.
[14] WEINBERGER K Q, BLITZER J, SAUL L K. Distance metric learning for large margin nearest neighbor classification[C] //Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press, 2006: 1473-1480.
[15] DAVIS J V, KULIS B, JAIN P, et al. Information-theoretic metric learning[C] //International Conference on Machine Learning. New York: ACM, 2007: 209-216.
[16] YING Yiming, LI Peng. Distance metric learning with eigenvalue optimization[J]. The Journal of Machine Learning Research, 2012, 13(1):1-26.
[17] ZUO Wangmeng, WANG Faqiang, ZHANG David, et al. Distance metric learning via iterated support vector machines[J]. IEEE Transactions on Image Processing, 2017, 26(10):4937-4950.
[18] BOHNE J, YING Y, GENTRIC S, et al. Large margin local metric learning[C] //European Conference on Computer Vision. Berlin: Springer, 2014: 679-694.
[19] YE Hanjia, ZHAN Dechuan, LI Nan, et al. Learning multiple local metrics: global consideration helps[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 42(7):1698-1712.
[20] MU Yang, DING Wei, TAO Dacheng. Local discriminative distance metrics ensemble learning[J]. Pattern Recognition, 2013, 46(8):2337-2349.
[21] WANG J, KALOUSIS A, WOZNICA A. Parametric local metric learning for nearest neighbor classification[C] //Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press, 2012: 1601-1609.
[22] CHEN Xianhong, HE Liang, XU Can, et al. Distance-dependent metric learning[J]. IEEE Signal Processing Letters, 2019, 26(2):357-361.
[23] WEINBERGER K Q, SAUL L K. Distance metric learning for large margin nearest neighbor classification[J]. Journal of Machine Learning Research, 2009, 10(2):207-244.
[24] SHI Y, BELLET A, SHA F. Sparse compositional metric learning[C] //The Association for the Advance of Artificial Intelligence. Menlo Park, CA: AAAI, 2014: 2078-2084.
[25] NGUYEN B, FERRI F J, MORELL C, et al. An efficient method for clustered multi-metric learning[J]. Information Sciences, 2019, 471: 149-163.
[26] YE Hanjia, ZHAN Dechuan, JIANG Yuan, et al. What makes objects similar: a unified multi-metric learning approach[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(5):1257-1270.
[27] YE Hanjia, ZHAN Dechuan, SI Xuemin, et al. Learning mahalanobis distance metric: considering instance disturbance helps[C] //International Joint Conference on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2017: 3315-3321.
[28] CHEN Shuo, GONG Chen, YANG Jian, et al. Adversarial metric learning[C] //International Joint Conference on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2018: 2021-2227.
[29] ABIN A, BASHIRI M A, BEIGY H. Learning a metric when clustering data points in the presence of constraints[J]. Advances in Data Analysis and Classification, 2020, 14(1):29-56.
[30] MAATEN L V D, HINTON G. Visualizing data using t-SNE[J]. The Journal of Machine Learning Research, 2008, 9(11):2579-2605.
[31] DEMSAR J. Statistical comparisons of classifiers over multiple data sets[J]. The Journal of Machine Learning Research, 2006, 7:1-30.
[1] 卢政宇,李光松,申莹珠,张彬. 基于连续特征的未知协议消息聚类算法[J]. 《山东大学学报(理学版)》, 2019, 54(5): 37-43.
[2] 刘国涛,张燕平,徐晨初. 一种优化覆盖中心的三支决策模型[J]. 山东大学学报（理学版）, 2017, 52(3): 105-110.
[3] 及歆荣,侯翠琴,侯义斌,赵斌. 基于筛选机制的L1核学习机分布式训练方法[J]. 山东大学学报（理学版）, 2016, 51(9): 137-144.
[4] 吴修国 曾广周. 目标描述逻辑研究[J]. J4, 2009, 44(11): 68-74.
[5] 许传轲 陈月辉 赵亚欧. 基于改进伪氨基酸组成的蛋白质相互作用预测[J]. J4, 2009, 44(9): 17-21.
[6] 修明 陈保会 史开泉. F-粗规律与它的属性控制[J]. J4, 2008, 43(12): 66-72.
[7] 付海艳,史开泉 . 知识过滤与属性f-迁移依赖[J]. J4, 2007, 42(10): 54-58 .
[8] 付海艳,卢昌荆,史开泉 . (F,F-)-规律推理与规律挖掘[J]. J4, 2007, 42(7): 54-57 .
[9] 张 凌,邱育峰,任雪芳 . 粗规律生成与它的分离[J]. J4, 2008, 43(3): 58-63 .
[10] 张 凌,邱育锋,任雪芳, . 基于单向S-粗集的知识堆垒与知识垛识别[J]. J4, 2008, 43(10): 12-17 .
[11] 林艳丽,刘晓东. 不完备序信息系统下的局部双量化决策粗糙集研究[J]. 《山东大学学报(理学版)》, 2020, 55(3): 89-97.
[12] 李敏,杨亚锋,雷宇,李丽红. 基于可拓域变化代价最小的最优粒度选择[J]. 《山东大学学报(理学版)》, 2021, 56(2): 17-27.
[13] 张聪, 于洪. 一种三支决策软增量聚类算法[J]. 山东大学学报（理学版）, 2014, 49(08): 40-47.
[14] 翟俊海, 张垚, 王熙照. 相容粗糙模糊集模型[J]. 山东大学学报（理学版）, 2014, 49(08): 73-79.
[15] 张要,马盈仓,杨小飞,朱恒东,杨婷. 结合流形结构与柔性嵌入的多标签特征选择[J]. 《山东大学学报(理学版)》, 2021, 56(7): 91-102.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2018 《山东大学学报(理学版)》编辑部　
地址: 山东省济南市山大南路27号山东大学中心校区（250100）　电话: +86-0531-88366917　E-mail: xblxb@sdu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发