Very large Asset and Liability Management (ALM) problems with millions of scenarios and decision variables are usually beyond the scope of general purpose QP and NLP solvers. On the other hand many specialised approaches have limitations on the type of models they can be applied to.
In this talk I will present a structure-exploiting parallel interior-point solver for quadratic and nonlinear programming problems. The solver allows the efficient exploitation of nested block structured constraint and Hessian matrices, such as appear in multistage stochastic programming models. Through its generic object-oriented design its applicability is not limited to stochastic programming: indeed it can be applied to virtually any problem displaying a nested block structure.
Numerical results are given for various QP and NLP reformulations of ALM problems with up to 50 million decision variables. Comparisons with CPLEX are given.
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