Unlabeled Data Help in Graph-Based Semi-Supervised Learning: A Bayesian Nonparametrics Perspective

被引：0
作者：
Sanz-Alonso, Daniel [1 ]
Yang, Ruiyi [2 ]
机构：
[1] Univ Chicago, Dept Stat, Chicago, IL 60637 USA
[2] Univ Chicago, Comm Comp & Appl Math, Chicago, IL 60637 USA
来源：
JOURNAL OF MACHINE LEARNING RESEARCH | 2022年 / 23卷
关键词：
semi-supervised learning; graph-Laplacians; Bayesian nonparametrics; Gaus-sian fields on manifold; POSTERIOR DISTRIBUTIONS; CONVERGENCE-RATES; LAPLACIAN;
D O I：
暂无
中图分类号：
TP [自动化技术、计算机技术];
学科分类号：
0812 ;
摘要：
In this paper we analyze the graph-based approach to semi-supervised learning under a manifold assumption. We adopt a Bayesian perspective and demonstrate that, for a suitable choice of prior constructed with sufficiently many unlabeled data, the posterior contracts around the truth at a rate that is minimax optimal up to a logarithmic factor. Our theory covers both regression and classification.
引用
收藏
页数：28
相关论文
共 52 条
[21]  
Garcia Trillos N., 2019, Foundations of Computational Mathematics, V20, P1
[22]   Convergence rates of posterior distributions [J].
Ghosal, S ;
Ghosh, JK ;
Van der Vaart, AW .
ANNALS OF STATISTICS, 2000, 28 (02) :500-531
[23]  
Ghosal S, 2017, CA ST PR MA, V44
[24]   Convergence rates of posterior distributions for noniid observations [J].
Ghosal, Subhashis ;
Van Der Vaart, Aad .
ANNALS OF STATISTICS, 2007, 35 (01) :192-223
[25]  
Hady M, 2006, IEEE Transactions on Neural Networks, V172, P530, DOI [DOI 10.1109/TNN.2009.2015974, 10.1109/TNN.2009.2015974]
[26]   Kernel Methods for Bayesian Elliptic Inverse Problems on Manifolds [J].
Harlim, John ;
Sanz-Alonso, Daniel ;
Yang, Ruiyi .
SIAM-ASA JOURNAL ON UNCERTAINTY QUANTIFICATION, 2020, 8 (04) :1414-1445
[27]  
Hein Matthias, 2005, P 22 INT C MACHINE L, P289, DOI [10.1145/1102351.1102388, DOI 10.1145/1102351.1102388]
[28]  
Hoffmann F, 2020, J MACH LEARN RES, V21
[29]   Estimating a smooth function on a large graph by Bayesian Laplacian regularisation [J].
Kirichenko, Alisa ;
van Zanten, Harry .
ELECTRONIC JOURNAL OF STATISTICS, 2017, 11 (01) :891-915
[30]  
Kushpel Alexander, 2015, ARXIV150406508
123456