Low rank metric learning via accelerated gradient optimization | |
Shen, Yuanyuan ; Yan, Yan ; Wang, Hanzi ; Yan Y(严严) ; Wang HZ(王菡子) | |
2014 | |
关键词 | Gradient methods Internet Learning algorithms |
英文摘要 | Conference Name:6th International Conference on Internet Multimedia Computing and Service, ICIMCS 2014. Conference Address: Xiamen, China. Time:July 10, 2014 - July 12, 2014.; National Natural Foundation of China; SIGMM China Chapter; Xiamen University; A basic assumption in many computer vision applications is that the underlying data lie on a low-dimensional subspace. This property has been widely exploited by metric learning algorithms to find a low-dimensional projection of the feature space such that relevant dimensions are high-weighted, while irrelevant dimensions are low-weighted. A sparse or a low rank metric is an effective solution in general. In this paper, we propose to perform metric learning with the low rank constraint. To effectively solve the proposed low rank metric learning problem, we develop an efficient algorithm based on the accelerated gradient method. The proposed algorithm not only learns an effective metric, but also projects data into an intrinsic low-dimensional subspace. Finally, the proposed algorithm is compared with several state-of-the-art metric learning algorithms in the applications of k nearest neighbor classification and face recognition and competitive results are achieved. Copyright 2014 ACM. |
语种 | 英语 |
出处 | http://dx.doi.org/10.1145/2632856.2632932 |
出版者 | Association for Computing Machinery |
内容类型 | 其他 |
源URL | [http://dspace.xmu.edu.cn/handle/2288/86904] ![]() |
专题 | 信息技术-会议论文 |
推荐引用方式 GB/T 7714 | Shen, Yuanyuan,Yan, Yan,Wang, Hanzi,et al. Low rank metric learning via accelerated gradient optimization. 2014-01-01. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论