{"id":"arxiv/2610.39239","url":"https://arxiv.org/abs/2610.39239","tags":["math.PR","cs.IT","math.OC","stat.CO"],"score":0,"title":"Bregman projections and partition design for accelerating finite Markov chains","domain":"math.PR","source":"arxiv","authors":["Michael C. H. Choi","Youjia Wang"],"reviews":0,"subject":"math.PR","summary":"We study Gibbs averaging of finite Markov chains with a prescribed stationary distribution $π$. Given a transition matrix $P$ and the Gibbs kernel $G$ induced by a state-space partition, we prove that $GPG$ is an information projection for every admissible Bregman generator $φ$, with divergences defined using transition probabilities relative to $π$. A Pythagorean identity characterizes this projection and recovers the relative entropy and weighted Frobenius cases. A dual characterization identifies $GPG$ as a kernel closest to stationarity among those with the same projection chain as $P$;…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-29","reactions":{}}