A seminar of potential interest will take place in the Department of Mathematics this week:
Thursday, Mar 27, 2025 2:00 PM – Aula Riunioni (Department of Mathematics)
Recent Progress in Steplengths in Gradient Methods for Unconstrained Optimization – Michiel Hochstenbach (Eindhoven University of Technology)
Abstract: We review recent progress in stepsizes in gradient methods for unconstrained numerical optimization, for both quadratic and general nonlinear problems, including several very valuable contributions from Italian experts such as Prof. Zanni and colleagues. We will review Barzilai-Borwein steps and secant equations, and show developments in harmonic stepsizes and homogeneous secant equations. The key idea is that the steplength is selected as the inverse Ritz value of an average Hessian. For limited memory methods, we will address several challenges, including how to avoid negative or non-real stepsize candidates. We will also present some open problems. This talk reflects joint work with Giulia Ferrandi.