Two PhD projects started in late 2025 at LITIS, INSA Rouen Normandie, exploring different aspects of graph machine learning and its applications.

🔬 Molecular Reactivity Prediction with Machine Learning

PhD student: Louisa Bouzidi, co supervised with Laurent Joubert from CARMEN, and Luc Brun from GREYC.

This PhD focuses on predicting key molecular reactivity parameters (such as electrophilicity and nucleophilicity) using machine learning. The work aims to combine geometric graph neural networks with quantum chemical features to better capture the underlying physico-chemical properties of molecules, while taking into account the dynamic aspect of molecular compounds.

📈 Towards More Expressive Dynamic Graph Neural Networks

PhD student: Noé Gille, co supervised with Sébastien Adam and Clément Chatelain (LITIS)

This PhD aims at improving the expressiveness of dynamic graph neural networks, and more generally improve the state of the art of dynamic graph machine learning, both from a theoretical and modeling perspective.

It explores new architectures capable of capturing structural and temporal patterns in evolving graphs, with applications in areas such as chemistry and sports analytics.

These projects contribute to ongoing research activities at LITIS on graph neural networks, representation learning, and applied machine learning.

More updates on results and publications will follow.