The field of federated learning and reinforcement learning is moving towards more robust and scalable methods. Researchers are exploring ways to improve the performance of federated learning algorithms in heterogeneous environments, such as through the use of novel global objective functions and decentralized architectures. Additionally, there is a growing interest in applying reinforcement learning to real-world problems, including network security and edge video analytics. Notable papers in this area are: Federated Reinforcement Learning in Heterogeneous Environments, which proposes a robust FRL-EH framework and achieves superior performance over existing state-of-the-art FRL algorithms. FCPO: Federated Continual Policy Optimization for Real-Time High-Throughput Edge Video Analytics, which combines Continual RL with Federated RL to achieve significant improvements in effective throughput, reduced latency, and faster convergence.
Advancements in Federated Learning and Reinforcement Learning
Sources
Federated Reinforcement Learning in Heterogeneous Environments
Adaptive Network Security Policies via Belief Aggregation and Rollout
Compatibility of Max and Sum Objectives for Committee Selection and $k$-Facility Location
Decentralized Federated Learning of Probabilistic Generative Classifiers
An FDM-sFEM scheme on time-space manifolds and its superconvergence analysis
HOTA: Hamiltonian framework for Optimal Transport Advection
Federated Majorize-Minimization: Beyond Parameter Aggregation
Enhancing Quantum Federated Learning with Fisher Information-Based Optimization
FCPO: Federated Continual Policy Optimization for Real-Time High-Throughput Edge Video Analytics