Moritz A. Zanger

Ph.D. Candidate at the TU Delft. Amateur Woodworker.

moritz.jpg

Me doing reinforcement learning in my spacious workshop.

My name is Moritz A. Zanger. I am a fourth year Ph.D. student under the supervision of Prof. Matthijs T.J. Spaan and Dr. Wendelin Böhmer at the Sequential Decision Making group at the TU Delft (Netherlands). Prior to this, I received a Master of Science (w. distinction) in Mechanical Engineering from the Karlsruhe Institute of Technology (Germany).

My research focuses on efficient uncertainty estimation in deep reinforcement learning, a topic I consider crucially important in making AI more trustworthy and reliable. I am, however, interesed in most things related to reinforcement learning and deep learning. For example, I am excited about recent algorithms in unsupervised reinforcement learning and generative applications of RL, like generative flow networks.

news

Sep 10, 2024 Contextual Similarity Distillation is on Arxiv!
Sep 10, 2024 One paper accepted at EWRL 2024 in Toulouse.
Apr 18, 2024 One paper accepted at ICLR 2024 in Vienna!
Feb 14, 2024 Our technical report on Epistemic AI passed the halftime review of EU Horizons.
Sep 03, 2023 One paper accepted at EWRL 2023 in Brussels.
Oct 27, 2021 Attended the ELLIS Doctoral Symposium in Tuebingen.
Sep 15, 2021 One paper accepted at IROS 2021.
Jun 01, 2021 Started a Ph.D. on Epistemic Uncertainty in RL with Matthijs Spaan at the TU Delft!

publications

  1. arXiv
    ntkgp.png
    Contextual Similarity Distillation: Ensemble Uncertainties with a Single Model
    Moritz A Zanger, Pascal R Vaart, Wendelin Böhmer, and 1 more author
    arXiv preprint arXiv:2503.11339, 2025
  2. arXiv
    ant.png
    Value Improved Actor Critic Algorithms
    Yaniv Oren, Moritz A Zanger, Pascal R Vaart, and 2 more authors
    arXiv preprint arXiv:2406.01423, 2024
  3. Diverse Projection Ensembles for Distributional Reinforcement Learning
    Moritz Akiya Zanger, Wendelin Boehmer, and Matthijs T. J. Spaan
    In The Twelfth International Conference on Learning Representations, 2024
  4. Safe continuous control with constrained model-based policy optimization
    Moritz A Zanger, Karam Daaboul, and J Marius Zöllner
    In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021