{"id":"arxiv/2608.22834","url":"https://arxiv.org/abs/2608.22834","tags":["math.OC","cs.LG","math.NA"],"score":0,"title":"Mirror descent algorithms with logarithmic barriers","domain":"math.OC","source":"arxiv","authors":["Alberto De Marchi","Yura Malitsky","Adrien B. Taylor"],"reviews":0,"subject":"math.OC","summary":"This work derives convergence guarantees for mirror descent and proximal mirror descent algorithms when a logarithmic barrier is used as a distance-generating function. Standard approaches cannot be applied when the solution lies on the boundary, where the Bregman divergence blows up. We show that, in a specific setting, both methods enjoy an $O(\\log k / k)$ rate, which is also tight. In addition, our contributions include: (i) a new technique for handling the blow-up; (ii) a resolution of a gap in the theory of relative smoothness; and (iii) a comparison of the proposed approach with…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-24","reactions":{}}