Heterogeneous federated RL agents can share knowledge through periodic distillation of action distributions toward a global average, though the paper's theoretical guarantees are not sound.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
FedHPD: Heterogeneous Federated Reinforcement Learning via Policy Distillation
Heterogeneous federated RL agents can share knowledge through periodic distillation of action distributions toward a global average, though the paper's theoretical guarantees are not sound.