Modified Limited memory BFGS with displacement aggregation
From MaRDI portal
Publication:6423293
arXiv2301.05447MaRDI QIDQ6423293FDOQ6423293
Authors: Manish Kumar Sahu, Suvendu Ranjan Pattanaik 
Publication date: 13 January 2023
Abstract: A displacement aggregation strategy is applied in modified limited memory Broyden Fletcher Goldfarb Shanno's (M-LBFGS) algorithm to solve large-scale unconstrained optimization problems. A displacement aggregation helps to store less memory while updating a new iterate to approximate the inverse Hessian matrix as it discards the generating linear dependence vectors. It has been observed that the M-LBFGS scheme achieves the same theoretical convergence properties as the full memory scheme or the limited memory BFGS. Also, Numerical results show that displacement aggregation in adaptive M-LBFGS outperforms M-LBFGS.
Nonlinear programming (90C30) Numerical methods based on nonlinear programming (49M37) Methods of quasi-Newton type (90C53)
This page was built for publication: Modified Limited memory BFGS with displacement aggregation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6423293)
