{"id":"arxiv/2606.23198","url":"https://arxiv.org/abs/2606.23198","tags":["math.OC","cs.LG","stat.ML"],"score":0,"title":"Incremental Learning in Mirror Flows","domain":"math.OC","source":"arxiv","authors":["Raphaël Berthier","Loucas Pillaud-Vivien"],"reviews":0,"subject":"math.OC","summary":"We study mirror flows generated by a convex quadratic loss and a general convex lower semicontinuous mirror potential. We show that, when initialized near the boundary of the domain of the mirror potential, their rescaled trajectories converge to a limiting mirror flow whose potential is the indicator function of the domain. In this limit, the primal variable minimizes the loss over a time-dependent hypothesis set: the subdifferential of the support function of the domain, evaluated at the dual variable. This characterization provides a general mechanism for incremental learning in mirror…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-22","reactions":{}}