Information-Theoretic Limits of Selecting Binary Graphical Models in High Dimensions

被引:128
作者
Santhanam, Narayana P. [1 ]
Wainwright, Martin J. [2 ,3 ]
机构
[1] Univ Hawaii, Dept Elect Engn, Honolulu, HI 96822 USA
[2] Univ Calif Berkeley, Dept Elect Engn & Comp Sci, Berkeley, CA 94720 USA
[3] Univ Calif Berkeley, Dept Stat, Berkeley, CA 94720 USA
基金
美国国家科学基金会;
关键词
High dimensional inference; KL divergence between Ising models; Markov random fields; sample complexity; structure of Ising models; DISTRIBUTIONS;
D O I
10.1109/TIT.2012.2191659
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
The problem of graphical model selection is to estimate the graph structure of a Markov random field given samples from it. We analyze the information-theoretic limitations of the problem of graph selection for binary Markov random fields under high-dimensional scaling, in which the graph size and the number of edges k, and/or the maximal node degree d, are allowed to increase to infinity as a function of the sample size n. For pairwise binary Markov random fields, we derive both necessary and sufficient conditions for correct graph selection over the class G(p,k) of graphs on p vertices with at most k edges, and over the class G(p,d) of graphs on p vertices with maximum degree at most. For the class , we establish the existence of constants and such that if, any method has error probability at least uniformly over the family, and we demonstrate a graph decoder that succeeds with high probability uniformly over the family for sample sizes n > c(1)k(2) log p. Similarly, for the class G(p,d), we exhibit constants and such that for n < cd(2) log p, anymethod fails with probability at least 1/2, and we demonstrate a graph decoder that succeeds with high probability for n > c(1) d(3) log p.
引用
收藏
页码:4117 / 4134
页数:18
相关论文
共 33 条
[1]  
Ahmedy A., 2008, CMUML08118
[2]  
Alon N., 2015, PROBABILISTIC METHOD
[3]  
[Anonymous], 2007, COMPLEX SOCIAL NETWO
[4]  
Banerjee O, 2008, J MACH LEARN RES, V9, P485
[5]  
Baxter R., 2007, Exactly Solved Models in Statistical Mechanics
[6]  
BESAG J, 1986, J R STAT SOC B, V48, P259
[7]  
Bresler G., 2008, RANDOM, P343
[8]  
Brown L. D., 1986, IMS Lecture Notes Monograph Series
[9]  
Chickering David Maxwell, 1996, Learning from data: Artificial intelligence and statistics V, P121
[10]   APPROXIMATING DISCRETE PROBABILITY DISTRIBUTIONS WITH DEPENDENCE TREES [J].
CHOW, CK ;
LIU, CN .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1968, 14 (03) :462-+