Marc Sebban

Marc Sebban is a Professor in Computer Science at the Hubert Curien Laboratory (LabHC) and Deputy Director of this research unit. He leads the Inria project-team MALICE, focusing on machine learning, domain adaptation, and metric learning.

Research Interests

• Metric learning with theoretical guarantees
• Domain adaptation via optimal transport
• Physics-informed neural networks
• Imbalanced data classification
• Tree-structured data similarity learning

Recent Publications

His 2025 work introduces provably accurate adaptive sampling for collocation points in PINNs and theoretically grounded quadrature methods using residual Hessians. 2024 publications explore physics-informed ML for laser-matter interaction, predictive modeling of body shape changes, and approximation error analysis in tanh neural networks. Earlier works address graph diffusion Wasserstein distances, metric learning for imbalanced data, and boosting algorithms with confidence oracles.