The GCG Classification

To be able to decompose a mixed-integer program into blocks, we have to understand what the underlying structure of it is. This is done using two different types of classifiers, namely variable classifiers and constraint classifiers. As the name suggests, a variable classifier groups the variables into different categories, e.g. according to its type (binary/integer/continuous), while a constraint classifier does the same for constraints.

In addition to grouping the variables/constraints into different (disjoint) classes, it is also possible to precompute how the variables/constraints of a class should be handled in the later detection. This information is stored for each variable/constraint class in VarClassDecompInfo and ConsClassDecompInfo respectively.

Variable Classifiers

Variable classifiers can store in VarClassDecompInfo how the variables of a class should be handled by a detector:

ALL = 0 -> not specified
LINKING = 1 -> as linking variables
MASTER = 2 -> as master variables
BLOCK = 3 -> as block variables

Constraint Classifiers

Constraint classifiers can store in ConsClassDecompInfo how the constraints of a class should be handled by a detector:

BOTH -> not specified
ONLY_MASTER -> assigned to master problem
ONLY_PRICING -> assigned to pricing problem

List of Classifiers

Here you find a list of all classifiers available in GCG and a short description of their functionality. They are divided into variable and constraint classifiers.

The following variable classifiers are available:

The following constraint classifiers are available:

Adding own Classifiers

If you want to write your own classifier, i.e. define by what criteria you want to classify variables or constraints, please consult the "How to" for that.

How to add classifiers