Skip to content

Latest commit

 

History

History
82 lines (73 loc) · 7.17 KB

CHANGELOG.md

File metadata and controls

82 lines (73 loc) · 7.17 KB

Changelog

Latest

  • #162: fix sign error in python implementation of parallel tempering
  • #161: persist markov chain proposal after parallel sampling

v1.5.2

  • #115: Improve initialization of custom PyProposals in MarkovChain
  • #114: Disable inheriting from PyProposal and PyModel

v1.5.1

  • #138: Ignore fixed dimensions when computing ESS & rhat
  • #157: Pin numpy<2.0
  • #141: Add minimal example for genome-scale sampling
  • #152: Add test-case for selecting start point with equality constraints
  • #69: Add CLA
  • #155: Increase numerical accuracy for parallel tempering when using python multiprocessing
  • #153: Add tuning target selection to setup function

v1.5.0

  • #145: Improve docs
  • #127: Add method to heuristically setup good sampling configurations depending on the problem
  • #139: Add transformation of start points during add_equality_constraints
  • #88: Fix parallelization for sampling with mcbackend as sample storage
  • #123: Add parallel tempering support for non-MPI users
  • #137: Fix rare bug for users implementing their own PyProposals for non-uniform sampling
  • #136: Improve numerical stability of mixture models
  • #133: Add default starting point to proposals and markov chains when none is given
  • #128: Add stepsize tuning with Python TuningTargets
  • #122: Fix rounding for Reversible Jump MCMC
  • #125: Update Readme.md & Fix automatic docs generation
  • #106: Add official support for python 3.12
  • #130: Update requirements-dev.txt
  • #108: For development: Fix setup.py warning on missing commit ID
  • #119: Fix numpy deprecation warnings in unit tests
  • #121: Update docs for problem

v.1.4.1

  • #120: Fix parallel sampling with additional equality constraints
  • #117: Update Optlang, sympy and numpy dependencies
  • #118: Fix back_transform when manually adding equality constraints and rounding

v1.4.0

  • #113: Fix small bug in back_transform
  • #111: Update documentation
  • #92: Add convenience functions for generating hybercubes, simplices and Birkhoff polytopes
  • #110: Add convenience method to problem for computing slacks
  • #82: Add option to compute_chebyshev to return point in transformed or original space
  • #102: Add RJMCMC
  • #104: Add multiphase sampling
  • #83: Readd vectorization for hopsy
  • #103: Fix callback example
  • #105: Drop support for unsupported python 3.7
  • #101: Add helper functions for testing if polytope is empty
  • #99: Fix benchmark example typo
  • #98: Consistent names for model functions
  • #97: Disable jekyll in automatic doc building
  • #96: Fix pickling for problem
  • #95: Fix parallel sampling for PyProposals
  • #91: Add benchmark example
  • #90: Reduce ci build times
  • #87: Reduce boilerplate for custom pyproposals
  • #71: Rename functions and deprecate old variants

v1.3.1

  • #85: Fix release for Python 3.10 and Python 3.11 windows

v1.3.0

  • #78: Make (mac+linux+win compatible) sdist
  • #79: Add installation example for arch linux
  • #33: Add direct support for equality constraints
  • #77: Add feasibility check to compute_chebyshev_center

v1.2.0

  • #70: Add correct license
  • #64: Optional in memory saving of states
  • #58: Add mcbackend support
  • #50: Bugfix: Fix state copying for parallel chains
  • #49: Parallelize computation of statistics on samples (arviz)
  • #48: Replace keyword n_threads by n_procs in sample function
  • #47: Add mcbackend compatibility
  • #11: Add parallel tempering with MPI