saealib.ReferencePointComparisonSet#

class saealib.ReferencePointComparisonSet(ref_source='population_best')[source]#

Bases: TrainingSet

Binary labels: 1 if archive point dominates the reference point, 0 otherwise.

Unlike PairwiseComparisonSet, train_x has shape (n_archive, dim), making it compatible with GlobalSurrogateManager and LocalSurrogateManager.

Parameters:

ref_source (str) – How to select the reference point: - "population_best" (default): best individual in the current population - "archive_best": best individual in the archive

Raises:

ValueError – If ctx is None (comparator required).
ValueError – If ref_source=”population_best” and population is None.

Methods

`__init__`
`build`	Return binary labels: 1 if archive point dominates reference, 0 otherwise.
`contract`	Return the training-set contract.

Method Details

ReferencePointComparisonSet.__init__(ref_source='population_best')[source]#

Parameters:: ref_source (str)
Return type:: None

ReferencePointComparisonSet.build(archive, population, ctx, candidate_x=None)[source]#

Return binary labels: 1 if archive point dominates reference, 0 otherwise.

Parameters:

archive (Archive)
population (Population | None)
ctx (OptimizationState | None)
candidate_x (np.ndarray | None)

Return type:

TrainingData

ReferencePointComparisonSet.contract()#

Return the training-set contract.

Return type:: ComponentContract