latin_hypercube#

exo_skryer.aux_functions.latin_hypercube(key: Array, n_samples: int, n_dim: int, *, scramble: bool = True, dtype=<class 'jax.numpy.float64'>) → tuple[Array, Array][source]#

Generate Latin hypercube samples in the unit hypercube [0, 1)^n_dim.

Parameters:

keyArray: JAX PRNG key for random number generation.
n_samplesint: Number of samples to generate. Must be positive.
n_dimint: Number of dimensions for each sample. Must be positive.
scramblebool, optional: If True (default), randomly permutes the stratum assignments for each dimension independently, reducing correlation between dimensions and improving space-filling properties. If False, strata are assigned sequentially without permutation.
dtypedtype, optional: Data type for the output array. Default is jax.numpy.float64.

Returns:

samplesndarray: Generated Latin hypercube samples with shape (n_samples, n_dim). Each value is in the range [0, 1).
keyArray: Updated PRNG key for subsequent random operations.