Purpose
Starts the global
search for an integer solution after solving the LP relaxation with
XPRSmaxim (
MAXIM) or
XPRSminim (
MINIM) or continues a global search if it has been interrupted.
Synopsis
int XPRS_CC XPRSglobal(XPRSprob prob);
GLOBAL
Argument
|
prob
|
The current problem.
|
Related Controls
Integer
|
|
Node selection criterion.
|
|
|
Once a global entity has been selected for branching, this control
determines whether the branch with the minimum or maximum estimate is followed first.
|
|
|
Limit for node selection criterion.
|
|
|
Number of rounds of lifted cover inequalities at top node.
|
|
|
1 for CPU time; 0 for elapsed time.
|
|
|
Maximum depth in the tree at which cuts are generated.
|
|
|
Frequency at which cuts are generated in the tree search.
|
|
|
Specifies the cut strategy.
|
|
|
Algorithm to use with the tree search.
|
|
|
Number of rounds of Gomory cuts at the top node.
|
|
|
Number of integer solutions to store.
|
|
|
Maximum number of MIP solutions to find.
|
|
|
Maximum number of nodes in Branch and Bound search.
|
|
|
Number of slave processors used for parallel MIP search.
|
|
|
Maximum time allowed.
|
|
|
Global print flag.
|
|
|
Type of integer preprocessing to be performed.
|
|
|
Node selection control.
|
|
|
Indicates whether to re-factorize the optimal basis.
|
|
|
Number of infeasible global entities on which to perform strong branching.
|
|
|
Number of dual iterations to perform strong branching.
|
|
|
The size of the candidate list of global entities for strong branching.
|
|
|
Number of rounds of lifted cover inequalities in the tree.
|
|
|
Number of rounds of Gomory cuts in the tree.
|
|
|
Node selection degradator estimate control.
|
Double
|
|
Factor to multiply estimated degradations by.
|
|
|
Cutoff set after an LP optimizer command.
|
|
|
Absolute optimality stopping criterion.
|
|
|
Amount added to objective function to give new cutoff.
|
|
|
Percentage cutoff.
|
|
|
Relative optimality stopping criterion.
|
|
|
Target object function for global.
|
|
|
Integer feasibility tolerance.
|
|
|
Default pseudo cost in node degradation estimation.
|
Example 1 (Library)
The following example inputs a problem
fred.mat, solves the LP and the global problem before printing the solution to file.
XPRSreadprob(prob,"fred","");
XPRSmaxim(prob,"");
XPRSglobal(prob);
XPRSwriteprtsol(prob);
Example 2 (Console)
The equivalent set of commands for the Console Optimizer are:
READPROB fred
MAXIM
GLOBAL
WRITEPRTSOL
Further information
1. When an optimal LP solution has been found with
XPRSmaxim (
MAXIM) or
XPRSminim (
MINIM), the
search for an integer solution is started using
XPRSglobal (
GLOBAL). In many cases
XPRSglobal (
GLOBAL) is to be
called directly after
XPRSmaxim (
MAXIM)/
XPRSminim (
MINIM). In
such circumstances this can be achieved slightly more efficiently using the
g flag to
XPRSmaxim (
MAXIM)/
XPRSminim (
MINIM).
2. If a global search is interrupted and
XPRSglobal (
GLOBAL) is
subsequently called again, the search will continue where it left off. To restart the search at the
top node you need to call either
XPRSinitglobal or
XPRSpostolve (
POSTSOLVE).
3. The controls described for XPRSmaxim (MAXIM) and XPRSminim (MINIM) can also be used to control the XPRSglobal (GLOBAL) algorithm.
4. (Console) The global search may be interrupted by typing CTRL-C as long
as the user has not already typed ahead.
5. A summary log of six columns of information is output every
n nodes,
where
-n is the value of
MIPLOG (see
The Global Log).
6. Optimizer library users can check the final status of the global search using
the
MIPSTATUS problem attribute.
7. The Optimizer supports global (i.e. active node list) files in excess of 2
GigaBytes by spreading the data over multiple files. The initial global file is given the name
probname.glb and subsequent files are
named
probname.glb.1,
probname.glb.2,.... No individual file will
be bigger than 2GB.
Related topics
If you have any comments or suggestions about these pages,
please send mail to docs@dashoptimization.com.