Tom Demeulemeester

Postdoctoral researcher at the Department of Economics, Faculty of Business and Economics (HEC), University of Lausanne (Switzerland), where I am fortunate to be supervised by Bettina Klaus.

My main research interests are operations research, combinatorial optimization, matching/assignment, computational social choice, and algorithmic game theory.

Email: tom (dot) demeulemeester (at) unil (dot) ch
Please feel free to contact me for questions about my research, or if you want to collaborate.

Paper in the spotlight

Can you truly make fair decisions using optimization techniques?

In Fair integer programming under dichotomous and cardinal preferences, we propose a unified framework to fairly select optimal solutions of integer programming formulations. As such, we embed the "black-box" procedure of solving an integer program into a framework that is explainable from start to finish.

We propose several general-purpose algorithms to maximize, for example, Nash welfare or the minimum selection probability, and we study their axiomatic and computational properties.

More details can be found in the paper.

News

Aug '24: The European Journal of Operational Research accepted our paper Fair integer programming under dichotomous and cardinal preferences (with Dries Goossens, Ben Hermans & Roel Leus).
Jun '24: Excited to announce that I will start a postdoc under the supervision of Bettina Klaus at the University of Lausanne from September.
May '24: I will hold the public defense of my PhD, entitled Fairness through randomization: An operations research perspective, on 22nd of May, 2024.
Aug '23: Our paper Relaxed core stability in hedonic games with size-dependent utilities (with Jannik Peter) was accepted at MFCS 2023.
Dec '22: A paper I co-authored got accepted at AAAI-23: Strategy-proofness and proportionality in party-approval multi-winner elections (with Théo Delemazure, Manuel Eberl, Jonas Israel & Patrick Lederer).
Apr '22: I presented our paper A pessimist’s approach to one-sided matching at the 2022 Easter Workshop on School Choice in Belfast.
Apr '22: The European Journal of Operational Research accepted our paper A pessimist’s approach to one-sided matching (with Dries Goossens, Ben Hermans & Roel Leus).

Other interesting results

Imagine a setting with an equal number of agents and objects. If the agents have random preferences over the objects, what is the worst-ranked object to which we can expect to assign everyone?

We show that the optimal expected worst rank grows at a logarithmic rate and is upper bounded by the expression in the figure. For a random market with 1,000 agents and objects, for example, this means that we can guarantee to assign everyone to one of their first 9 choices, in expectation.

More details can be found in the paper, where we also describe how traditional mechanisms in this setting perform poorly on this measure.

Page updated

Google Sites

Report abuse