Recent Releases of dynesty
dynesty - v2.1.4
This is bug-fix release. The main user-visible changes is that npdim= option of dynesty is removed. Also because of the code change, you will not be able to resume previous dynesty runs from earlier (<2.1.3) dynesty versions. Detailed changelog is below:
- Get rid of npdim option that at some point may have allowed the prior transformation to return higher dimensional vector than the inputs. Note that due to this change, restoring the checkpoint from previous version of the dynesty won't be possible) (issues #456, #457) (original issue reported by @MichaelDAlbrow, fixed by @segasai ) ### Fixed
- Fix the way the additional arguments are treated when working with dynesty's pool. Previously those only could have been passed through dynesty.pool.Pool() constructor. Now they can still be provided directly to the sampler (not recommended) ( #464 , reported by @eteq, fixed by @segasai )
- change the .ptp() method to np.ptp() function as it is deprecated in numpy 2.0 ( #478 , reported and patched by @joezuntz)
- Fix an error if you use run_nested() several times (i.e. with maxiter option) while using blob=True. ( #475 , reported by @carlosRmelo)
- Python
Published by segasai almost 2 years ago
dynesty - v2.1.0
This a major release with a significant change of the sampler. It is now able to correctly sample likelihood functions with plateaus, (i.e likelihood functions with areas where likelihood value is exactly the same).
The previous behaviour also led to inaccurate log(Z) estimates if some fraction of the prior volume had log(P)=-LARGENUMBER
- Python
Published by segasai over 3 years ago
dynesty - v2.0.3
This is a release with a major performance bug fix for the dynamical sampler
A major performance regression has been discovered that lead to dynamic nested sampling batches becoming slower and slower. The regression was introduced in 1.2 and can easily lead to a factor of several slower performance if a large number of batches is used. ( see #415 ) Some small fixes related to resuming of interrupted sampling were applied as well.
- Python
Published by segasai over 3 years ago
dynesty - v2.0.2
Minor bug fix release.
- When checkpointing is on the dynamic sampler will always checkpoint in the end of the runnested() irrespective of checkpointtime
- Equally weighted samples are now randomly shuffled ( #408 )
- The livepoints option was somewhat broken when blob option was introduced requiring a tuple of 4 elements irrespective of whether your likelihood returns blobs or not. Now if you use blob=True and want to provide livepoints you need to provide 4 elements (u,v,logl,blobs). If you use blob=False you will need to provide just 3 elements as before (u,v,logl)
- Python
Published by segasai over 3 years ago
dynesty - v2.0.0
This is a major release with several significant improvements.
- The implementation of the check-points to save progress and allow restarting of fits.
- A new simple interface to obtain equally weighted samples directly from results object.
- Allow likelihood functions to return additional computed quantities (blobs) that will be saved together with samples.
- Random slice sampling now supports interval doubling scheme which may increase the performance in the case of bad posteriors.
See the full changelog and links to the documentation of the new features here https://dynesty.readthedocs.io/en/v2.0.0/#changelog
- Python
Published by segasai over 3 years ago
dynesty - v1.2.2
Bug fix release which addresses quite a serious bug which can lead to biased posteriors.
- The problem with biased posteriors was fixed when using multi-ellipsoid bounds and rslice and rwalk samplers. Previously the chains did not satisfy detailed balance. (issue #364). Original discovery of the problem and help by Colm Talbot. In the case of complex posteriors, somewhat slower performance may be seen.
- Fix the issue introduced in 1.2.1 when the prior_transform returns a tuple or or a list (rather than numpy array). Now that should be accepted.
- Python
Published by segasai about 4 years ago
dynesty - v1.2.1
Small bugfix release
- The arguments of prior_transform and likelihood function are now explicitely copied, so the sampling can work if those function apply changes to argument vectors ( #362 )
- Fix the compilation of the docs, and update them a bit
- Python
Published by segasai about 4 years ago
dynesty - v1.2.0
This version has multiple changes that should improve stability and speed. Some of the main changes include also * The default dynamic sampling behaviour has been changed to focus on the effective number of posterior samples as opposed to KL divergence. * The rstagger sampler has been removed and the default choice of the sampler may be different compared to previous releases depending on the dimensionality of the problem. * dynesty should now provide 100% reproduceable results if the rstate object is provided. It needs to be a new generation numpy Random Generator (as opposed to numpy.RandomState).
Most of the changes in the release have been contributed by Sergey Koposov who has joined the dynesty project. A more detailed list of changes is available in the changelog
- Python
Published by segasai about 4 years ago
dynesty - Version 1.0
This is the first "out of beta" release. Woohoo!
- Python
Published by joshspeagle over 6 years ago
dynesty - dynesty v0.9.7
Creating a Zenodo link and DOI for this.
- Python
Published by joshspeagle almost 7 years ago