Change Log
0.4.2 (2024-05-01)
Enhancement: identify the cause of an unknown distribution. This may be useful if the user names a distribution and expects it to be found in a module other than
scipy.stats
.Enhancement: support NumPy 2.0 and newer. This package is not affected by the breaking changes in NumPy 2.0.
0.4.1 (2023-01-20)
Bug fix: correct
lhs.dist.inverse_uniform
so that it actually returns samples drawn from an inverse-uniform distribution. Previously, it returned samples drawn from a uniform distribution.
0.4.0 (2022-12-23)
Breaking change: require Python 3.8 or newer.
Breaking change: draw samples using the order in which the distributions are defined.
0.3.3 (2022-08-19)
Enhancement: add the
lhs.dist.inverse_uniform
distribution.
0.3.2 (2022-01-30)
Bug fix: correct the documentation example for
lhs.dist.constant
.Enhancement: add a simpler example to the documentation home page.
0.3.1 (2021-12-13)
Bug fix: use the correct subspace samples for dependent parameters.
0.3.0 (2021-12-07)
Breaking change: remove
lhs.Sampler
, this will be provided by pypfilt 0.7.Breaking change: remove
lhs.LhsSamples
, this is no longer needed because pypfilt 0.7 saves all model prior samples.Breaking change: replace
lhs.sample
functions withlhs.sample.lhs_values
andlhs.sample.draw
(also available aslhs.draw
).Enhancement: non-scalar samples and shape parameters are now supported.
Enhancement: add support for broadcasting samples to add extra dimensions.
Enhancement: distributions can now be identified by name, by providing a frozen SciPy distribution, or by providing the percent point function (inverse of the CDF).
Remove the pypfilt dependency; pypfilt 0.7 will depend on this package instead.
0.2.0 (2020-12-18)
Implement an LHS sampler for pypfilt >= 0.7.0 (not yet released).
0.1.0 (2020-12-03)
Initial release.