Inexact Restoration Methods for Finite Sum Minimization

Abstract: Convex and nonconvex finite-sum minimization arises in many scientific computing and machine learning applications. Recently, first-order and second-order methods where objective functions, gradients and Hessians are approximated by randomly sampling components of the sum have received great attention. We discuss a class of methods which employs suitable approximations of the objective function, gradient and Hessian built via random subsampling techniques. The choice of the sample size is deterministic and ruled by the Inexact Restoration approach. Local and global properties for finding approximate first- and second-order optimal points and function evaluation complexity results are discussed in the framework of line search and trust region methods.

Mathematics

Audience: researchers in the topic

Women in Mathematics in South-Eastern Europe

Series comments: The webinar will be held in Zoom. A direct link to the virtual hall will appear on icms.bg/ on the day of the webinar.