Short bio

• 2022. I founded Viking Conseil to extend my PhD work and build adaptive forecasting models that are able to adapt to an increasingly unstable world. The objective of Viking Conseil is to continue research in order to apply our solution in the industry.

• 2019 - 2022. I obtained my PhD in Statistics under an industrial contract (CIFRE). I worked jointly at the LPSM, Sorbonne Université under the supervision of Olivier Wintenberger and at EDF R&D under the supervision of Yannig Goude and Thi Thu Huong Hoang. I defended on June 22, 2022, my PhD manuscript is available here, and my defense slides here.
  I received the male thesis prize of Think Smartgrids, as well as the Marie-Jeanne Laurent-Duhamel thesis prize of the Société Française de Statistiques.

• 2015 - 2019. I graduated from the École Normale Supérieure and got a Master Degree from the École Normale Supérieure Paris-Saclay in mathematics and machine learning (MVA).

Research

I am mostly interested in time series forecasting and the application I consider in particular is electricity load forecasting. I consider adaptive methods that are able to learn regime changes in the time series, focusing in particular on state-space representations. Indeed, state-space models capture very well the evolution of the electricity demand; in particular, they perform well in the context of abrupt regime changes.

Publications

• J. de Vilmarest and N. Werge (2024). An adaptive volatility method for probabilistic forecasting and its application to the M6 financial forecasting competition. International Journal of Forecasting.
• J. de Vilmarest and O. Wintenberger (2024). Viking: Variational Bayesian Variational Tracking. Statistical Inference for Stochastic Processes.
• J. de Vilmarest, J. Browell, M. Fasiolo, Y. Goude and O. Wintenberger (2023). Adaptive Probabilistic Forecasting of Electricity (Net-)Load. IEEE Transactions on Power Systems.
• G. Lambert, B. Hamrouche and J. de Vilmarest (2023). Frugal day-ahead forecasting of multiple local electricity loads by aggregating adaptive models. Nature Scientific Reports.
• J. de Vilmarest and Y. Goude (2022). State-Space Models for Online Post-Covid Electricity Load Forecasting Competition. IEEE Open Access Journal of Power and Energy.
  Implementation on gitlab.
• J. de Vilmarest and O. Wintenberger (2021). Stochastic online optimization using kalman recursion. Journal of Machine Learning Research.
• D. Obst, J. de Vilmarest and Y. Goude (2021). Adaptive Methods for Short-Term Electricity Load Forecasting During COVID-19 Lockdown in France. IEEE Transactions on Power Systems.

Preprints

• B. Abélès, J. de Vilmarest and O. Wintenberger (2024). Adaptive time series forecasting with markovian variance switching.
• J. de Vilmarest and O. Wintenberger (2019). Logarithmic Regret for parameter-free Online Logistic Regression.

Competitions

• February 2022 - February 2023 We participated with Nicklas Werge in the M6 Financial Forecasting Competition under the team name AdaGaussMC. We focus on the Forecasts leaderboard, ranking third in the first quarter, and second globally as of July 7. See our preprint and implementation for details.
• June 2021. I obtained the first place in Competition on building energy consumption forecasting.
• March - April 2021. We obtained with Y. Goude the first place in the competition Day-Ahead Electricity Demand Forecasting: Post-COVID Paradigm.
  See our publication and implementation for details.

Communications

• May 27-31, 2024. Journées de Statistiques (Bordeaux, France).
• September 11-14, 2023. WPI-Workshop: Stochastics, Statistics, Machine Learning and their Applications to Sustainable Finance and Energy Markets (Vienna, Austria).
• January 16, 2023. Machine Learning Meetup (Nantes, France).
• September 27-28, 2022. International Ruhr Energy Conference (Essen, Germany).
• August 30, 2022. Journées MAS.
• June 22 2022. PhD Defense, Jussieu, Paris.
• June 13 2022. Journées de Statistiques, Lyon.
• April 5 2022. Rencontre des Jeunes Statisticiens, Porquerolles.
• January 20 2022. Séminaire Images Optimisation et Probabilités at Bordeaux.
• January 13 2022. Séminaire DATA at the Laboratoire Jean Kuntzmann, Grenoble.
• December 1 2021. PGMO days.
• October 4 2021. Assessment of ensemble forecasts. See the slides.
• July 27 2021. Panel session Performance evaluation of artificial intelligence methods for energy consumption forecasting using open data sets at the IEEE Power & Energy Society General Meeting. See the slides.
• July 24 2021. Poster presentation at the time-series workshop of ICML. See the article, poster and short video.
• June 7 2021. Oral presentation at the Journées de Statistiques. See the slides.
• September 15 2020. Oral presentation at New Results on Time Series and their Statistical Applications, Marseille, France.