- Function GMP::Instance::CreateDual(GMP, name)
GMP::Instance::CreateDual
The function GMP::Instance::CreateDual
generates a mathematical
program that is the dual representation of the specified generated
mathematical program. The generated mathematical program should have
model type LP.
GMP::Instance::CreateDual(
GMP, ! (input) a generated mathematical program
name ! (input) a string expression
)
Arguments
- GMP
An element in the set
AllGeneratedMathematicalPrograms
.- name
A string that holds the name for the dual of the generated mathematical program.
Return Value
A new element in the set
AllGeneratedMathematicalPrograms
with the name as specified by the name argument.
Note
The name argument should be different from the name of the original generated mathematical program.
If an element with name specified by the name argument is already present in the set
AllGeneratedMathematicalPrograms
the corresponding generated mathematical program will be replaced (or updated in case the same symbolic mathematical program is involved).Before a generated mathematical program is dualized, AIMMS first transforms it temporary into a standard form using the following rules:
Each column \(x_i\) that is frozen to 0 is deleted.
For each column \(x_i\) with upper bound \(u_i\), \(u_i \neq 0\) and \(u_i < \infty\), an extra row \(x_i \leq u_i\) is added.
For each column \(x_i\) with lower bound \(l_i\), \(l_i \neq 0\) and \(l_i > -\infty\), an extra row \(x_i \geq l_i\) is added.
Each ranged row \(l_j \leq a^T x \leq u_j\) (\(l_j > -\infty\) and \(u_j < \infty\)) is replaced by two rows \(l_j \leq a^T x\) and \(a^T x \leq u_j\).
By using the suffix
.ExtendedConstraint
it is possible to refer to the rows that are added to create the standard form:The constraint
v.ExtendedConstraint('DualUpperBound',i)
is added for a variablev(i)
with an upper bound unequal to 0 and inf.The constraint
v.ExtendedConstraint('DualLowerBound',i)
is added for a variablev(i)
with a lower bound unequal to 0 and -inf.The constraints
c.ExtendedConstraint('DualLowerBound',j)
andc.ExtendedConstraint('DualUpperBound',j)
replace a ranged constraintc(j)
.
The objective variable for the dual mathematical program will become
mp.ExtendedConstraint(DualObjective)
and the objective constraint will bemp.ExtendedVariable(DualDefinition)
, wheremp
denotes the symbolic mathematical program.
Example
Assume that ‘PrimalModel’ is a mathematical program with the following declaration (in ams format):
Variable x1 { Range: [0, 5]; } Variable x2 { Range: nonnegative; } Variable obj { Definition: - 7 * x1 - 2 * x2; } Constraint c1 { Definition: - x1 + 2 * x2 <= 4; } MathematicalProgram PrimalModel { Objective: obj; Direction: minimize; Type: LP; }Then
GMP::Instance::CreateDual
will create a dual mathematical program with variablesname lower upper c1 -inf 0 obj_definition -inf inf x1.ExtendedConstraint('DualUpperBound') -inf 0 PrimalModel.ExtendedConstraint('DualObjective') -inf infand constraints
x1: - c1 + 7 * obj_definition + x1.ExtendedConstraint('DualUpperBound') >= 0 ; x2: + 2 * c1 + 2 * obj_definition >= 0 ; obj: obj_definition = 1 ; PrimalModel.ExtendedVariable('DualDefinition'): - 4 * c1 - 5 * x1.ExtendedConstraint('DualUpperBound') + PrimalModel.ExtendedConstraint('DualObjective') = 0 ;
See also
The function GMP::Instance::Generate
. See Modifying an Extended Math Program Instance of the Language Reference for
more details on extended suffixes.