Mean-Field Control for Restless Bandits and Weakly Coupled MDPs

Explore a technical lecture where Dr. Nicolas Gast from Inria discusses mean-field control approaches for addressing the curse of dimensionality in reinforcement learning, particularly focusing on resource allocation problems with weakly coupled Markov Decision Processes (MDPs). Learn about LP-based relaxations including the Whittle index and their applications in scenarios where multiple entities' states evolve over time, coupled through controller actions. Discover how these solutions become asymptotically optimal as resource numbers approach infinity, drawing from Dr. Gast's extensive research experience in stochastic models and optimization methods for large-scale system control algorithms. Gain insights from his work at Inria Grenoble and academic background spanning Ecole Normale Superieure and University of Grenoble, enhanced by his visiting position at MIT.