Hector Kohler

About

Hector Kohler is an Assistant Professor (tenure track) at Télécom Paris. He is funded by Hi! Paris to research Reinforcement Learning. He is interested in solving hard (computer) science problems with Reinforcement Learning. Such hard problems include optimization on graphs, and communications in multi-agent networks, for which the DIG team of LTCI is internationally recognized.

Hector Kohler is also developing methodologies to comprehensively evaluate new Reinforcement Learning algorithms and to promote open and reproducible science. Before joining IP Paris, he was a Postdoctoral Research Fellow at École Polytechnique and a visiting scholar at the RLAI lab from the University of Alberta in Canada. Hector Kohler completed his Ph.D. in the Scool team from Inria Lille.

Publications Full list on Scholar →

  1. EWRL 2026

    Limits of reinforcement learning for decision trees in Markov decision processes

    Timothee Mathieu, Juliette Achddou, Alex Davey, Hector Kohler, Philippe Preux, Julien Teigny
    European Workshop on Reinforcement Learning, Lille

    PDF
  2. RLJ 2026

    Limits of reinforcement learning for decision trees in Markov decision processes

    Hector Kohler, Riad Akrour, Philippe Preux
    Reinforcement Learning Journal

    PDF
  3. KDD 2025

    Breiman meets Bellman: Non-Greedy Decision Trees with MDPs

    Hector Kohler, Riad Akrour, Philippe Preux
    ACM SIGKDD Conference on Knowledge Discovery and Data Mining

    PDF Code
  4. PRL 2025

    Evaluating Interpretable Reinforcement Learning by Distilling Policies into Programs

    Hector Kohler, Quentin Delfosse, Waris Radji, Riad Akrour, Philippe Preux
    Workshop on Programmatic Reinforcement Learning, Edmonton

    arXiv Code
  5. EWRL 2024

    Interpretable and Editable Programmatic Tree Policies for Reinforcement Learning

    Hector Kohler, Quentin Delfosse, Riad Akrour, Kristian Kersting, Philippe Preux
    European Workshop on Reinforcement Learning, Toulouse

    arXiv Code
  6. TMLR 2024

    AdaStop: adaptive statistical testing for sound comparisons of Deep RL agents

    Timothée Mathieu, Matheus Medeiros Centa, Riccardo Della Vecchia, Hector Kohler, Alena Shilova, Odalric-Ambrym Maillard, Philippe Preux
    Transactions on Machine Learning Research

    PDF Code
  7. IFAC 2022

    PID Tuning using Cross-Entropy Deep Learning: a Lyapunov Stability Analysis

    Hector Kohler, Benoit Clement, Thomas Chaffre, Gilles Le Chenadec
    IFAC-PapersOnLine

    PDF

Open Source

  • rlberry

    Contributor to this library for reinforcement learning research: deep RL agents and robust statistical analysis of training runs.

    Docs GitHub

  • DPDTreeEstimator

    A scikit-learn estimator for DPDT, non-greedy decision tree induction with dynamic programming.

    Code

  • interpretable-rl-zoo

    Fully transparent control policies for games and robots, following the Gymnasium API.

    Code

  • AdaStop

    A statistical test for reliable and cheap comparisons of reinforcement learning agents.

    Code

Teaching assistant

Course notebooks

Supervisions

Service