FedKL: Tackling Data Heterogeneity in Federated Reinforcement Learning by Penalizing KL Divergence

Computer Science

• Reinforcement Learning 100%
• Data Heterogeneity 100%
• Leibler Divergence 100%
• Federated Learning 50%
• Learning System 25%
• Objective Function 25%
• Training Process 25%
• Heterogeneous Data 25%
• Total Variation 25%
• Slow Convergence 25%

Mathematics

• Kullback-Leibler Divergence 100%
• Objective Function 50%
• Step Size 50%
• Necessary Condition 50%
• Training Process 50%
• Constrains 50%