saealib.FeasibilityClassificationSet#

class saealib.FeasibilityClassificationSet(source='archive')[source]#

Binary labels derived from constraint violation: 1 if feasible, 0 otherwise.

A sample is feasible when cv <= eps_cv, where eps_cv is taken from ctx.problem.eps_cv (defaults to 1e-6 when ctx is None).

This set is algo-agnostic and works with any surrogate that accepts binary classification targets (shape (n_train,)).

Parameters:: source (str) – "archive" to use the evaluated archive, or "population" to use the current population.
Raises:: ValueError – If source="population" but population is None.

Methods

`__init__`
`build`	Return feasibility binary labels (1 = feasible, 0 = infeasible).
`contract`	Return the training-set contract.

Method Details

FeasibilityClassificationSet.__init__(source='archive')[source]#

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

Return feasibility binary labels (1 = feasible, 0 = infeasible).

Parameters:

Return type:

TrainingData

FeasibilityClassificationSet.contract()#

Return the training-set contract.