End-to-End Bayesian Networks Exact Learning in Shared Memory

被引:0
作者
Karan, Subhadeep [1 ]
Sayed, Zainul Abideen [1 ]
Zola, Jaroslaw [1 ]
机构
[1] Univ Buffalo, Dept Comp Sci & Engn, Buffalo, NY 14203 USA
关键词
Bayes methods; Dynamic programming; Lattices; Search problems; Optimization; Task analysis; Directed acyclic graph; Bayesian networks; exact learning; task parallelism; PARALLEL ALGORITHM;
D O I
10.1109/TPDS.2024.3366471
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Bayesian networks are important Machine Learning models with many practical applications in, e.g., biomedicine and bioinformatics. The problem of Bayesian networks learning is NP-hard and computationally challenging. In this article, we propose practical parallel exact algorithms to learn Bayesian networks from data. Our approach uses shared-memory task parallelism to realize exploration of dynamic programming lattices emerging in Bayesian networks structure learning, and introduces several optimization techniques to constraint and partition the underlying search space. Through extensive experimental testing we show that the resulting method is highly scalable, and it can be used to efficiently learn large globally optimal networks.
引用
收藏
页码:634 / 645
页数:12
相关论文
共 31 条
  • [11] Parent assignment is hard for the MDL, AIC, and NML costs
    Koivisto, Mikko
    [J]. LEARNING THEORY, PROCEEDINGS, 2006, 4005 : 289 - 303
  • [12] Koller D., 2009, Probabilistic graphical models: principles and techniques
  • [13] Bayesian networks in biomedicine and health-care
    Lucas, PJF
    van der Gaag, LC
    Abu-Hanna, A
    [J]. ARTIFICIAL INTELLIGENCE IN MEDICINE, 2004, 30 (03) : 201 - 214
  • [14] A parallel algorithm for Bayesian network structure learning from large data sets
    Madsen, Anders L.
    Jensen, Frank
    Salmeron, Antonio
    Langseth, Helge
    Nielsen, Thomas D.
    [J]. KNOWLEDGE-BASED SYSTEMS, 2017, 117 : 46 - 55
  • [15] Concurrent Hash Tables: Fast and General(?)!
    Maier, Tobias
    Sanders, Peter
    Dementiev, Roman
    [J]. ACM TRANSACTIONS ON PARALLEL COMPUTING, 2019, 5 (04)
  • [16] Empirical hardness of finding optimal Bayesian network structures: algorithm selection and runtime prediction
    Malone, Brandon
    Kangas, Kustaa
    Jarvisalo, Matti
    Koivisto, Mikko
    Myllymaki, Petri
    [J]. MACHINE LEARNING, 2018, 107 (01) : 247 - 283
  • [17] Parallel globally optimal structure learning of Bayesian networks
    Nikolova, Olga
    Zola, Jaroslaw
    Aluru, Srinivas
    [J]. JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING, 2013, 73 (08) : 1039 - 1048
  • [18] Nikolova O, 2012, INT CONF HIGH PERFOR
  • [19] Ott S, 2003, PACIFIC SYMPOSIUM ON BIOCOMPUTING 2004, P557
  • [20] Pearl J., 2009, Causality: Models, reasoning, and inference, V2nd