pyABC 0.10.15

Sampler:

• Allow redis dynamical sampler to only wait for relevant particles after
  a generatio, giving a speed-up without drawbacks (#448)
• Add option to limit number of delayed look-ahead samples to limit memory
  usage (#428)

Logging:

• Standardize output of floats (#450)
• Use hierarchical logging (ABC.Submodule) (#417)

General:

• Refactor: Remove deprecated nr_samples_per_parameter, internal
  simplifications (#422)
• Tidy up and minimize dependencies (#436, #441)
• External: Remove simulation files after collecting results (#434)
• Make feather/pyarrow dependency optional for older hardware (#442)

Documentation:

• Add description of JupyterHub to documentation (#439)

CI:

• Test webserver basic functionality
• Rerun stochastically failing tests (all #436)
• Test whether dataframe storage routines work properly (#442)

pyABC 0.10.14

General:

• Speed up parameter handling (30% internal speed-up) (#387)
• Streamline testing via tox and pre-commit hooks, add flake8 config file (#408)
• Update to python 3.9 (#411)
• Extract PEtab bounds and nominal parameters (#416)
• Allow specifying parameter names in density plots (#416)
• Normalize look-ahead sampling by subpopulation ESS (#418, #421)

Documentation:

• Update contribution and issue guidelines (#408)
• Add example on yaml2sbml usage (#416)
• Clean up user docs on parallelization and storage (#419)

Fixes:

• Fix redis for later started workers (#410)
• Fix PEtab parameter scale import, support all PEtab prior distributions (#413)

Database:

• Add database migration scheme based on alembic (#419)
• Store proposal ids, increment database version to 0 (#419)

pyABC 0.10.13

• Update branch name master -> main in agreement with
  https://github.com/github/renaming (#406).
• Notebook improvements:
  • Add ground truth to noise notebook.
  • Rename notebook "quickstart" -> "model_selection".
  • Split notebook tests in 2, extend, reduce test matrix (python 3.7).
• Improve output at the beginning and end (e.g. no end_time at beginning;
  show duration).
• Add walltime plots (plot_total_walltime, plot_walltime,
  plot_eps_walltime).
• Make sure ABCSMC.run() is always properly finished (sampler, history)
  by a wrapper (all #401).
• Redis sampler with look-ahead mode:
  • Fix insufficient logging of look-ahead samples.
  • Log all accepted particles.
• Add plot_lookahead_... plots for look-ahead mode diagnostics.
• Add global figure parameter settings for pyABC (all #405).

pyABC 0.10.12

• Check components for their adaptivity for correct application of look-ahead
  mode (#397).

pyABC 0.10.11

• Require pandas >= 1.2.0 for PEtab.

pyABC 0.10.10

• Uniquely identify redis runs via a uuid.
• Secure proper termination of processes for redis and dask (all #338).
• Rework redis sampler, implement a look-ahead mode for pre-defined (#338)
  and adaptive algorithms (#374) for near-perfect parallel efficiency.
• Allow subprocessing in multiprocessed redis workers (#377).
• Add a static-scheduling redis sampler (#379).
• Check whether epsilon is nan before sampling (#382).
• Remove useless IdentityFakeDistance (#390).

pyABC 0.10.9

0.10.9 (2020-11-28)

• Add stopping criterion for total walltime (#370).

0.10.8 (2020-11-27) (forgotten to release 🙈 )

• Allow to normalize acceptance rate plots by ESS (#346).
• Add a list of pyABC references (#348).
• Update to petabtests 0.0.0a5 (#362).
• Add stopping criterion for total number of samples (#364).
• Remove dill dependency, thus fixing a cloudpickle error, and run selected tests also with python 3.7 (#367).

pyABC 0.10.7

• Move progress usage to attribute level (#336).
• Add check for whether redis is up already (#337).
• Add a self-tuned finite-space discrete transition (#341).

pyABC 0.10.6

• Refactor and modularize ABCSMC inference module (#333).
• Make fast random choice function robust across dimensions (#333).

pyABC 0.10.5

• Remove bkcharts dependency (#328).
• Add optional progress bar to various samplers (#330).
• Refactor package metadata (#330).
• Refactor CI build, add code quality tests (#331).
• Add warning when many zero-prior samples are generated (#331).