Research on parallel algorithms for large scale Bayesian network structure search

2016 Fiscal Year Final Research Report

• Project/Area Number: 15K20965
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Computational science; Intelligent informatics
• Research Institution: The University of Tokyo
• Principal Investigator: Tamada Yoshinori 東京大学, 医科学研究所, 特任講師 (80435495)
• Project Period (FY): 2015-04-01 – 2017-03-31
• Keywords: ベイジアンネットワーク / グラフ探索アルゴリズム / 並列アルゴリズム

Outline of Final Research Achievements

We have successfully developed the novel parallel algorithm for searching for the optimal DAG structure using the supercomputer with 2D torus interconnect network. Using this algorithm we have succeeded in searching for the optimal DAG structure of the 36 variables and 200 samples using FX10 4800 computation nodes. This is the first example of searching for the optimal DAG structure with 36 variables in the literature. We have then improved this algorithm and succeeded in the optimal 37 variable DAG search using K computer 20768 computation nodes. Currently we are preparing the submission of the paper of this result.

Free Research Field

バイオインフォマティクス