A Closed-Form Upper Bound for Admissible Learning-Rate Steps in Belief-Space Dynamics
Machine Learning
2026-05-11 v1
Abstract
Learning-rate steps are usually treated as hyperparameters. This paper isolates a local beliefspace calculation: when an update is modeled as a projected forward step on the probability simplex, admissibility means contractivity in the natural KL/Bregman geometry. Under this model, the upper bound of an admissible step is not a tuning slogan but a formula.
Cite
@article{arxiv.2605.06741,
title = {A Closed-Form Upper Bound for Admissible Learning-Rate Steps in Belief-Space Dynamics},
author = {Zixi Li and Youzhen Li},
journal= {arXiv preprint arXiv:2605.06741},
year = {2026}
}