<ul><li>Federated reinforcement learning (FRL) allows multiple agents to collaborate and learn a shared policy in different environments.</li><li>The actor-critic (AC) algorithm is known for its low variance and high sample efficiency in RL.</li><li>However, theoretical understanding of AC in a federated manner with different environments is limited.</li><li>The Single-loop Federated Actor Critic (SFAC) algorithm is proposed, showing convergence to a near-stationary point and linear speed-up in sample complexity.</li></ul>

Single-Loop Federated Actor-Critic across Heterogeneous Environments

Discover more