Optimal transport (OT) is a powerful framework to define and compute distances between distributions (a.k.a. Wasserstein or earth mover’s distance), with a tractable computation thanks to the Sinkhorn algorithm [1]. Entropic regularization, which enables fast iterative scaling and GPU-friendly computation of the OT, providing the backbone of modern scalable OT pipelines.

Major challenges arise in exploring OT for domain adaptation on streaming data. While an online Sinkhorn algorithm was introduced in [2] to address the OT distances computation from sample streams, it does not operate in an online manner. The major issue is that the functions are evaluated on an increasing amount of newly available samples. which yields a memory complexity of O(n) and a time complexity of O(n^2). In order to fully operate in an online manner, the evaluations should not rely on all the previously available samples. Some attempts were provided to mitigate this major issue, such as by performing measure compression techniques (which are computationally expensive) [3] and by combining streaming quantile approximation with sliced OT [4].

This post-doc fellowship aims to provide theoretical foundations and algorithmic developments for OT on streaming data, mainly time series. For this purpose, the post-doc fellow will leverage earlier research results and take full advantage of the literature of adaptive signal processing and representation learning with deep learning.

This post-doc is an integral part of the global project OOD (Online Deep anomaly Detection), bringing together 4 PhD students and several permanent researchers from the Machine Learning group of the LITIS Lab.

[1] G. Peyré and M. Cuturi, “Computational optimal transport: With applications to data science,” Foundations and Trends® in Machine Learning, vol. 11, no. 5-6, pp. 355–607, 2019.
[2] A. Mensch and G. Peyré, “Online sinkhorn: Optimal transport distances from sample streams,” in NeurIPS, vol. 33, pp. 1657–1667, 2020.
[3] F. Wang, C. Poon and T. Shardlow, « Compressed online Sinkhorn,” arXiv preprint arXiv:2310.05019, 2023.
[4] K. Nguyen, « Streaming Sliced Optimal Transport, » arXiv preprint arXiv:2505.06835, 2025.

Duration : 18 months

Start of contract : when available

Host lab : LITIS Lab (Rouen metropolitan area)

Requirements:
PhD in applied mathematics, machine learning, advanced statistics, computer science or related. Strong background in advanced optimization and machine learning. Proficiency in Python.

If interested, please send CV in a motivational email to paul.honeine@univ-rouen.fr, gilles.gasso@insa-rouen.fr, maxime.berar@univ-rouen.fr and fannia.pacheco@univ-rouen.fr.

The research team « Apprentissage » of the LITIS laboratory is recruiting at INSA Rouen Normandy for a postdoctoral position in machine learning/deep learning for reaction optimization in organic synthesis.

The postdoctoral fellow will conduct high-impact research in machine learning and deep learning applied to the organic synthesis. It is expected development of theoretical, algorithmic contributions in machine learning for chemistry, in particular the modeling of organic reactions based on large amounts of experimental data. The intended research will target applications to fundamental problems in organic chemistry such as regioselectivity, optimization of the chemical yield of a reaction, minimization of by-products and finally prediction of diastereoisomeric ratios and/or enantiomeric ratios. First, classical machine learning models will be analyzed to identify the most relevant predictive features and to build robust models of reaction outcomes. The research will then move toward geometric deep learning, leveraging graph neural networks (GNNs) to represent reacting molecules as structured objects. GNNs naturally encode both the topology of molecular graphs and the chemical properties of atoms and bonds, providing a powerful framework for predicting reaction behavior and optimizing synthetic pathways. A major scientific challenge will lie in the integration of quantum-level information into these graph-based representations. Such information is not only difficult to incorporate, but also crucial, as quantum effects fundamentally govern reaction outcomes. By targeting advanced geometric deep learning enriched with quantum information, this research is expected to advance both the theoretical foundations of molecular modeling and the practical applications of data-driven organic synthesis.

The successful candidate will work at the LITIS Laboratory in collaboration with Institut CARMeN (UMR 6064), an internationally renowned laboratory specializing in methodology development in organic synthesis.

Application deadline: 2025-08-31
Apply here: https://jobs.inria.fr/public/classic/en/offres/2025-08984
Feel free to contact us if you have any questions: paul.viallard@inria.fr and romaric.gaudel@irisa.fr

Dans le cadre du projet ANR MeSMRise, une thèse est disponible à partir du 1/10/2025, à Grenoble (LJK). Le sujet porte sur la combinaison d’apprentissage profond auto-supervisé avec les théories sensori-motrices pour l’apprentissage de représentations de séquences d’interactions. Tous les détails peuvent se trouver ici: https://projet.liris.cnrs.fr/mesmrise/job/OfferPhDWP2.pdf

Recrutement ATER 2025 – Département ITI / Laboratoire LITIS (INSA Rouen Normandie)
L’INSA Rouen Normandie ouvre un poste d’Attaché Temporaire d’Enseignement et de Recherche (ATER), à pourvoir à partir du 1er septembre 2025, au sein du Département ITI et du laboratoire LITIS.

🔍 Profil enseignement
La personne recrutée interviendra principalement dans le domaine de l’Ingénierie des données, avec des enseignements en :
– Traitement du signal (notamment aléatoire)
– Apprentissage statistique et machine learning
– Analyse numérique, statistiques, théorie de l’information
– Encadrement de projets et de stages

🔬 Profil recherche
L’activité de recherche sera menée au LITIS, au sein d’une des équipes suivantes :
– Multi-agent, Interaction, Décision (MIND)
– Systèmes de Transport Intelligent (STI)
– Apprentissage (machine learning)

📍 Lieu : INSA Rouen Normandie, Saint-Étienne-du-Rouvray

📅 Prise de fonction : 01/09/2025

🌐 Candidature via Galaxie, Lien vers l’offre: https://www.insa-rouen.fr/insa-rouen-normandie/offres-demploi/recrutement-61-ater-002-iti-litis-1

📧 Contacts
– Enseignement : Géraldine Del Mondo (Directrice ITI) – geraldine.del_mondo@insa-rouen.fr
– Recherche : Laurent Vercouter (Co-directeur LITIS) – laurent.vercouter@insa-rouen.fr

L’offre complète est disponible à l’adresse : https://recrutement.isae-supaero.fr/fr/annonce/3576511-enseignant-chercheur-en-intelligence-artificielle-et-apprentissage-automatique-31400-toulouse

Le profil enseignement est centré sur l’apprentissage profond, l’optimisation pour l’apprentissage automatique et les architectures de calcul pour l’apprentissage automatique.
Le profil recherche est en apprentissage automatique et reste assez ouvert ; le projet de recherche devra s’intégrer au sein de l’équipe existante. Les candidat.e.s sont donc invités à se renseigner et prendre contact avec les membres de l’équipe. N’hésitez pas à contacter Thomas Oberlin (thomas.oberlin@isae-supaero.fr) et moi-même (emmanuel.rachelson@isae-supaero.fr) pour toute question ou pour venir rencontrer l’équipe.

Dans le cadre du projet MeSMRise (Multimodal deep SensoriMotor Representation learning), un stage de M2 est proposé à partir de février (date flexible) sur l’apprentissage de représentations prédictives d’objets dans le cadre des théories sensorimotrices. Une thèse sur la continuation de ce sujet sera ouverte en septembre 2025 (date flexible). Tous les détails peuvent se trouver ici: https://projet.liris.cnrs.fr/mesmrise/job/OfferInternWP2.pdf