Francisco Sobral (State University of Maringá, Brazil)

Quasi-Newton and the unreduced matrix in interior point methods
Monday 26 March 2018 at 14.00, JCMB 6201

Abstract

In this work we study and implement a quasi-Newton approximation to the unreduced matrix that arises in Interior Point Methods. Given the good structure and huge size of the unreduced matrix, the most common approaches reduce it to Augmented Systems or to Normal equations in order to compute the Newton directions. We show that a quasi-Newton approximation to the inverse of the unreduced matrix has the ability to decrease the overall number of expensive matrix factorizations is some linear programming problems. We also discuss interesting properties of this approach.

Seminars by year

Current 2019 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996