math.OCNov 18, 2026

AG-BFGS: Foundations for Accelerated Geometry Learning

Authors: Regina Carla Sousa, Denis Vargas, Elizabeth Wanner, Joshua Knowles

Organizations: Dept of Mathematics, UFOP, Brazil · PPGMMC, CEFET-MG, Brazil · CSDT School, Aston University - PPGMMC, CEFET-MG · SLB Cambridge Research, High Cross, Madingley Road, Cambridgeshire, UK

Abstract

We introduce an accelerated geometry-learning framework for BFGS that incorporates momentum into the learning of Hessian inverse approximations. We prove that the resulting accelerated updates satisfy explicit stability conditions and preserve positive definiteness within admissible parameter regions. Numerical experiments demonstrate consistent improvements over the corresponding non-accelerated schemes while maintaining numerical stability. These results provide the first rigorous theoretical foundation for accelerated quasi-Newton learning and bridge the gap between empirical observations and mathematical guarantees for accelerated second-order methods.

Figures & tables

Explore similar work

CardsList
  1. Primal Acceleration of Newton's Method

    Aug 21, 2026Nikita DoikovSmooth Convex OptimizationSecond-Order Optimization