A contrastive-divergence-based reward stabilizes reinforcement learning of position-dependent step sizes for gradient-based MCMC, beating constant-step tuning on most of 44 benchmark posteriors.
Drawing on the reward-centring framework of Naik et al
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.CO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Harnessing the Power of Reinforcement Learning for Adaptive MCMC
A contrastive-divergence-based reward stabilizes reinforcement learning of position-dependent step sizes for gradient-based MCMC, beating constant-step tuning on most of 44 benchmark posteriors.