cz.cuni.mff.ksi.jinfer.iss.heuristics.construction.glpk
Class GlpkInputGenerator

java.lang.Object
  cz.cuni.mff.ksi.jinfer.iss.heuristics.construction.glpk.GlpkInputGenerator

public final class GlpkInputGenerator
extends Object

Utility class for conversion from AMModel to GLPK MathProg plaintext format.

Method Summary

static String	generateGlpkInput(AMModel model, double alpha, double beta)
static String	generateGlpkInput(AMModel model, double alpha, double beta, List<AttributeMappingId> incumbentAMs, long k) Generates a GLPK MathProg representation of the ID set optimization for the provided model, in this variant specifically for Local Branching heuristic.
static String	generateGlpkInput(AMModel model, List<AttributeMappingId> fixed, double alpha, double beta) Generates a GLPK MathProg representation of the ID set optimization for the provided model.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

generateGlpkInput

public static String generateGlpkInput(AMModel model,
                                       double alpha,
                                       double beta,
                                       List<AttributeMappingId> incumbentAMs,
                                       long k)
                                throws IOException,
                                       InterruptedException

Generates a GLPK MathProg representation of the ID set optimization for the provided model, in this variant specifically for Local Branching heuristic.

Please see generateGlpkInput(AMModel, List, double, double) for general info.

This variant will add one more constraint to ensure the Hamming distance from the incumbent solution in the form:
SUMI (1 - ami) + SUMJ amj <= k
where I is the set of AMs in the incumbent solution, J is the set of remaining AMs in the model.

Parameters:

model - Model from which the problem formulation should be generated.

alpha - Weight of the attribute mapping support in its total weight.

beta - Weight of the attribute mapping coverage in its total weight.

incumbentAMs - List of AMs contained in the currently incumbent solution.

k - Maximal Hamming distance from the specified incumbent solution.

Returns:

String representation of the problem formulation in MathProg language, that can be directly passed to GLPK Solver.

Throws:

IOException

InterruptedException

See Also:

LocalBranching

generateGlpkInput

public static String generateGlpkInput(AMModel model,
                                       double alpha,
                                       double beta)
                                throws IOException,
                                       InterruptedException

Throws:

IOException

InterruptedException

See Also:

generateGlpkInput(cz.cuni.mff.ksi.jinfer.iss.objects.AMModel, java.util.List, double, double)

generateGlpkInput

public static String generateGlpkInput(AMModel model,
                                       List<AttributeMappingId> fixed,
                                       double alpha,
                                       double beta)
                                throws IOException,
                                       InterruptedException

Generates a GLPK MathProg representation of the ID set optimization for the provided model. In this representation, we need to provide variables, constraints and the objective function.

• Every candidate attribute mapping corresponds to one binary variable (ami).
• For each two AMs ami, amj that cannot be in the same ID set together, a constraint is added: ami + amj <= 1. Two AMs cannot be in the same ID set together, if:
  • Their type (element name) is the same, or
  • Their images intersect.
• The objective function is a weighted sum of active variables, where their weight is defined by MappingUtils#weight(AttributeMappingId, AMModel, double, double) .

Parameters:

model - Model from which the problem formulation should be generated.

alpha - Weight of the attribute mapping support in its total weight.

beta - Weight of the attribute mapping coverage in its total weight.

fixed - List of attribute mappings that should be present in the solution (their value is fixed to 1).

Returns:

String representation of the problem formulation in MathProg language, that can be directly passed to GLPK Solver.

Throws:

IOException

InterruptedException