Thibaut Vidal

IVADO Digital Futures 2025 | IVADO

Propulsing vehicle routing into space... At IVADO's Digital Futures Event, my brilliant postdoc and collaborator Théo Guyard will show how #Optimization & #MachineLearning meet orbital dynamics to clean up space debris, in collaboration with NASA Ames Research Center.🚀
ivado.ca/en/events/iv...

Work done at the SCALE-AI Chair at @polymtl.bsky.social with my fabulous co-authors Arthur Ferraz, Quentin Cappart, Axel Parmentier, Alexandre Forel, and Cheikh Ahmed... Happy #ORMS, #StrategicOptimization, and #MachineLearning, everyone!

GitHub - vidalt/Districting-Routing: Source code associated with the paper "Deep Learning for Data-Driven Districting-and-Routing", authored by A. Ferraz, Q. Cappart, and T. Vidal Source code associated with the paper "Deep Learning for Data-Driven Districting-and-Routing", authored by A. Ferraz, Q. Cappart, and T. Vidal - vidalt/Districting-Routing

Related links:
🔗Paper 1: arxiv.org/pdf/2402.06040
🔗Paper 2 (NeurIPS 2024): arxiv.org/pdf/2412.08287
🔗Source code: github.com/vidalt/Distr...

Optimizing Supply Chains with GNN

How can GNNs be harnessed for an efficient strategic optimization of delivery districts using ML+OR pipelines and end-to-end learning? Check Data Skeptic's latest podcast discussing two of our recent works on this topic (my intervention starts around 5:00): open.spotify.com/episode/3GAP...

Stay tuned as we regularly share new discoveries on trustworthy #MachineLearning in connection with #GraphTheory, #ORMS, and Combinatorial Optimization. All the related papers are openly accessible, as well as the source codes:
github.com/vidalt
Thanks for following us! 🙌

This study was funded by SCALE-AI Canada through its "Research Chairs" program and Polytechnique Montréal (@polymtl.bsky.social). It has been a privilege to collaborate with the brilliant Julien Ferry, Ricardo Fukasawa, and Timothée Pascal on this!

Votez pour votre découverte préférée! Notre jury a sélectionné les 10 découvertes québécoises les plus impressionnantes de la dernière année. À votre tour de choisir la découverte qui vous surprend ou vous inspire le plus.

Until February 14, 2025, you can vote for your favorite discovery on the list! If you would like to support our project, "L’intelligence artificielle : toujours confidentielle?", cast your vote here:
www.quebecscience.qc.ca/decouverte20...

Our work stood out for its critical focus on AI safety, with the jury emphasizing: "The rapid development of AI sometimes comes at the expense of public safety. Highlighting the risks of data non-confidentiality is a crucial step in establishing ethical guidelines."

L’intelligence artificielle : toujours confidentielle ? - Québec Science Les modèles d’intelligence artificielle peuvent à l’occasion se montrer trop bavards ! En les étudiant, on peut parfois reconstituer des données confidentielles qui ont servi à leur entraînement.

Hot off the press! Our research on the risks of training-data reconstruction from random forest models* has just been nominated on Quebec Science's (@quebecscience.bsky.social) Top 10 Scientific Discoveries of 2024! 🌟 🚀
www.quebecscience.qc.ca/sciences/les...

Alice Gorgé just completed her research internship at the SCALE AI Chair at Polytechnique Montréal on the privacy risks of ML models. She has just received the prestigious Louis-Edouard Rivot medal, an honor given annually to Polytechnique Paris (X) students who excel in their research work! 😎

GitHub - alexforel/AdaptiveCC: Code for paper on "Adaptive Partitioning for Chance-Constrained Problems" Code for paper on "Adaptive Partitioning for Chance-Constrained Problems" - alexforel/AdaptiveCC

All the source code and material to reproduce the experiments is available under an MIT license at github.com/alexforel/Ad.... Many thanks to my fabulous coauthors as well as SCALE-AI and @polymtl.bsky.social for the research support. Happy optimization, everyone... 😎

In a nutshell, the method works by iteratively refining and merging scenario sets to obtain tightened bounds. Convergence is guaranteed over a finite number of iterations, and we experimentally measure very significant speed-ups over direct solution approaches.

Interested in solving chance-constrained optimization problems at scale? Buckle-up, the most recent work of
Marius Roland and Alexandre Forel (optimization-online.org?p=25061) is now in the press at SIAM JOpt... Congratulations on this excellent work! 🚀 #ORMS #Stochastic #Optimization

Last week was my first time in a radio studio, joining @matthieudugal.bsky.social & Moteur de Recherche on Radio Canada. What an experience! The vibe was fantastic & there's something so exciting about chatting with the columnists and sharing cool science facts🎙️🤩 ici.radio-canada.ca/ohdio/premie...

Hello, world!
Hummm... hello, blue sky? 😏

A bit late to this party... but can you add me? ;)