From JCunniff at tiaa-cref.org Fri Jan 27 10:34:02 2012 From: JCunniff at tiaa-cref.org (Cunniff, John M) Date: Fri, 27 Jan 2012 10:34:02 -0500 Subject: [SMI] Using COIN-OR SMI to solve a QP Message-ID: First, thank you for providing these optimization programs. I tried using SMI to solve a multi-stage stochastic QP mean-variance asset allocation problem (by adding a QUADOBJ section to the .core file), but the optimizer appears to ignore quadratic terms as these terms are not in the output deterministic equivalent mps file and do not appear to affect the optimal solution. We have a subscription to CPLEX here at TIAA-CREF, and it is able to read and solve the .core file as a single period QP, so I believe my formatting is correct. I read in Alan King's presentation to INFORMS in Seattle in Nov. 2007 that it "would be easy to add integer variables and quadratic objectives", so I am assuming that SMI cannot solve a QP at this time. 1. Is it true that the current version of SMI cannot solve a QP? 2. Which .cpp and .hpp files would need to be altered to expand SMI to handle a QP problem, and what would those alterations be? Thank you in advance your help. John Cunniff ************************************************************************* This e-mail may contain confidential or privileged information. If you are not the intended recipient, please notify the sender immediately and then delete it. TIAA-CREF ************************************************************************* -------------- next part -------------- An HTML attachment was scrubbed... URL: From kingaj at us.ibm.com Fri Jan 27 11:29:44 2012 From: kingaj at us.ibm.com (Alan King) Date: Fri, 27 Jan 2012 11:29:44 -0500 Subject: [SMI] Using COIN-OR SMI to solve a QP In-Reply-To: References: Message-ID: Yes, that is correct, SMI does not handle quadratic terms. If you would be willing to provide a test problem (hopefully for which you know the answer) then I would be willing to implement the needed changes. It is not hard to do, but does require deep knowledge of the SMI infrastructure. Best Alan ------------------------------------------------------- Alan King, PhD IBM Thomas J. Watson Research Center Yorktown Heights, New York 914-945-1236 (o) 914-260-9888 (m) http://www.research.ibm.com/people/k/kingaj/ From: "Cunniff, John M" To: Date: 01/27/2012 10:36 AM Subject: [SMI] Using COIN-OR SMI to solve a QP Sent by: smi-bounces at list.coin-or.org First, thank you for providing these optimization programs. I tried using SMI to solve a multi-stage stochastic QP mean-variance asset allocation problem (by adding a QUADOBJ section to the .core file), but the optimizer appears to ignore quadratic terms as these terms are not in the output deterministic equivalent mps file and do not appear to affect the optimal solution. We have a subscription to CPLEX here at TIAA-CREF, and it is able to read and solve the .core file as a single period QP, so I believe my formatting is correct. I read in Alan King?s presentation to INFORMS in Seattle in Nov. 2007 that it ?would be easy to add integer variables and quadratic objectives?, so I am assuming that SMI cannot solve a QP at this time. 1. Is it true that the current version of SMI cannot solve a QP? 2. Which .cpp and .hpp files would need to be altered to expand SMI to handle a QP problem, and what would those alterations be? Thank you in advance your help. John Cunniff ************************************************************************* This e-mail may contain confidential or privileged information. If you are not the intended recipient, please notify the sender immediately and then delete it. TIAA-CREF ************************************************************************* _______________________________________________ SMI mailing list SMI at list.coin-or.org http://list.coin-or.org/mailman/listinfo/smi -------------- next part -------------- An HTML attachment was scrubbed... URL: