Baptiste Abélès

Hi, I am a PhD student in Learning Theory at Universitat Pompeu Fabra in Barcelona. I am interested in Online Learning Theory and its connection with Statistical Learning Theory in particular in the setting of dependent data.

More specifically, I have been focusing on the use of Online Learning theoretical results to provide better statistical learning guarantees. My research interests also include: Bandits, PAC-Bayes, Graphs Theory, and more broadly learning theory.

I am supervised by Gergely Neu. You can reach out to me at: baptistabeles[at] gmail [dot] com.

CV  /  Google Scholar  / 

profile photo
News
  • 06/2025 - I have been accepted to MLSS Senegal, see you in M'bour!
  • 05/2025 - Presentation of our paper "Online-to-PAC generalization bounds under graph-mixing dependencies" at AISTATS 2025 in Phuket, Thailand.
  • 04/2025 - I will give a talk about online-to-PAC for Non-i.i.d. data at INRIA Paris.
  • 02/2025 - Presentation of our paper "Generalization bounds for mixing processes via delayed online-to-PAC conversions" at ALT 2025 in Milan, Italy.
  • show more
Publications & Preprints

2025

Linear Bandits with Non-i.i.d. Noise
Baptiste Abélès, Eugenio Clerico, Hamish Flynn, Gergely Neu
preprint
arxiv

Online-to-PAC generalization bounds under graph-mixing dependencies Baptiste Abélès, Eugenio Clerico, Gergely Neu
(AISTATS 2025) 28th International Conference on Artifical Intelligence and Statistics.
arxiv

Generalization bounds for mixing processes via delayed online-to-PAC conversions Baptiste Abélès, Eugenio Clerico, Gergely Neu
(ALT 2025) 36th International Conference on Algorithmic Learning Theory.
arxiv

Talks

Generalization bounds for mixing processes via delayed online-to-PAC conversions
04/2024 - INRIA. Paris, France.


Thanks Jon Barron for this nice template.