Event info

Date: Friday, 16 January 2026 16:00 - 17:00

Place: Seminar Room (W1-D-710) and Live streaming with Zoom

Speaker: Yuji Shinano (Zuse Institute Berlin, Germany)

Title: Challenges to solve previously unsolved instances by using a large-scale parallel MIP solver

Abstract:
Mixed-integer programming (MIP) solvers have achieved dramatic performance improvements over the past thirty years and can be regarded as the most commercially successful class of solvers based on mathematical optimization. Because mixed-integer programming problems are NP-hard, they are theoretically difficult to solve exactly. To address real-world problems, the solver-development community has established the benchmark dataset MIPLIB, which models practical problem instances, and has continuously advanced solver technology capable of handling these instances. Although all state-of-the-art commercial solvers now support parallel computation, only a very limited number implement the latest algorithms in a truly large-scale parallel environment. The Ubiquity Generator framework (UG), developed by the speaker, is a software framework designed to realize large-scale parallel solvers built on top of modern state-of -the-art MIP solvers. Parallel solvers based on UG have already been the first to obtain optimal solutions for more than twenty benchmark instances.
This talk reports on the current state of large-scale parallel mixed-integer programming solvers developed using UG.

IMI Colloquium Organizers
IKEMATSU, Yasuhiko
KURATA, Sumito
TAGAMI, Daisuke