maximize, minimize


Purpose
Maximize/minimize the current problem.
Synopsis
procedure maximize(alg:integer, obj:linctr)
procedure maximize(obj:linctr)
procedure maximize(alg:integer, qobj:qexp)
procedure maximize(qobj:qexp)
procedure maximize(alg:integer, nlobl:nlctr)
procedure maximize(nlobl:nlctr)
Arguments
alg 
Algorithm choice:
XPRS_BAR 
Newton-Barrier to solve LP
XPRS_DUAL 
Dual simplex
XPRS_NET 
Network solver
XPRS_LIN 
Only solve LP ignoring all global entities
XPRS_PRI 
Primal simplex
XPRS_ENUM 
Start a search for the n-best MIP solutions
XPRS_LPSTOP 
Stop the MIP solution process after solving the first LP
XPRS_CONT 
Continue a previously interrupted solution process
XPRS_LOCAL 
Solve the linearization of the problem (mmxnlp only)
XPRS_CORELP 
Solve the linear part of the problem (mmxnlp only)
obj 
Objective function constraint
qobj 
Quadratic objective function (with module mmquad)
nlobj 
Non linear objective function (with module mmnl)
Example
The following maximizes Profit using the dual simplex algorithm and stops before the global search:
declarations
 Profit:linctr
end-declarations
	  	  
maximize(XPRS_DUAL+XPRS_LPSTOP, Profit)  
The following minimizes MinCost using the Newton-Barrier algorithm and ignoring all global entities
declarations
 MinCost:linctr
end-declarations
	  	  
minimize(XPRS_BAR+XPRS_LIN, MinCost) 
Further information
1. This procedure calls the Optimizer to maximize/minimize the current problem (excluding all hidden constraints) using the given constraint as objective function. Optionally, the algorithm to be used can be defined. By default, the global search is executed automatically if the problem contains any global entities. Where appropriate, several algorithm choice parameters may be combined (using plus signs).
2. If XPRS_LIN is specified, then the discreteness of all global entities is ignored, even during the presolve procedure.
3. If XPRS_LPSTOP is specified, then just the LP at the top node is solved and no Branch-and-Bound search is initiated. But the discreteness of the global entities is taken into account in presolving the LP at the top node. Note also that getprobstat still returns information related to the MIP problem when this option is used although only an LP solve has been executed and the solution information returned by getsol corresponds to the current LP solution. However, if the the MIP is solved to optimality during this call, the MIP optimal solution will be returned by getsol.
4. If XPRS_CONT is used after a solve has completed, the routine returns immediately without altering the current problem status.
5. If XPRS_ENUM is specified, the optimiser starts a search for the n-best MIP solutions. The maximum number of solutions to store may be specified using the XPRS_enummaxsol (default: 10). After the execution of the enumeration, the number of solutions found during the search is returned by the control parameter XPRS_enumsols. The procedure selectsol can then be used to select one of these solutions.
6. If XNLP_LOCAL is specified for a non-linear problem having been loaded using mmxnlp and which have been solved using XSLP, then the current linearization will be reoptimized.
7. If XNLP_CORELP is specified for a non-linear problem having been loaded using mmxnlp, then only the linear part of the problem will be loaded and optimized. This is usefull for checking if the linear part of the problem is well posed.
8. Support for quadratic programming requires the module mmquad or mmnl.
9. Support for general nonlinear programming requires the module mmxnlp.
Related topics
postsolve, loadprob, selectsol.
Module
mmxprs


If you have any comments or suggestions about these pages, please send mail to support@fico.com.

© Copyright 2001-2013 Fair Isaac Corporation. All rights reserved.