Implementation of trust region methods in optimization. (c1998)

LAUR Repository

Show simple item record

dc.contributor.author Hajj, Mohammed Omar
dc.date.accessioned 2010-11-26T13:56:54Z
dc.date.available 2010-11-26T13:56:54Z
dc.date.copyright 1998 en_US
dc.date.issued 2010-11-26
dc.date.submitted 1998-05
dc.identifier.uri http://hdl.handle.net/10725/143
dc.description Includes bibliographical references (l. 37). en_US
dc.description.abstract This project presents a new approach to Quasi-Newton methods for unconstrained optimization. Quasi-Newton Methods update at each iteration the existing Hessian approximation (or its inverse) cheaply by integrating data derived from the previously completed one, which is soon ignored. These methods are based on the so-called Secant equation. In our project we focus on solving a critical subproblem of the Quasi-Newton algorithm that requires determining a proper, suitable step size that takes from the current approximation to the minimum to a new 'better' one. The subproblem can either be posed as doing a Line Search along some generated search direction in order to determine a minimum along the search vector. Another technique, on which we focus primarily in this work, is to use a Trust Region method that directly computes the step vector without doing a focused Line Search. The subproblem is critical to the numerical success of Q-N methods. We emphasize features of successful implementation to pinpoint assess merits of Trust Region methods. Our Numerical Results reveal that Trust Region algorithms seem to markedly improve as the dimension of the problem increases, while for small dimensional problems performance of both methods is comparable. en_US
dc.language.iso en en_US
dc.subject Mathematical optimization en_US
dc.title Implementation of trust region methods in optimization. (c1998) en_US
dc.type Thesis en_US
dc.term.submitted Spring en_US
dc.author.degree MS in Computer Science en_US
dc.author.school Arts and Sciences en_US
dc.author.commembers May Abboud
dc.author.woa RA en_US
dc.description.physdesc 1 bound copy: vii, 44 leaves; ill.; 30 cm. available at RNL. en_US
dc.author.division Computer Science en_US
dc.author.advisor Issam Moghrabi
dc.identifier.doi https://doi.org/10.26756/th.1998.1 en_US
dc.publisher.institution Lebanese American University en_US
dc.author.affiliation Lebanese American University en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search LAUR

Advanced Search


My Account