{"id":"arxiv/2602.02250","url":"https://arxiv.org/abs/2602.02250","tags":["math.OC","cs.LG"],"score":0,"title":"Well-Posed KL-Regularized Control via Wasserstein and Kalman-Wasserstein KL Divergences","domain":"math.OC","source":"arxiv","authors":["Viktor Stein","Adwait Datar","Nihat Ay"],"reviews":0,"subject":"math.OC","summary":"Kullback-Leibler (KL) divergence regularization is widely used in reinforcement learning, but it becomes infinite under support mismatch and can degenerate in low-noise regimes. Using a unified information-geometric framework, we introduce KL analogs by replacing the Fisher-Rao geometry in the dynamical formulation of the KL with transport-based geometries, and derive closed-form expressions for common distribution families. Between elliptic distributions, these divergences remain finite for degenerating equal covariances and yield a geometric interpretation of regularization heuristics used…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-02-02","reactions":{}}