Jules Depersin

About me

I’m an Assistant Professor at the University of Amsterdam in the Quantitative Economics section. I received my PhD in Statistics from Center for research in economics and statistics (CREST), Paris in December 2021: here are the manuscript and the slides. I was supervised by Guillaume Lecué.

I have a strong background in high-dimensional geometry and robust machine learning. Having earned a Ph.D. in this field, my expertise lies in navigating the intricacies of statistical and computational complexities in high-dimensional tasks.

Over the years, I have transitioned from primarily theoretical work to exploring more concrete applications in the realm of economics. I have for instance studied the performance of standard estimators in non-IID settings with two-way fixed effects for trade data, and I have done work on measuring and understanding the role of search frictions in international trade. In these works I try to understand how to characterize the inherent "complexity" of a problem or a model.

With an interest in both theoretical and applied research, I continue to explore the intersection of Machine Learning, statistics, computation, and economics.

Work in progress

• Depersin Jules, Patault Bérengère (2023). Revisiting the effect of search frictions on market concentration.

Publications

• Depersin Jules, Lecué Guillaume (2022). Robust subgaussian estimation of a mean vector in nearly linear time, Annals of Statistics.

• Depersin Jules, Lecué Guillaume (2021). Optimal robust mean and location estimation via convex programs with respect to any pseudo-norms, Probability theory and related fields.

• Depersin Jules, Lecué Guillaume (2021). On the robustness to adversarial corruption and to heavy-tailed data of the Stahel-Donoho median of means, Information and Inference: a Journal of the IMA.

• Depersin Jules (2020). A spectral algorithm for robust regression with subgaussian rates, Submitted.

• Depersin Jules (2020). Robust subgaussian estimation with VC-dimension, Annales de l’Institut Henri Poincaré.

• Depersin Jules, Barthelemy Marc (2019). From global scaling to the dynamics of individual cities, Proceedings of the National Academy of Sciences.

• Depersin Jules, Gaubert Stéphane, Joswig Michael (2017). A tropical isoperimetric inequality. Proceedings of the 29th Conference on Formal Power Series and Algebraic Combinatorics.

Events

• 03/2023: Talk on Revisiting the Effect of Search Frictions on Market Concentration for the Econometrics Seminars and Workshop Series of the Tilbergen Institute.

• 08/2021: Talk on Robustness of the Stahel-Donoho estimator at Journées MAS.

• 02/2021: Talk on Using VC-dimension in robust estimation at Conference on robustness and privacy.

• 02/2020: Talk on Robust and Fast Estimation for Heavy-Tailed Distributions at Stat-ML-Eco Seminar at CREST.

• 12/2019: Talk on Robust and fast estimation of a mean vector for heavy-tailed distributions at Eurandom Heavy Tails Workshop.

Teaching

I've had and will have the pleasure of assisting in teaching several courses, here are some.

2022-2023

• Programming and numerical analysis - University of Amsterdam
• Reinforcement Learning - University of Amsterdam with Yi He

2021-2022

• Théorie des processus (Martingales and Markov Chains) - ENSAE Paris with Nicolas Chopin
• Advanced machine learning - ENSAE Paris with Vianney Perchet
• Theoretical Foundations of Machine Learning - ENSAE Paris with Vianney Perchet

2020-2021

• Théorie des processus (Martingales and Markov Chains) - ENSAE Paris with Nicolas Chopin
• Advanced machine learning - ENSAE Paris with Vianney Perchet
• Theoretical Foundations of Machine Learning - ENSAE Paris with Vianney Perchet

2019-2020

• Théorie des processus (Martingales and Markov Chains) - ENSAE Paris with Nicolas Chopin
• Advanced machine learning - ENSAE Paris with Vianney Perchet

2018-2019

• Théorie des processus (Martingales and Markov Chains) - ENSAE Paris with Pierre Alquier
• Probability Theory - ENSAE Paris with Sandie Souchet
• Numerical analysis - ENSAE Paris :
  • Download the first Jupiter Notebook
  • Download the second Jupiter Notebook

In 2019, 2020 and 2021, I've entirely held (lectures+ tutorials) a crash course on basic statistics, differential equations, optimization and Machine Learning for students joining ENSAE with a background in economics. Download the lecture notes or see the page of the course.

I supervised some students' Applied Statistics Projects with Nicolas Schreuder.

Education

• Ph.D. in Applied Mathematics, ENSAE and University Paris Saclay [2018-2021]
• M.Sc. in Statistics & Data analysis, ENSAE, Paris [2017]
• Engeneering degree, Ecole Polytechnique [2016]

Coordinations of courses