Felix Leibfried

Papers

Total Citations

H-Index

About

Felix Leibfried is a researcher at the intersection of reinforcement learning, information theory, and robotics, whose work explores how agents can learn more efficiently by balancing reward maximization with intrinsic motivation. His key research areas include information-theoretic approaches to perception-action coupling, empowerment-driven learning, and mutual-information regularization in Markov decision processes. Leibfried’s most cited paper, "An information-theoretic on-line update principle for perception-action coupling" (2017, 10 citations), proposes a principled framework inspired by sensorimotor coupling in biological systems, offering a novel way to integrate action and perception in robotic agents. In "A Unified Bellman Optimality Principle Combining Reward Maximization and Empowerment" (2019, 8 citations), he introduces a unified framework that combines traditional reward-based reinforcement learning with empowerment—an information-theoretic measure of an agent’s control over its environment—enabling more exploratory and robust behavior. His work on "Mutual-Information Regularization in Markov Decision Processes and Actor-Critic Learning" (2019, 3 citations) further advances the field by showing how entropy regularization can improve exploration and policy robustness. Leibfried’s contributions are notable for bridging theoretical insights from information theory with practical algorithmic advances, offering a foundation for more autonomous and adaptive learning systems.