We are opening several reinforcement learning position at Inria Scool. Please, take also a look at this page. https://team.inria.fr/scool/job-offers.
- One 18m postdoc position about recommender systems in the context of the Pl@ntNet Inria research project. The focus is on estimation/improvement of annotation expert users by means of contextual multi-armed bandits approaches, with a unique opportunity to consider real data. Please apply here. Preferred starting date is early spring 2024.
Please contact O-A. Maillard by email with 3 of your main publications, CV, motivation letter and recommendation letter. The positions can be filled as soon as possible.
Currently no funding available. Last position:
Within the Chaire of Artificial Intelligence on Reinforcement Learning (AppRenf project, R-PILOTE-19-004-APPRENF), we are opening a fully-funded PhD position starting next fall 2022. The topic is about Real-life Challenges for Reinforcement Learning. The proposal is part of the Fondation I-SITE ULNE within the project PILOTE from cluster HumAIn@Lille. This is a PhD in Machine Learning, more specifically in Reinforcement Learning. See this document for further details. To apply, please contact O-A. Maillard by email with CV, motivation letter.
Interns (summer 2023)
Below is a list of (funded) internships proposals, please contact me in case you are interested. We expect students with a solid mathematical background specifically in statistics, information theory and/or dynamical systems. I only sketch topics below for obvious reasons, please contact me for details.
- Risk in sequential decision making, within the Three-Risk-Proof Sequential Decision Making (3R-SDM) project, linked to Inrae-Inria call on Environmental Risks (filled)
- Bandit tools for MDP theory: In particular we want to revisit LSPI, MCTS, VI adapting modern optimal bandit strategies (filled)
- Revisiting Regression Trees, Random Forest and variable selection with multi-armed bandits for finite-time error bound handling.
- Use-case: Analysis of on-farm sequential data for decision making and recommender systems. This is a unique opportunity to analyse state-action-reward trajectories coming from real agriculture experiments.
- Farm-gym v2: We are continuing the development of the atari of farming, fully-oriented towards RL.
These are intended for Master 2 or outstanding Master 1 students, and generally open the possibility to start a PhD later on. If you are interested, go ahead and contact me directly.
In case you want to apply for PhD, I strongly encourage you to read (a substantial part of) the following books and lecture notes:
- Prediction Learning ang Games
Nicolo Cesa-Bianchi, and Gábor Lugosi. Cambridge University Press, 2006.
- Concentration inequalities: A nonasymptotic theory of independence
Stéphane Boucheron, Gábor Lugosi, and Pascal Massart. OUP Oxford, 2013.
- Self-normalized processes: Limit theory and Statistical Applications
Victor H. Peña, Tze Leung Lai, and Qi-Man Shao. Springer Science & Business Media, 2008.
- Pac-Bayesian supervised classification: The thermodynamics of statistical learning
Catoni, Olivier. IMS, 2007.
- Bandits algorithms
Tor Lattimore, Csaba Szepesvári.
- Algorithms for Reinforcement Learning
Csaba Szepesvári. Synthesis Lectures on Artificial Intelligence and Machine Learning 4.1 (2010): 1-103.
- Markov Decision Processes: Discrete Stochastic Dynamic Programming
- Mathematics of Statistical Sequentiel Decision Making.
Odalric-ambrym Maillard (my Habilitation dissertation).
- Statistical Learning Theory and Sequential Prediction
Alexander Rakhlin, Karthik Sridharan
- Concentration of Measure Inequalities in Information Theory, Communications and Coding,
Raginsky, Maxim, and Igal Sason. Now Publishers Inc., 2014.
- Course on Reinforcement Learning