Jonas Rothfuss

ETH Zurich, Karlsruhe Institute of Technology

Papers

Total Citations

H-Index

About

Jonas Rothfuss is a researcher at the forefront of safe and data-efficient machine learning, with a primary focus on Bayesian optimization, meta-reinforcement learning, and robotic control. His work addresses the critical challenge of enabling robots to learn and adapt quickly while operating under strict safety constraints. Rothfuss’s major contributions include the development of PACOH-RL, a model-based meta-reinforcement learning algorithm that meta-learns priors for dynamics models, allowing for swift adaptation to new environments with minimal interaction data. This work has garnered significant attention, with his most-cited paper on meta-learning priors for safe Bayesian optimization accumulating 7 citations. Additionally, his research on deep episodic memory for robot action execution, which encodes, recalls, and predicts experiences, represents a novel approach to enabling robots to leverage past experiences for more efficient learning. Rothfuss’s work is particularly notable for its practical implications in robotics, where the ability to safely and efficiently optimize controller parameters is paramount. His contributions are shaping the future of autonomous systems that can learn from limited data while maintaining safety guarantees.