Michel Minoux, professeur émérite au LIP6 (Sorbonne Université) et chercheur de renommée internationale en recherche opérationnelle, nous a quittés en septembre 2023.
Ancien élève de l'École Polytechnique et de Télécom Paris, il a mené l'essentiel de sa carrière scientifique au CNET (Centre National d'Études des Télécommunications), à l'Université Paris IX-Dauphine puis au LIP6, où il a dirigé l'axe de recherche « Optimisation des grands systèmes » au sein de l'équipe DECISION. Auteur de plus d'une centaine d'articles et de plusieurs ouvrages de référence, dont Programmation Mathématique : théorie et algorithmes et Graphes et algorithmes, il a marqué durablement notre discipline, tant par ses apports théoriques que par son attention constante aux applications industrielles.
La conférence lui rendra hommage à travers des témoignages et des présentations scientifiques dans la continuité de ses travaux.
Programme définitif avec les résumés de la conférence
Thursday, 17 September 2026
| 08:15–08:45 | Welcome breakfast and registration |
| 08:45–09:00 | Welcome Address Fabrice Kordon · Director of LIP6, Sorbonne Université A Few Words from the family Read by Riadh Zorgati on behalf of Madame Claudine Minoux Opening Remarks Patrice Perny · LIP6 Decision Team and GDR ROD |
| 09:00–09:45 | Dans le sillage de Michel Minoux : graphes, dioïdes et programmation linéaire Patrice Perny · LIP6, Sorbonne Université Résumé Dans cet exposé j'évoquerai mon premier contact avec l'univers de Michel Minoux grace au célèbre ouvrage ``Graphes et Algorithmes'', puis les développements en collaboration avec Michel Gondran sur les algèbres de chemins et les dioïdes et l'influence qu'ils ont eu sur quelques travaux réalisés au sein de l'équipe décision du LIP6. J'évoquerai ensuite deux collaborations avec lui au carrefour de la théorie de la décision et de l'optimisation, publiées dans Algorithmica et Discrete Applied Mathematics, l'une sur le problème de l'affectation OWA optimale, l'autre sur l'optimisation d'une intégrale de Choquet discrete. Je terminerai en mentionnant quelques suites données à ces travaux au sein de l'équipe plus récemment. |
| 09:45–10:30 | Tribute to Michel Minoux: An Overview of Probability Maximization for Gaussian Inequality Systems and Applications Riadh Zorgati · EDF R&D Abstract I had the honor and pleasure of working with Michel Minoux for over twenty years. An affable man of integrity, he was at once an effective scientific advisor attuned to the concerns of industry, an excellent educator, and a rigorous, passionate researcher. I wish to pay tribute to him here and—on behalf of all my colleagues at EDF R&D—thank him for his pivotal contributions to improving energy management and distribution grid management, as well as to the training of our engineers. I also want to thank him personally for his commitment to our joint research on probability maximization and its numerous applications (energy management, inverse problems, finance, and statistics). |
| 10:30–11:00 | Coffee break |
| 11:00–11:45 | Michel Minoux et l’aventure des graphes, des dioïdes et des mathématiques tropicales Michel Gondran · EDF R&D and AEIS Résumé Il est intéressant de se rappeler les différentes étapes de l’aventure intellectuelle qui ont permis à Michel d’inventer de nombreux concepts sur les cheminement dans un graphe, puis de construire les algèbres de dioïdes, et finalement de bâtir, à coté des mathématiques construites à partir de la trilogie de groupe-anneau-corps, un nouveau type de mathématiques, les mathématiques tropicales, bâties sur la trilogie de monoïdes canoniquement ordonnés- diodes-semicorps canoniquement ordonnés. Malgré la modestie de Michel, ces mathématiques nouvelles seront peut-être celles du 21ième siècle. |
| 11:45–12:15 | Testimonies Kamel Barkaoui · Cédric, CNAM |
| 12:15–13:45 | Lunch break |
| 13:45–14:30 | Zero-Sum Discounted Stochastic Games with Random Rewards Abdel Lisser · CentraleSupélec, Université Paris-Saclay Abstract We consider a two-person zero-sum discounted stochastic game with random rewards and known transition probabilities. The players have opposite objectives and are interested in optimizing the expected discounted reward which they can obtain with a given confidence level when both the players play the worst possible move against each other. We model such a game problem by defining the chance-constrained optimization problem for each player. In this framework, risk attitudes are determined by confidence levels, where values of 0.5 or below correspond to risk-seeking behavior and values above 0.5 correspond to risk-averse behavior. When the reward vector follows a multivariate elliptically symmetric distribution, the game is equivalent to a minimax formulation. We consider the game with risk-seeking and risk-averse players separately. We show that the risk-seeking problem is equivalent to a constrained optimization of a parameterized zero-sum stochastic game and the optimal payoff of player 1 and optimal cost of player 2 can be computed using Riemann gradient sampling algorithms. Later we use the solution of the constrained optimization problem of each player to compute its optimal strategy by solving a linear programming problem. We reformulate the risk-averse problem as a discrete minimax problem. We propose an algorithm based on a linearization method and discuss its convergence properties. Alternatively, we reformulate the risk-averse problem as a second-order cone programming problem with bilinear constraints. The numerical experiments on randomly generated instances are performed to illustrate our theoretical results. |
| 14:30–15:15 | On the Complementarity Between Large-Scale Optimization and Reinforcement Learning for Segment Routing in Optical Networks Brigitte Jaumard · Concordia University Abstract Mathematical programming and reinforcement learning serve different purposes in optimization. Mathematical programming is a traditional approach that uses mathematical equations to model and solve optimization problems. It is often used for problems with well-defined objectives and constraints. Reinforcement learning, on the other hand, is a more dynamic and adaptive approach that uses trial and error to learn and improve over time. It is particularly useful for problems with complex, non-linear objectives and where the environment is not fully known. In terms of how they can help each other, mathematical programming can provide the theoretical framework and constraints for reinforcement learning to operate within. Reinforcement learning can then use this framework to learn and improve decision-making processes, potentially leading to more efficient and effective solutions to optimization problems. In other words, the combination of these two approaches can lead to more robust and scalable optimization solutions that can handle the complexities of real-world problems. We will explore this new paradigm combining mathematical programming and reinforcement learning. We will illustrate it with segment routing in optical networks, which is now widespread thanks to a simplified control plane and increased scalability to meet the evolving traffic needs of 6G and beyond. |
| 15:15–15:45 | Coffee break |
| 15:45–16:30 | My Friendship and Scientific Collaboration with Michel Minoux since 1976 Nelson Maculan · Universidade Federal do Rio de Janeiro Abstract While attending the ISMP 1976 (International Symposium on Mathematical Programming, Budapest, August 1976)—following my presentation on the design optimization of a telecommunications network—our colleague Michel Gondran suggested I meet Michel Minoux in Paris. A week later, I was already meeting Michel in Paris. Since then, we have collaborated scientifically and maintained a close friendship. In my presentation, I will highlight the extensive scientific and academic collaboration between us. |
| 16:30–17:15 | Collaboration avec Michel Minoux: souvenirs d'un ami Celso Ribeiro · Universidade Federal Fluminense |
| 17:15–17:45 | Testimonies |
Friday, 18 September 2026
| 08:30–09:00 | Welcome breakfast and registration |
| 09:00–09:45 | On the Complexity of Some Problems in Cooperative Game Theory Michel Grabisch · Charles University and Université Paris 1 Panthéon-Sorbonne Abstract We study the computational complexity of fundamental algorithmic problems — membership testing, separation, valid-inequality testing, and linear optimization — over polytopes and cones arising from cooperative games (also known as pseudo-Boolean functions). A central obstacle in the study of such problems is that a general cooperative game on n players requires 2^n values, so the input size is 2^n for a game with n players, making these computational tasks theoretically trivial. Restricting to k-additive games reduces the input size to O(n^k ), making such games a natural target for meaningful questions about the existence of efficient algorithms. On the positive side, we give an explicit extended formulation of size O(n^k ) for the core of k-additive k-monotone games, allowing all four problems to be solved by a single polynomial-size linear program — in particular, circumventing the ellipsoid method that is needed when building from earlier tractability results of Deng and Papadimitriou, or of Edmonds. For the cone of k-additive (k−1)-monotone games, we give a complete characterization of its extreme rays and derive the same O(n^k ) bound on extension complexity, yielding a geometry-based proof and generalization of a result of Billionnet and Minoux. On the negative side, we show that for l ≤ k − 2 the cone of k-additive l-monotone games is computationally intractable: membership testing is not in NP (unless NP = coNP), valid-inequality testing is NP-complete, and extension complexity is at least 1.5^n. Our hardness reduction works by identifying, through a sequence of facial operations on the dual cone, a copy of the correlation polytope — a canonical hard 0/1 polytope for which the same problems are known to be intractable. Our hardness results yield, as a special case, a result of Crama and of Gallo and Simone. Furthermore, our hardness results also explain the lack of any good characterization of the extreme rays of the cone of k-additive (k−2)-monotone games. Together, the results draw a sharp algorithmic boundary within the k-additive family: tractability holds exactly when the monotonicity order is at least k−1, and hardness sets in at order k−2 and below. Joint work with Hans Raj Tiwary |
| 09:45–10:30 | Taking Uncertainty into Account: Collaborations with Michel Minoux Wim Van Ackooij · EDF R&D Abstract In this talk we will discuss two paths of research related to and done in collaboration with Michel Minoux. These have been united under the moniker of “taking uncertainty into account”. We will first discuss key insights obtained in a joint work on the differentiability of singular multivariate Gaussian distribution functions. A second part of the talk will be dedicated to discussing two stage robust optimization with applications in energy and how the state space representable uncertainty sets introduced by Prof. Minoux fit into this picture. |
| 10:30–11:00 | Coffee break |
| 11:00–11:45 | Multi-Stage Stochastic Optimization for Lot-Sizing with Uncertain Demand and Renewable Energy Supply Céline Gicquel · LISN, Université Paris-Saclay Abstract One way to achieve energy efficiency in manufacturing is to equip plants with on-site renewable energy generation systems to partially power industrial processes. However, renewable energy sources are highly intermittent and their availability is difficult to predict accurately. Therefore, we study an integrated industrial production and energy supply planning problem under uncertain renewable energy availability. We first propose a multi-stage stochastic programming approach to model this problem. We then investigate the development of a hybrid solution algorithm combining branch-and-cut with stochastic dual dynamic programming. Computational results obtained on randomly generated instances show the practical efficiency of the proposed approach. Joint work with R. Liao, F. Quezada and S. Kedad-Sidhoum |
| 11:45–12:15 | Testimonies |
| 12:15–13:45 | Lunch break |
| 13:45–14:30 | Lagrangian Duality in Combinatorics: 30 Years of Industrial Applications, Sometimes Fruitful Benoît Rottembourg · Inria Abstract Decomposition and relaxation methods have, for over half a century, enabled the solving of combinatorial optimization problems too large or complex to be tackled directly by MILP solvers or dedicated Branch & Bound techniques. Among these, Lagrangian duality stands out for its ability to provide both precise estimates of the sought discrete solution and embryonic solutions that can be repaired to satisfy the problem's constraints (sometimes referred to as the "dual ascent" technique). Similar to column generation—mathematically closely related—it partially exploits the underlying combinatorial structure of the problem and convexifies it to estimate the desired economic objective. Michel Minoux has been one of its most ardent advocates since the 1980s, and through his influence, he has passed on this passion to his students and readers. As a willing doctoral "victim" since 1989, I would like to use this talk to recount 30 years of Lagrangian decomposition in industry through five use cases in construction, media, tourism, banking, and the second-hand market. Beyond this bouquet of increasingly stochastic applications, I will attempt to convexify the anecdotes shared to outline broader trends, transcending my own experience as an engineer. |
| 14:30–15:00 | Coffee break |
| 15:00–15:45 | Graph Laplacian Matrix: Theory and Applications Arnaud Knippel · INSA Rouen Normandie Abstract The graph Laplacian matrix is defined as L = D - A, where D is the diagonal matrix of vertex degrees and A is the adjacency matrix. We illustrate its role in various applications. This matrix exhibits fundamental properties related to graph structure, notably eigenvector-based transformations. We focus on a specific class of eigenvectors with entries in {-1, 0, 1} and characterize graphs admitting them, both in the general case and for certain planar graphs like trees and cactus graphs. Finally, we prove that finding these eigenvectors is NP-hard and introduce an Integer Programming approach to enumerate them for the {-1,1} case. |
| 15:45–16:30 | From Nash Fairness to Bottleneck–Sum Optimization: Following a Path Opened by Michel Minoux Viet Hung Nguyen · LIMOS, Université Clermont Auvergne Abstract What began as a question about Nash proportional fairness leads to a path opened by Michel Minoux: how can apparently incompatible combinatorial criteria be optimized together? Minoux’s formalization of combined bottleneck–sum optimization shows how threshold restrictions turn this question into a sequence of tractable subproblems. I revisit this idea for upper and lower bottleneck criteria, connect it with the algorithms of Martello et al. and Duin–Volgenant, and extend it to the three-criterion two-sided bottleneck–sum problem. An assignment application and polyhedral insights illustrate the scope of this framework. |
| 16:30 | Closing of the conference |