ISyE Seminar - November 2025
(07-11-2025) Another look into our monthly ISyE seminars! For November's internal ISyE seminar in Zwijnaarde, we heard from our new researcher Alisson and special guest Gwen Maudet, postdoctoral researcher at the University of Luxembourg.
Our monthly ISyE seminars bring together researchers to share insights on industrial systems and engineering. At our November ISyE seminar in Zwijnaarde, our new researcher Alisson Constantine-Macías kicked off, followed by our special guest Dr. Gwen Maudet.
Our new researcher Alisson Constantine-Macías kicked off with a warm introduction titled “Who am I”, sharing insights into her background and research journey.
We were honoured to host Dr. Gwen Maudet, postdoctoral researcher at the University of Luxembourg, who presented his research on "A Distance between Mixed Integer Programming Instances: on the Road to Generalizability of Machine Learning Solving Strategies".
This talk presents my postdoctoral research on using machine learning (ML) to enhance Mixed Integer Programming (MIP) solvers. We focus on defining distances between MIP instances, enabling to measure heterogeneity of evaluation sets and guide the selection of ML based solving strategies.
You can learn more on about Dr. Maudet's work on https://gwenmaudet.github.io.
Get involved on LinkedIn.
Picture copyright: Gwen Maudet, gwenmaudet.github.io
Our new researcher Alisson Constantine-Macías kicked off with a warm introduction titled “Who am I”, sharing insights into her background and research journey.
We were honoured to host Dr. Gwen Maudet, postdoctoral researcher at the University of Luxembourg, who presented his research on "A Distance between Mixed Integer Programming Instances: on the Road to Generalizability of Machine Learning Solving Strategies".
This talk presents my postdoctoral research on using machine learning (ML) to enhance Mixed Integer Programming (MIP) solvers. We focus on defining distances between MIP instances, enabling to measure heterogeneity of evaluation sets and guide the selection of ML based solving strategies.
A Distance between Mixed Integer Programming Instances: on the Road to Generalizability of Machine Learning Solving Strategies
This work, published at ECAI 2025 (https://arxiv.org/abs/2507.1106), introduces a distance metric for mixed-integer linear programming (MILP) instances, derived directly from their mathematical formulations and inspired by the Earth Mover’s Distance. The metric enables users to quantify the heterogeneity of instance sets and to guide solver behavior, particularly in machine learning–enhanced solvers. Evaluated on the StrIPLIB dataset, our method outperforms all non-learned baselines and matches the performance of a classifier trained on 760 labeled instances.You can learn more on about Dr. Maudet's work on https://gwenmaudet.github.io.
Get involved on LinkedIn.
Picture copyright: Gwen Maudet, gwenmaudet.github.io