Foundations of AI

Bild: Emilie Orgler

Prof. Dr. Leif Döring

Welcome to the research group where math meets magic. We’re a team of mathematicians obsessed with the stochastic heart of modern AI. Our playground is reinforcement learning, from principled decision‑making agents to the clever tweaks that make large models work. We chase clean theorems and hope for applied surprises: rigorous insight that brings algorithmic improvements and measurable wins. If you like beautiful proofs, daring experiments, and smarter systems, you’re in the right place.

This website will give you more information on

our research projects on the mathematical foundations, mostly reinforcement learning,
our lectures the foundations of AI and probability theory
events we organise to bring together researcher working in the interface of mathematics and AI

Team

For individual information on the members of the chair please check their profiles. If you are interested in writing a Bachelor's or Master's thesis supervised by a member of the team please contact them directly.

Research

Please check the research page for an overview of the research interests in the group. Typically, we study theoretical questions for stochastic processes but also study related real-world problems.

Teaching

The chair offers courses in the interplay of Analysis and pure and applied Probability Theory. Please check the course websites for information on current and future lectures and also for material of running lectures.

News

Target update frequency paper in ICML 2026

Our paper “The Role of Target Update Frequencies in Q-Learning” was accepted at ICML! The paper, based on the Masters thesis of Tilman Aach, rigorously analyses the target network trick of DeepMind's famous DQN paper. We show in the tabular setting, how decreasing update frequencies can improve the ...

New PhD student

A new PhD student joint our group. Daniel Schmidt finished his Master's degree in Mathematics at TU Darmstadt after pursuing his Bachelors at University of Mainz. He will work on Reinforcement Learning projects, both on theoretical and practical questions. Welcome!

Two papers accepted for NeuriPS 2025

Our group has two papers at NeurIPS 2025. The first is the final paper of Sara Klein's PhD thesis, where we study a policy gradient algorithm with receding horizon. The second is a paper with Sebastian Kassing from Wuppertal, where we study SGD with vanishing Tikhonov regularisation and show that ...

Vacancy: Assistant professor in applied probability

We are seeking for applications on a non-tenure track assistant professor position in the area of applied probability.