Skip to content

Releases: prmiles/pymcmcstat

v1.9.1

24 Oct 16:49
Compare
Choose a tag to compare
  • Fixed #79 - Chainstats display now auto updates formatting for really large or small numbers.
  • Fixed #82 - Initial value check routine in Parallel MCMC now correctly removes extra columns.

v1.9.0

19 Jul 14:45
a936aa6
Compare
Choose a tag to compare
  • Added feature to results structure to store all parameters names, regardless of whether or not they are included in sampling.
  • Added mcmcplot package to requirements. MCMCPlotting module is noted as deprecated.
  • Added new module for uncertainty propagation. Aims to provide more flexible API for user to plot different combinations of credible and prediction intervals.
  • Added a plotting routine so that you can plot a 2-D interval in 3-D space.

v1.8.0

28 Jun 17:36
Compare
Choose a tag to compare
  • Added acceptance rate display feature when calling chain statistics
  • User can specify skip or maxpoints in pairwise correlation and chain panel plots in order to thin chain.
  • User can request item definitions when calling chainstats.
  • Resolved #51 by adding method to check if limits are incompatible. User is provided with more descriptive error message.
  • Updated documentation for Bayesian components.
  • Added saving routines to accomodate post-processing. A lighter version of the results dictionary is saved to json when log routines are used for chains. This should address #55.
  • Option to return KDE objects when using plot_density_panel. User can then evaluate pdf directly. This should address #66.

v1.7.1

03 May 10:07
d233760
Compare
Choose a tag to compare
  • Fixed bug #41.
  • Updated how version attribute is defined in package and referenced in setup.py and documentation.

v1.7.0rc2

05 Apr 21:41
Compare
Choose a tag to compare

Updated docs

v1.7.0rc1

05 Apr 17:16
eccdec8
Compare
Choose a tag to compare

Test release for v1.7.0

v1.7.0

05 Apr 21:56
fee6471
Compare
Choose a tag to compare
  • Added custom sampler feature. Interface requires creation of a class, but allows for inclusion of things like Gibbs samplers. User can explicitly pass in features to the sum-of-squares function.
  • Added check routines for reading in parallel MCMC results to ensure only directories with name 'chain_' are queried.
  • All files were updated to comply with formatting standards using flake8.
  • Added coverage report check to ensure minimal testing requirements are met.
  • Added original covariance matrix to results structure for reference.

v1.6.0rc1

31 Aug 16:17
d3a3f07
Compare
Choose a tag to compare

Initial test release of v1.6.0.

  • Added optimal handling of prior function used numpy array multiplication.
  • Fixed bug with saving results to json file - output directory now automatically made.
  • Updated PI plotting to allow any valid matplotlib input.
  • Added Gelman-Rubin diagnostics to ChainStatistics module, with display feature.
  • Added numpy error settings option to MCMC initialization.
  • Setup no adaptation feature. Can sample parameters without adapting them.
  • Added restart routine to ParallelMCMC. Can restart from json files.
  • Added parallel chain organization routines to the chain subpackage.

v1.6.0

31 Aug 16:28
6bb2b63
Compare
Choose a tag to compare

Final release of v1.6.0.

  • Added optimal handling of prior function used numpy array multiplication.
  • Fixed bug with saving results to json file - output directory now automatically made.
  • Updated PI plotting to allow any valid matplotlib input.
  • Added Gelman-Rubin diagnostics to ChainStatistics module, with display feature.
  • Added numpy error settings option to MCMC initialization.
  • Setup no adaptation feature. Can sample parameters without adapting them.
  • Added restart routine to ParallelMCMC. Can restart from json files.
  • Added parallel chain organization routines to the chain subpackage.

v1.5.0rc2

27 Jul 17:21
6a5a0a5
Compare
Choose a tag to compare

Added Zenodo link to release