Walter A. Kosters

About

Walter A. Kosters is a leading researcher in artificial intelligence, with a primary focus on **reinforcement learning** and its application to high-dimensional, complex decision-making problems. His major contributions center on advancing **model-based reinforcement learning (MBRL)** , a paradigm that aims to dramatically improve sample efficiency compared to traditional model-free methods. Kosters’ work is particularly notable for addressing the critical challenge of sample complexity in deep RL, which has historically limited the deployment of these algorithms in real-world robotics and game-playing scenarios. His most cited paper, "High-accuracy model-based reinforcement learning, a survey" (2023), has already garnered over 40 citations, underscoring its impact as a key reference for researchers seeking to balance accuracy with computational feasibility. Through comprehensive surveys published in 2020 and 2021, Kosters has systematically mapped the landscape of MBRL for high-dimensional problems, providing essential taxonomies and benchmarks that guide both new and experienced practitioners. His research is instrumental in bridging the gap between theoretical RL advances and practical, sample-efficient applications, making him a pivotal voice in the ongoing evolution of deep reinforcement learning.

Research Focus

Key Achievements

Top Papers

1. 1
   High-accuracy model-based reinforcement learning, a survey

   40 citations · 2023
2. 2
   Model-Based Deep Reinforcement Learning for High-Dimensional Problems, a Survey.

   12 citations · 2020
3. 3
   Deep Model-Based Reinforcement Learning for High-Dimensional Problems, a Survey

   7 citations · 2020
4. 4
   High-Accuracy Model-Based Reinforcement Learning, a Survey

   5 citations · 2021

Key Collaborators

Contact & Links