The paper claims the first asymptotically optimal regret algorithm, achieving the exact logarithmic constant K(M), for average-reward communicating Markov decision processes.
2 2.1.1 Randomized policies, their gain, bias & gap functions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Asymptotically optimal regret in communicating Markov decision processes
The paper claims the first asymptotically optimal regret algorithm, achieving the exact logarithmic constant K(M), for average-reward communicating Markov decision processes.