Model

class pygcgopt.gcg.Model

Main class for interaction with the GCG solver.

addDecomposition(self, PartialDecomposition partialdec)

Adds a user specified decomposition to GCG.

The passed PartialDecomposition can be partial or finnished. A partial decomposition will be completed by GCG using its detector loop. If a finnished decomposition is passed, GCG will skip the detection loop and use the provided decomposition right away.

addDecompositionFromConss(self, master_conss, *block_conss)

Adds a user specified decomposition to GCG based on constraints.

Parameters:

  • master_conss – An iterable of Constraint objects. Can be the empty list.

  • block_conss – Any number of lists. The Constraints from each list will be turned into their own block. (optional)

Returns:

The created PartialDecomposition object

Creates a PartialDecomposition object using createDecomposition(). Fixes the master constraints with fixMasterConss() and the block constraints with fixBlockConss(). The decomposition is added with addDecomposition().

addPreexistingPartialDecomposition(self, PartialDecomposition partialdec)
addVar(self, *args, **kwargs)
createDecomposition(self)

Creates a new empty PartialDecomposition.

The created PartialDecomposition object can be used to fix constraints and variables. Afterwards, it can be passed to the model through addPreexistingPartialDecomposition().

See also

detect(self)

Detect the problem.

Can be executed before or after presolving. If executed before presolving, the structure is detected on the original problem and presolving is skiped when solving the problem later.

See also

getDetprobdataOrig(self)

returns the detprobdata for unpresolved problem

getDetprobdataPresolved(self)

returns the detprobdata for presolved problem

getDualbound(self)

Retrieve the best dual bound.

This retrieves the same dual bound that GCG reports in the console log. The dual bound is based on the objective value of the optimized linear programming relaxation at the current node.

Note

The dual bound at the root node is not always equal to the solution of the restricted master problem LP relaxation. This can be due to master cuts or abortion of the pricing loop before the restricted master problem is optimal.

Returns:

The best dual bound of the current node.

getMasterProb(self)

Provides access to the GCG master problem.

Returns:

An instance of scip#Model that represents the master problem.

getMastervars(self, var)

Returns the master variables corresponding to the variable of original problem

Parameters:

var – Variable of original problem

Returns:

List of master variables

getNConsClassifiers(self)

returns the number of all constraint classifiers

Returns:

number of constraint classifiers

Return type:

int

getNVarClassifiers(self)

returns the number of all variable classifiers

Returns:

number of variable classifiers

Return type:

int

getPartDecompFromID(self, id)

returns a partial decomposition regarding to the given partialdecomp id

Parameters:

id – patial decomposition id as int

Returns:

PartialDecomposition object

includeConsClassifier(self, ConsClassifier consclassifier, consclassifiername, desc, priority=0, enabled=True)

includes a constraint classifier

Parameters:

  • consclassifier – an object of a subclass of consclassifier#ConsClassifier.

  • consclassifiername – name of constraint classifier

  • desc – description of constraint classifier

includeDefaultPlugins(self)

Includes all default plug-ins of GCG into SCIP

Called automatically during initialization of the model.

includeDetector(self, Detector detector, detectorname, decchar, desc, freqcallround=1, maxcallround=INT_MAX, mincallround=0, freqcallroundoriginal=1, maxcallroundoriginal=INT_MAX, mincallroundoriginal=0, priority=0, enabled=True, enabledfinishing=False, enabledpostprocessing=False, skip=False, usefulrecall=False)

includes a detector

Parameters:

  • detector – An object of a subclass of detector#Detector.

  • detectorname – name of detector

  • desc – description of detector

For an explanation for all arguments, see GCGincludeDetector().

includePricingSolver(self, PricingSolver pricingSolver, solvername, desc, priority=0, heuristicEnabled=False, exactEnabled=False)
includeScore(self, Score score, scorename, shortname, desc)

includes a score

Parameters:

  • detector – An object of a subclass of detector#Detector.

  • detectorname – name of the detector

For an explanation for all arguments, see GCGincludeDetector().

includeVarClassifier(self, VarClassifier varclassifier, varclassifiername, desc, priority=0, enabled=True)

includes a variable classifier

Parameters:

  • varclassifier – an object of a subclass of varclassifier#VarClassifier.

  • varclassifiername – name of variable classifier

  • desc – description of variable classifier

listConsClassifiers(self)

lists all constraint classifiers that are currently included

Returns:

list of strings of the constraint classifier names

listDecompositions(self)

Lists all finnished decompositions found during the detection loop or provided by the user.

listDetectors(self)

Lists all detectors that are currently included

Returns:

A list of strings of the detector names

Note

The detectors can be enabled or disabled using the appropriate methods by passing the name.

See also

listPricingSolvers(self)
listScores(self)

Lists all scores that are currently included

Returns:

A list of strings of the score names

listVarClassifiers(self)

lists all variable classifiers that are currently included

Returns:

list of strings of the variable classifier names

optimize(self)

Optimize the problem.

This will transform, presolve and detect the problem if neccessary. Otherwise, GCG will solve the problem directly.

presolve(self)

Presolve the problem.

printStatistics(self)

Print solving statistics of GCG to stdout.

printVersion(self)

Print version, copyright information and compile mode of GCG and SCIP

setDetectorEnabled(self, detector_name, is_enabled=True)

Enables or disables a detector for detecting partial decompositions.

Parameters:

  • detector_name – The name of the detector.

  • is_enabled – Decides weather the detector should be enabled or diabled.

This is a convenience method to access the boolean parameter “detection/detectors/<name>/enabled”.

Note

Disabling a detector using this method is not enough to ensure that it will not run. In addition setDetectorFinishingEnabled() and setDetectorPostProcessingEnabled() have to be used.

Use listDetectors() to obtain a list of all detectors.

setDetectorFinishingEnabled(self, detector_name, is_enabled=True)

Enables or disables a detector for finishing partial decompositions.

Parameters:

  • detector_name – The name of the detector.

  • is_enabled – Decides weather the detector should be enabled or diabled.

This is a convenience method to access the boolean parameter “detection/detectors/<name>/finishingenabled”.

See also

setDetectorPostprocessingEnabled(self, detector_name, is_enabled=True)

Enables or disables a detector for postprocessing partial decompositions.

Parameters:

  • detector_name – The name of the detector.

  • is_enabled – Decides weather the detector should be enabled or diabled.

This is a convenience method to access the boolean parameter “detection/detectors/<name>/postprocessingenabled”.

See also

setGCGSeparating(self, setting)

Sets parameter settings of all separators

Parameters:

setting – the parameter settings (SCIP_PARAMSETTING)

setPricingSolverEnabled(self, pricing_solver_name, is_enabled=True)

Enables or disables exact and heuristic solving for the specified pricing solver.

Parameters:

  • pricing_solver_name – The name of the pricing solver.

  • is_enabled – Decides weather the pricing solver should be enabled or diabled.

This is a convenience method to access the boolean parameters “pricingsolver/<name>/exactenabled” and “pricingsolver/<name>/heurenabled”.

Use listPricingSolvers() to obtain a list of all pricing solvers.

setPricingSolverExactEnabled(self, pricing_solver_name, is_enabled=True)

Enables or disables exact solving for the specified pricing solver.

Parameters:

  • pricing_solver_name – The name of the pricing solver.

  • is_enabled – Decides weather the pricing solver should be enabled or diabled.

This is a convenience method to access the boolean parameter “pricingsolver/<name>/exactenabled”.

Use listPricingSolvers() to obtain a list of all pricing solvers.

setPricingSolverHeuristicEnabled(self, pricing_solver_name, is_enabled=True)

Enables or disables heuristic solving for the specified pricing solver.

Parameters:

  • pricing_solver_name – The name of the pricing solver.

  • is_enabled – Decides weather the pricing solver should be enabled or diabled.

This is a convenience method to access the boolean parameter “pricingsolver/<name>/heurenabled”.

Use listPricingSolvers() to obtain a list of all pricing solvers.

setVarClassifierEnabled(self, varclassifier_name, is_enabled=True)

enables or disables a variable classifier

Parameters:

  • varclassifier_name – the name of the variable classifier

  • is_enabled – decides weather the variable classifier should be enabled or diabled.

This is a convenience method to access the boolean parameter “detection/classification/varclassifier/<name>/enabled”.

writeAllDecomps(self, directory=u'alldecompositions/', extension=u'dec', bool original=True, bool presolved=True, createDirectory=True)

Writes all decompositions to disk

Parameters:

  • directory – A path to a folder where to store the decomposition files

  • extension – Extension without a dot. Decides the output format. Use “dec” to output decomposition files

  • createDirectory – Automatically create the directory specified in directory if it does not exist