PDR-ANPG achieves last-iterate ε-optimality gap and ε constraint violation in CMDPs with sample complexity Õ(ε^{-2} min{ε^{-2}, ε_bias^{-1/3}}) for parameterized policies with transferred compatibility error ε_bias.
Improved samp le complexity analysis of natural pol- icy gradient algorithm with general parameterization for i nfinite horizon discounted reward Markov decision processes
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Last-Iterate Convergence of General Parameterized Policies in Constrained MDPs
PDR-ANPG achieves last-iterate ε-optimality gap and ε constraint violation in CMDPs with sample complexity Õ(ε^{-2} min{ε^{-2}, ε_bias^{-1/3}}) for parameterized policies with transferred compatibility error ε_bias.