Éloi Campagne

I am a PhD candidate at ENS Paris-Saclay, in the Learning and Information Processing Systems (LIPS) team at Centre Borelli, in collaboration with the demand forecasting team at EDF Lab. My PhD is supervised by Argyris Kalogeratos, Mathilde Mougeot, Yvenn Amara-Ouali, and Yannig Goude.

My research studies how explicit structure can improve forecasting when data and computational resources are limited. In the context of electricity demand and renewable-energy forecasting, I investigate three complementary scales: dependencies between time series, transfer between related learning tasks, and the organization of training examples. This work has led me to develop and evaluate graph-based forecasting models, curriculum-learning methods, and cascaded transfer strategies for many-task learning under a shared optimization budget. A recurring objective of my work is to identify the conditions under which these structures provide information beyond strong unstructured baselines. I combine statistical learning, time-series forecasting, graph methods, and optimization, with particular attention to interpretability, controlled comparisons, and the operational constraints of power-system forecasting.

Before starting my PhD, I graduated from Télécom SudParis with a specialization in mathematics, and completed the Mathematics, Vision, Learning (MVA) master's program at ENS Paris-Saclay in 2023. In 2022, I studied at the University of Twente through the Erasmus program.