Submodular Hypergraphs: p-Laplacians, Cheeger Inequalities and Spectral Clustering

被引:0
作者
Li, Pan [1 ]
Milenkovic, Olgica [1 ]
机构
[1] Univ Illinois, Dept Elect & Comp Engn, Urbana, IL 61801 USA
来源
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80 | 2018年 / 80卷
基金
美国国家科学基金会;
关键词
1-LAPLACIAN;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We introduce submodular hypergraphs, a family of hypergraphs that have different submodular weights associated with different cuts of hyper-edges. Submodular hypergraphs arise in clustering applications in which higher-order structures carry relevant information. For such hypergraphs, we define the notion of p-Laplacians and derive corresponding nodal domain theorems and k-way Cheeger inequalities. We conclude with the description of algorithms for computing the spectra of 1- and 2-Laplacians that constitute the basis of new spectral hypergraph clustering methods.
引用
收藏
页数:10
相关论文
共 34 条
[11]   Nodal domains of eigenvectors for 1-Laplacian on graphs [J].
Chang, K. C. ;
Shao, Sihong ;
Zhang, Dong .
ADVANCES IN MATHEMATICS, 2017, 308 :529-574
[12]   Spectrum of the 1-Laplacian and Cheeger's Constant on Graphs [J].
Chang, K. C. .
JOURNAL OF GRAPH THEORY, 2016, 81 (02) :167-207
[13]  
Chekuri C, 2017, PROCEEDINGS OF THE TWENTY-EIGHTH ANNUAL ACM-SIAM SYMPOSIUM ON DISCRETE ALGORITHMS, P1085
[14]  
Chung F., 1992, Spectral Graph Theory
[15]  
Davies EB, 2001, LINEAR ALGEBRA APPL, V336, P51
[16]  
Ene A, 2015, PR MACH LEARN RES, V37, P787
[17]   Structured learning of sum-of-submodular higher order energy functions [J].
Fix, Alexander ;
Joachims, Thorsten ;
Park, Sung Min ;
Zabih, Ramin .
2013 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 2013, :3104-3111
[18]  
Gould Sydney Henry, 1966, VARIATIONAL METHODS, V22
[19]  
Hein M., 2013, P ADV NEURAL INFORM, P2427
[20]  
Jegelka S., 2013, Advances in Neural Information Processing Systems, P1313