The paper proves global convergence of a distributed neural policy gradient algorithm for networked cooperative multi-agent RL, provided the true Q-functions and density ratios lie in a Barron-type function class.
Finite-time analysis of dis- tributed TD(0) with linear function approximation on multi-agent rein- forcement learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.MA 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Distributed Neural Policy Gradient Algorithm for Global Convergence of Networked Multi-Agent Reinforcement Learning
The paper proves global convergence of a distributed neural policy gradient algorithm for networked cooperative multi-agent RL, provided the true Q-functions and density ratios lie in a Barron-type function class.