CADETProcess.parameter_space.LatinHypercubeSampler

CADETProcess.parameter_space.LatinHypercubeSampler#

class CADETProcess.parameter_space.LatinHypercubeSampler[source]#

Bases: SamplerBase

Latin Hypercube sampler via scipy.stats.qmc.

Generates exactly n points, so the one-point-per-stratum property holds for the returned set; a subset of a larger design would not be a Latin Hypercube. Categorical parameters are stratified as extra design dimensions, giving balanced category counts. Box-bounded spaces only: raises if linear constraints are present, and any candidate rejected during postprocessing exhausts the pool (a filtered design loses stratification). Use HopsySampler for spaces that require rejection.