{"id":"arxiv/2610.40608","url":"https://arxiv.org/abs/2610.40608","tags":["math.OC"],"score":0,"title":"Improved Gradient Complexity for Quasi-Newton Proximal Extragradient Methods via Certified Curvature Reuse","domain":"math.OC","source":"arxiv","authors":["Aryan Mokhtari","Francisco Patitucci"],"reviews":0,"subject":"math.OC","summary":"We study how curvature estimated from gradient differences can accelerate proximal-extragradient quasi-Newton methods. Existing global guarantees for QNPE, A-QNPE, and OL-OQN learn Hessian approximations online and incur dimension dependence through regret bounds. We introduce certified curvature reuse: one finite-difference probe per iteration refreshes a group of Hessian columns, while a path-length certificate controls the error of stale columns and determines the step. Each iteration uses at most three gradient queries and one regularized linear solve (or trust-region step), without a…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-29","reactions":{}}