Éloi Campagne

Éloi Campagne

PhD student in Applied Mathematics — ENS Paris-Saclay

Research on forecasting electricity consumption using interpretable graph-based neural networks.

I am currently a PhD student at ENS Paris-Saclay, in the Centre Borelli's Learning and Information Processing Systems (LIPS) team, in collaboration with EDF Lab's consumption forecasting team, under the supervision of Argyris Kalogeratos, Mathilde Mougeot, Yvenn Amara-Ouali and Yannig Goude.

My research focuses on developing and testing new methods for forecasting electricity consumption and production to support power grid management. I work with graph-based neural networks, aiming to design temporal and dynamic graph models that remain interpretable, in the spirit of generalized additive models.

Before starting my PhD, I graduated from Télécom SudParis with a specialization in mathematics, and from the Master MVA program at ENS Paris-Saclay in 2023. In 2022, I spent a semester at the University of Twente in the Netherlands through the Erasmus program, where I deepened my interest in Machine Learning.

Education

PhD in Machine Learning
École Normale Supérieure Paris-Saclay
2023 – 2027 (expected) | Centre Borelli & EDF (CIFRE)

Graphical Models for Joint Forecasting of Short-Term Consumption and Renewable Production.
Supervisors: A. Kalogeratos, M. Mougeot & Y. Amara-Ouali.

MSc — MVA (Mathematics, Vision, Learning)
École Normale Supérieure Paris-Saclay
2022 – 2023

Time Series Learning, Optimization, Topological Data Analysis, Responsible ML. High honors.

Engineering Degree
Télécom SudParis
2019 – 2023 | Mathematics & Data Science

Probability, Optimization, Statistical Learning, Bayesian Estimation, Algorithms. GPA: 3.86/4.

Erasmus Exchange
University of Twente (NL)
Spring 2022

Deep Learning, Differential Privacy, Reinforcement Learning, Data-Driven Modeling.

Work Experience

Research Intern in Machine Learning
EDF & Centre Borelli
2023

Developed Graph Neural Network methods for short-term electricity demand forecasting using regional and individual data. Thesis available here.
Supervisors: Y. Amara-Ouali, Q. Chan-Wai-Nam & A. Kalogeratos.

Data Science Intern
Okwind
2022

Optimized solar production prediction algorithms using NeuralProphet, NHiTS, DeepAR, and TFT.
Supervisor: T. Riou.

Software Development Intern
Aerometrik
2021

Designed Arduino programs for particle counter interaction.
Supervisor: D. Rouault.

Research

Go check out my latest publications in the field of Graph Machine Learning applied to Electricity Load Forecasting!

My personal touch

A more personal page that gathers my analog photographs and a non-exhaustive list of books I recommend.