Automated configuration of general-purpose optimization algorithms via machine learning techniques

2017 Fiscal Year Final Research Report

• Project/Area Number: 26282085
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Partial Multi-year Fund
• Section: 一般
• Research Field: Social systems engineering/Safety system
• Research Institution: Osaka University
• Principal Investigator: Umetani Shunji 大阪大学, 情報科学研究科, 准教授 (80367820)
• Co-Investigator(Kenkyū-buntansha): 河原 吉伸 大阪大学, 産業科学研究所, 准教授 (00514796)
• Project Period (FY): 2014-04-01 – 2018-03-31
• Keywords: 組合せ最適化 / アルゴリズム

Outline of Final Research Achievements

For most non-expert users suffers, it is desirable to formulate their real world problems into general-purpose combinatorial optimization model such as mixed integer programming problem (MIP) and constraint satisfaction problems (CSP) and solve them by the state-of-the-art algorithms. However, it would be certainly difficult to attain comparable results to those of the specially tailored algorithms, because the hardness of extracting the useful features for improving the efficiency of algorithms from the general form of MIP and CSP.
In this research project, in order to achieve a general-purpose algorithm with highest performance, we developed automated system that configures an metaheuristic algorithm for the instance to be solved by combining a wide variety of components and tuning program parameters at runtime.

Free Research Field

組合せ最適化