Remarks on input to state stability on open sets, motivated by perturbed gradient flows in model-free learning
Eduardo D. Sontag · arXiv (Cornell University) · 2021
This note presents an expository discussion, with poofs, of input to state stability for systems that evolve in open subsets of Euclidean space. The results are then applied to the study of noisy gradient systems, which are of interest in data-driven model-free machine learning.