Log-barrier regularization in online mirror descent attains the optimal Õ(t^{-1/4}) last-iterate convergence rate in uncoupled zero-sum matrix games under bandit feedback.
Last-iterate convergence of decentralized optimistic gradient descent/ascent in infinite-horizon competitive markov games http://proceedings.mlr.press/v134/wei21a.html
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Optimal last-iterate convergence in matrix games with bandit feedback using the log-barrier
Log-barrier regularization in online mirror descent attains the optimal Õ(t^{-1/4}) last-iterate convergence rate in uncoupled zero-sum matrix games under bandit feedback.