A Generalized Limited-Memory BNS Method Based on the Block BFGS Update

0476437 - ÚI 2018 RIV CZ eng C - Konferenční příspěvek (zahraniční konf.)
Vlček, Jan - Lukšan, Ladislav
A Generalized Limited-Memory BNS Method Based on the Block BFGS Update.
Programs and algorithms of numerical mathematics 18. Prague: Institute of Mathematics CAS, 2017 - (Chleboun, J.; Kůs, P.; Přikryl, P.; Segeth, K.; Šístek, J.; Vejchodský, T.), s. 164-171. ISBN 978-80-85823-67-7.
[Programs and Algorithms of Numerical Mathematics /18./. Janov nad Nisou (CZ), 19.06.2016-24.06.2016]
Grant CEP: GA ČR GA13-06684S
Institucionální podpora: RVO:67985807
Klíčová slova: unconstrained minimization * block variable metric methods * limited-memory methods * the BFGS update * global convergence * numerical results
Obor OECD: Applied mathematics
http://dml.cz/handle/10338.dmlcz/703010

A block version of the BFGS variable metric update formula is investigated. It satisfies the quasi-Newton conditions with all used difference vectors and gives the best improvement of convergence in some sense for quadratic objective functions, but it does not guarantee that the direction vectors are descent for general functions. To overcome this difficulty and utilize the advantageous properties of the block BFGS update, a block version of the limited-memory BNS method for large scale unconstrained optimization is proposed. The algorithm is globally convergent for convex sufficiently smooth functions and our numerical experiments indicate its efficiency.
Trvalý link: http://hdl.handle.net/11104/0272937