Calibrated multi-task learning

doi:10.1145/3219819.3219951

CORC > 西安光学精密机械研究所 > 中国科学院西安光学精密机械研究所 > 光学影像学习与分析中心

	Calibrated multi-task learning
	Nie, Feiping 1; Hu, Zhanxuan 1; Li, Xuelong 2
	2018-07-19
会议日期	2018-08-19
会议地点	London, United kingdom
DOI	10.1145/3219819.3219951
页码	2012-2021
英文摘要	This paper proposes a novel algorithm, named Non-Convex Calibrated Multi-Task Learning (NC-CMTL), for learning multiple related regression tasks jointly. Instead of utilizing the nuclear norm, NC-CMTL adopts a non-convex low rank regularizer to explore the shared information among different tasks. In addition, considering that the regularization parameter for each regression task desponds on its noise level, we replace the least squares loss function by square-root loss function. Computationally, as proposed model has a non-smooth loss function and a non-convex regularization term, we construct an efficient re-weighted method to optimize it. Theoretically, we first present the convergence analysis of constructed method, and then prove that the derived solution is a stationary point of original problem. Particularly, the regularizer and optimization method used in this paper are also suitable for other rank minimization problems. Numerical experiments on both synthetic and real data illustrate the advantages of NC-CMTL over several state-of-the-art methods. © 2018 Association for Computing Machinery.
产权排序	2
会议录	KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
会议录出版者	Association for Computing Machinery
语种	英语
ISBN号	9781450355520
内容类型	会议论文
源URL	[http://ir.opt.ac.cn/handle/181661/30575]
专题	西安光学精密机械研究所_光学影像学习与分析中心
作者单位	1.School of Computer Science, Center for OPTIMAL, Northwestern Polytechnical University, Xi'an, China; 2.OPTIMAL, Xian Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi'an, China
推荐引用方式 GB/T 7714	Nie, Feiping,Hu, Zhanxuan,Li, Xuelong. Calibrated multi-task learning[C]. 见:. London, United kingdom. 2018-08-19.