Best view here.
Breaking changes
- Following parameters are renamed:
COS_TILE_1
COS_TILE_2
TILE_1
TILE_2
- Most assisting commands have been shifter from using index to column names.
- Following models have been depreciated,
MultiSourceModel
MultiSpinModel
TruncatedPowerLaw
New features
- New models:
FlexibleMixtureModel
MassGapModel
NPowerLawMGaussian
NPowerLawMGaussianWithDefaultSpinMagnitude
NPowerLawMGaussianWithSpinMisalignment
- New transforms:
ComponentMassesAndRedshiftToDetectedMassAndRedshift
ComponentMassesToMassRatioAndSecondaryMass
ComponentMassesToPrimaryMassAndMassRatio
ComponentMassesToTotalMassAndMassRatio
SourceMassAndRedshiftToDetectedMassAndRedshift
- scaling factor in vt, during inference and synthetic data generation.
- Debug mode for inference.
New Design
- Introduced a refined base class for smoothed mass distribution models, enhancing functionality and maintainability.
Breaking changes
gwkokab.inference.BayesianHierarchicalModel
is deprecated in favor ofgwkokab.inference.PoissonLikelihood
gwkokab.priors.UnnormalizedUniform
deprecated in favor ofnumpyro.distributions.ImproperUniform
.- Trivial errors are removed from
gwkokab.errors
module.normal_error
truncated_normal_error
uniform_error
- After the release of numpyro-0.15.1,
less_than_equals_to
andgreater_than_equals_to
constraints are removed fromgwkokab.constraints
module. - The directory containing individual injections is removed from the synthetic data generation process.
gwkokab.population.PopFactory
is removed from the public API; instead, an instance of the class is provided in thegwkokab.population
module.gwkokab.population.NoisePopInfo
,gwkokab.population.PopInfo
, andgwkokab.population.run_noise_factory
are removed.gwkokab.inference.BayesianHierarchicalModel
in favor ofgwkokab.inference.PoissonLikelihood
gwkokab.inference.ModelPack
removed.
New features
- Constraints related to closed intervals, for details, see PR#125.
- Bijective transformations on different mass coordinates, for details, see PR#125.
- Wrapper for scaling models.
- Model registration and retrieval.
gwkokab.population.popmodel_magazine
andgwkokab.population.error_magazine
are introduced to register the population model and error models, respectively. - New models:
MultiSourceModel
MultiSpinModel
NPowerLawMGaussianWithDefaultSpinMagnitudeAndSpinMisalignment
- New API for inference, compatible with
gwkokab.inference.flowMChandler
gwkokab.inference.PoissonLikelihood
gwkokab.inference.Bake
- Progress Bar for the synthetic data generation process.
- Initial release.