HIGH-DIMENSIONAL CONSISTENCY IN SCORE-BASED AND HYBRID STRUCTURE LEARNING

被引:65
作者
Nandy, Preetam [1 ]
Hauser, Alain [2 ]
Maathuis, Marloes H. [1 ]
机构
[1] Swiss Fed Inst Technol, Seminar Stat, Ramistr 101, CH-8092 Zurich, Switzerland
[2] Bern Univ Appl Sci, Dept Engn & Informat Technol, Jlcoweg 1, CH-3400 Burgdorf, Switzerland
基金
瑞士国家科学基金会;
关键词
Bayesian network; directed acyclic graph (DAG); linear structural equation model (linear SEM); structure learning; greedy equivalence search (GES); score-based method; hybrid method; high-dimensional data; consistency; MARKOV EQUIVALENCE CLASSES; MODEL SELECTION; COVARIANCE ESTIMATION; MAXIMUM-LIKELIHOOD; DISTRIBUTIONS; ALGORITHM; PACKAGE; GRAPHS;
D O I
10.1214/17-AOS1654
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
Main approaches for learning Bayesian networks can be classified as constraint-based, score-based or hybrid methods. Although high-dimensional consistency results are available for constraint-based methods like the PC algorithm, such results have not been proved for score-based or hybrid methods, and most of the hybrid methods have not even shown to be consistent in the classical setting where the number of variables remains fixed and the sample size tends to infinity. In this paper, we show that consistency of hybrid methods based on greedy equivalence search (GES) can be achieved in the classical setting with adaptive restrictions on the search space that depend on the current state of the algorithm. Moreover, we prove consistency of GES and adaptively restricted GES (ARGES) in several sparse high-dimensional settings. ARGES scales well to sparse graphs with thousands of variables and our simulation study indicates that both GES and ARGES generally outperform the PC algorithm.
引用
收藏
页码:3151 / 3183
页数:33
相关论文
共 59 条
[1]   Scaling up the Greedy Equivalence Search algorithm by constraining the search space of equivalence classes [J].
Alonso-Barba, Juan I. ;
delaOssa, Luis ;
Gamez, Jose A. ;
Puerta, Jose M. .
INTERNATIONAL JOURNAL OF APPROXIMATE REASONING, 2013, 54 (04) :429-451
[2]  
Anandkumar A, 2012, J MACH LEARN RES, V13, P2293
[3]  
Andersson SA, 1997, ANN STAT, V25, P505
[4]  
[Anonymous], 2009, PROBABILISTIC GRAPHI
[5]  
[Anonymous], 2009, ELEMENTS STAT LEARNI, DOI DOI 10.1007/978-0-387-84858-7
[6]  
Banerjee O, 2008, J MACH LEARN RES, V9, P485
[7]   A Constrained l1 Minimization Approach to Sparse Precision Matrix Estimation [J].
Cai, Tony ;
Liu, Weidong ;
Luo, Xi .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 2011, 106 (494) :594-607
[8]   Extended Bayesian information criteria for model selection with large model spaces [J].
Chen, Jiahua ;
Chen, Zehua .
BIOMETRIKA, 2008, 95 (03) :759-771
[9]  
Chickering D. M., 2003, Journal of Machine Learning Research, V3, P507, DOI 10.1162/153244303321897717
[10]  
CHICKERING D.M., 2015, UAI 2015