第37回工学院特別セミナー「最適化法講演」

ドイツZuse Institute Berlinにおいて数理最適化ソフトウェア開発に携わるお二人による講演会を開催します。是非ふるってご参加ください。

2024年1月9日（火）15:00 - 17:00

東京工業大学 大岡山西8号館（E） 10階大会議室

本学学生・教職員、学外一般

工学院

情報・システム研究機構統計数理研究所、九州大学マス・フォア・インダストリ研究所、Zuse Institute Berlin、株式会社NTTデータ数理システム

• 15:00-16:00

  講演 1：Global Optimization of Mixed-Integer Nonlinear Programs with SCIP8
  講演者：Dr.Stefan Vigerske（Zuse Institute Berlin／GAMS）

  SCIP is a branch-and-price framework for the solution of Constraint Integer Programs, which combines the generality of constraint programming with the strength of linear（or more general）relaxations. Next to a focus on mixed-integer linear programming, SCIP has been extended since 2009 to handle first quadratic and then general nonlinear constraints（given as expression tree）. With version 8, these capabilities have been largely reworked and extended. This talk is a reminiscence of this work and gives an overview on the state of MINLP solving in SCIP 8.

• 16:00-17:00

  講演 2：Solving challenging instances of discrete optimization problems in parallel using more than 100,000 cores
  講演者：品野勇治博士（Zuse Institute Berlin）

  We have been developing the Ubiquity Generator framework (UG) which is a software framework designed to parallelize state-of-the-art solvers on large-scale computing environments, for more than 10 years. The parallel solvers instantiated by UG have successfully solved more than 20 previously unsolved instances from MIPLIB, 5 instances from SteinLib, and 3 instances from QAPLIB. In this talk, we will present the latest status of UG and its parallel solvers.

参加申込不要