Globality and locality incorporation in distance metric learning

Globality and locality incorporation in distance metric learning
复制标题

远程度量学习中的全局性和局部性合并

DOI:
10.1016/j.neucom.2013.09.041
复制
发表时间:
2014-04
期刊:
影响因子:
6
通讯作者:
Zeng-Fu Wang
Zeng-Fu Wang
中科院分区:
计算机科学2区
文献类型:
--
作者:
Wei Wang;胡包钢;Zeng-Fu Wang

文献摘要

参考文献

被引文献

相似文献

有监督的距离度量学习对统计分类和信息检索的成功起着重要的作用。尽管提出了许多相关的算法,但关于将几何信息(即,位置)和标签信息(即,在度量学习中的全局性。在本文中,我们提出了一种新的度量学习框架,称为“依赖最大化的度量学习”(DMML),它可以有效地将这两个信息源集成到一个统一的结构作为凸规划的实例,而不需要平衡权重。在DMML中,通过最大化再生核希尔伯特空间(RKHS)中数据分布之间的依赖性来训练度量。然而,与现有的信息论算法中的学习不同,DMML不需要估计或假设数据分布。在此框架下,我们提出了两种方法,分别采用不同的独立性标准,即,Hilbert-Schmidt独立性准则和广义距离协方差。分类、可视化和图像检索的综合实验结果表明,DMML的性能优于现有的度量学习算法,同时也说明了这两种方法在相关应用中的各自优势。
Supervised distance metric learning plays a substantial role to the success of statistical classification and information retrieval. Although many related algorithms are proposed, it is still an open problem about incorporating both the geometric information (i.e., locality) and the label information (i.e., globality) in metric learning. In this paper, we propose a novel metric learning framework, called “Dependence Maximization based Metric Learning” (DMML), which can efficiently integrate these two sources of information into a unified structure as instances of convex programming without requiring balance weights. In DMML, the metric is trained by maximizing the dependence between data distributions in the reproducing kernel Hilbert spaces (RKHSs). Unlike learning in the existing information theoretic algorithms, however, DMML requires no estimation or assumption of data distributions. Under this proposed framework, we present two methods by employing different independence criteria respectively, i.e., Hilbert–Schmidt Independence Criterion and the generalized Distance Covariance. Comprehensive experimental results for classification, visualization and image retrieval demonstrate that DMML favorably outperforms state-of-the-art metric learning algorithms, meanwhile illustrate the respective advantages of these two proposed methods in the related applications.
DOI: 10.1145/1273496.1273599
发表时间: 2007-06
期刊: The Journal of general virology
影响因子: --
作者:
Le Song;Alex Smola;A. Gretton;Karsten M. Borgwardt
通讯作者: Le Song;Alex Smola;A. Gretton;Karsten M. Borgwardt
DOI: 10.1007/3-540-33486-6_8
发表时间: 2001-01
期刊: --
影响因子: --
作者:
N. Cristianini;J. Shawe-Taylor;A. Elisseeff;J. Kandola
通讯作者: N. Cristianini;J. Shawe-Taylor;A. Elisseeff;J. Kandola
DOI: 10.1145/1281192.1281209
发表时间: 2007-08
影响因子: 1.6
作者:
Jianhui Chen;Zheng Zhao;Jieping Ye;Huan Liu
通讯作者: Jianhui Chen;Zheng Zhao;Jieping Ye;Huan Liu
DOI: 10.1023/a:1011139631724
发表时间: 2001-01-01
影响因子: 19.5
作者:
Oliva, A;Torralba, A
通讯作者: Torralba, A
自适应超图学习及其在图像分类中的应用
DOI: 10.1109/tip.2012.2190083
发表时间: 2012-07-01
影响因子: 10.6
作者:
Yu, Jun;Tao, Dacheng;Wang, Meng
通讯作者: Wang, Meng