The OPTMODEL Procedure
- Overview
-
Getting Started
-
Syntax
-
DetailsNamed ParametersIndexingTypesNamesParametersExpressionsIdentifier ExpressionsFunction ExpressionsIndex SetsOPTMODEL Expression ExtensionsConditions of OptimalityData Set Input/OutputControl FlowFormatted OutputODS Table and Variable NamesConstraintsSuffixesInteger Variable SuffixesDual ValuesReduced CostsPresolverModel UpdateMultiple SubproblemsMultiple SolutionsProblem SymbolsOPTMODEL OptionsAutomatic DifferentiationConversionsFCMP RoutinesMore on Index SetsThreaded and Distributed ProcessingMacro Variable _OROPTMODEL_Rewriting PROC NLP Models for PROC OPTMODEL
-
Examples
- References
Integer Variable Suffixes
The suffixes .relax, .priority, and .direction are applicable to integer variables.
For an integer variable x, setting x.relax to a nonzero, nonmissing value relaxes the integrality restriction. The value of x.relax is read as either 1 or 0, depending on whether or not integrality is relaxed. This suffix is ignored for noninteger variables.
The value that is contained in x.priority sets the priority of an integer variable x for use with branching in the MILP solver or selection in the CLP solvers. This value can be any nonnegative, nonmissing
number. The default value is 0, which indicates default branching priority. Variables with positive .priority values are assigned greater priority than the default. Variables with the highest .priority values are assigned the highest priority. Variables with the same .priority value are assigned the same branching priority.
The value of x.direction assigns a branching direction to an integer variable x. This value should be an integer in the range –1 to 3. A noninteger value in this range is rounded on assignment. The default
value is 0. The significance of each integer is found in Table 5.16.
Table 5.16: Branching Directions
|
Value |
Direction |
|
–1 |
Round down to nearest integer |
|
0 |
Default |
|
1 |
Round up to nearest integer |
|
2 |
Round to nearest integer |
|
3 |
Round to closest presolved bound |
Suppose the solver branches next on an integer variable x whose last LP relaxation solution is 3.3. Suppose also that after passing through the presolver, the lower bound of x is 0 and the upper bound of x is 10. If the value in x.direction is –1 or 2, then the solver sets x to 3 for the next iteration. If the value in x.direction is 1, then the solver sets x to 4. If the value in x.direction is 3, then the solver sets x to 0.
The MPS data set created by the SAVE MPS statement (SAVE MPS Statement) includes a BRANCH section if any nondefault .priority or .direction values have been specified for integer variables.