XPRSmaxim, XPRSminim

MAXIM, MINIM


Purpose
Begins a search for the optimal LP solution.
Synopsis
int XPRS_CC XPRSmaxim(XPRSprob prob, const char *flags);
int XPRS_CC XPRSminim(XPRSprob prob, const char *flags);
MAXIM [-flags]
MINIM [-flags]
Arguments
prob 
The current problem.
flags 
Flags to pass to XPRSmaxim (MAXIM) or XPRSminim (MINIM). The default is "" or NULL, in which case the algorithm used is determined by the DEFAULTALG control. If the argument includes:
the model will be solved using the Newton barrier method;
the model will be solved using the primal simplex algorithm;
the model will be solved using the dual simplex algorithm;
(lower case L), the model will be solved as a linear model ignoring the discreteness of global variables;
(lower case N), the network part of the model will be identified and solved using the network simplex algorithm;
the global model will be solved, calling XPRSglobal (GLOBAL).
Certain combinations of options may be used where this makes sense so, for example, pg will solve the LP with the primal algorithm and then go on to perform the global search.
Related Controls
Integer
Whether automatic perturbation is performed.
Maximum number of Newton Barrier iterations.
Ordering algorithm for the Cholesky factorization.
Newton barrier: level of solution output.
Max number of threads to run.
Specifies "Big M" method, or phaseI/phaseII.
Cache size in Kbytes for the Newton barrier.
1 for CPU time; 0 for elapsed time.
Type of crash.
Newton barrier crossover control.
Algorithm to use with the tree search.
Columns with this many elements are considered dense.
Pricing method for the dual algorithm.
Invert frequency.
Minimum number of iterations between inverts.
Whether to use previously loaded basis.
Iteration limit for the simplex algorithm.
Frequency and type of simplex algorithm log.
Maximum time allowed.
Degree of presolving to perform.
Specifies the operations performed during presolve.
Type of pricing to be used.
Indicates whether to re-factorize the optimal basis.
Control of the infeasibility diagnosis during presolve.

Double
Newton barrier tolerance for dual infeasibilities.
Newton barrier tolerance for relative duality gap.
Newton barrier tolerance for primal infeasibilities.
Newton barrier minimal step size.
BIGM 
Infeasibility penalty.
Zero tolerance in the Cholesky decomposition.
Markowitz tolerance for elimination phase of presolve.
Zero tolerance on eta elements.
Zero tolerance on RHS.
Markowitz tolerance for the factorization.
Cutoff set after an LP optimizer command. (Dual only)
Reduced cost tolerance.
Maximum absolute penalty variable coefficient.
Perturbation value.
Pivot tolerance.
Partial pricing candidate list sizing parameter.
Relative pivot tolerance.

Example 1 (Library)
XPRSmaxim(prob,"b");
This maximizes the current problem using the Newton barrier method.
Example 2 (Console)
MINIM -g
This minimizes the current problem and commences the global search.
Further information
1. The algorithm used to optimize is determined by the DEFAULTALG control. By default, the dual simplex is used for LP and MIP problems and the barrier is used for QP problems.
2. The d and p flags can be used with the n flag to complete the solution of the model with either the dual or primal algorithms once the network algorithm has solved the network part of the model.
3. The b flag cannot be used with the n flag.
4. The dual simplex algorithm is a two phase algorithm which can remove dual infeasibilities.
5. (Console) If the user prematurely terminates the solution process by typing CTRL-C, the iterative procedure will terminate at the first "safe" point.
Related topics
XPRSglobal (GLOBAL), XPRSreadbasis (READBASIS), XPRSgoal (GOAL), Performance Issues, The Simplex Log.


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