Aymen Al Marjani

Aymen Al Marjani

PhD Student

UMPA (ENS Lyon)

Scool (Inria)

Biography

I am a third-year PhD student at UMPA (ENS Lyon) and the Inria Scool team (formerly SequeL), supervised by Aurélien Garivier and Emilie Kaufmann. Prior to my PhD, I graduated from ENS Paris-Saclay and École Polytechnique.

My main research interests lie in Reinforcement Learning and Multi-Armed Bandits. In particular, I have worked on Pure Exploration problems, where the goal is to learn a “good” decision using few rounds of interaction with the environment. I focus on designing theoretically-grounded algorithms with good empirical performance.

I am also interested in online learning, information theory, differential privacy and machine learning in general.

Interests
  • Machine Learning
  • Reinforcement Learning
  • Multi-armed Bandits
  • Online Learning
Education
  • PhD in Reinforcement Learning, 2021-Now

    ENS Lyon & Inria Lille

  • Msc in Machine Learning, 2019-2020

    ENS Paris-Saclay

  • Diplôme d'Ingénieur (MSc) in Applied Mathematics, 2016-2019

    École Polytechnique

Publications

(2023). Active Coverage for PAC Reinforcement Learning. In COLT 2023.

PDF Slides Arxiv

(2023). Optimistic PAC Reinforcement Learning: the Instance-Dependent View. In ALT 2023.

Cite PDF Arxiv

(2022). Near Instance-Optimal PAC Reinforcement Learning for Deterministic MDPs. In NeurIPS 2022.

Cite PDF Slides Arxiv

(2022). On the complexity of All ε-Best Arms Identification. In ECMLPKDD 2022.

Cite PDF Arxiv

(2021). Navigating to the Best Policy in Markov Decision Processes. In NeurIPS 2021.

Cite PDF Slides Video Arxiv

(2020). Adaptive Sampling for Best Policy Identification in Markov Decision Processes. In ICML 2021.

Cite PDF Slides Arxiv

Experience

 
 
 
 
 
KTH Royal Institute of Technology
Research Intern
Apr 2020 – Sep 2020 Stockholm, Sweden
Sample Complexity of Best Policy Identification(PAC RL) with access to a generative model.
 
 
 
 
 
TotalEnergies
Data Analyst Intern
Apr 2019 – Aug 2019 Paris, France
Performed Sobol Analysis and constructed a surrogate model for Black-Box refinery simulators.
 
 
 
 
 
Philsmile
Summer Intern
Philsmile
Jun 2018 – Aug 2018 Manilla, Philippines
Built a plug-and-play Facial Recognition module for automated classroom-attendance monitoring.

Contact

  • aymen (dot) al (underscore) marjani (at) ens-lyon.fr