Post by Albrecht Schmidt

Reinforcement learning for interactive systems. Sven Mayer started with an overview of the advances in RL we saw over the last 10 years (including Atari games, AlphaGo, openAI rubrics cube). To get an idea of the challenges, he walked us through a simple pong example. Good discussions on reward predictions and the overall credit assignment problem in real world problems. #winterschool2026