Changelog#

Versions follow Semantic Versioning (<major>.<minor>.<patch>).

Backward incompatible (breaking) changes will only be introduced in major versions with advance notice in the Deprecations section of releases.

Expert prior elicitation method v0.7.0 (2025-11-12)#

- add possibility for a dry run when initializing the Elicit object
- include in Elicit a new meta_settings parameter which includes a dry_run=True argument
adjust the summary information when printing eliobj
- include information about shape
- differentiate between target quantity and elicited summary
- correct the computation of the number of hyperparameters for parametric method
- compute number of weights (incl. biases) of NNs when using deep_prior method and include as info for computing number of hyperparameters (#30)

- add tests for elicit.py module
use Enum where appropriate to make valid options of an argument explicit. (#30)
- add tests for functions in utils.py module
discovered bug in computation of inverse_logif in DoubleBounded transformation (#38)
- outsource checks for Elicit object into new module called _checks.py
include checks when initializing the Elicit object and when using the .update method
add unittest for checks in test_init.py (#39)
- The hyperparameter names were not correctly matched with the hyperparameter values in the final output
This issue has been fixed in src\elicito\_outputs.py
A corresponding test has been added in tests\unit\test_outputs.py (#41)
- update elicito such that it is compatible with Python 3.13
This involved updating the dependencies in pyproject.toml to versions that support Python 3.13
relaxing scipy and pandas dependencies to allow for future versions
closes Issue#19 (#42)
- an error occured when using the __str__ method for the fitted eliobj
reason was a wrong assignment of the fitted results to the attribute (overwriting the self object)
fixed assignment in fit() method in __init__.py
added test to tests\unit\test_init.py::test_str_method (#43)

- add information on print method for eliobj to How-To-Guides: save-and-load (#25)

No significant changes.

No significant changes.

No significant changes.

- integrated InvertibleNetwork implementation from BayesFlow==1.1.6 into elicito.
integration has been approved by BayesFlow maintainer Stefan Radev
removal of BayesFlow dependency enabled removing version constraints (#11)

- fix check for number of model parameters. tfd.Sequential/Joint distributions were not considered in the initial check
add tensorflow-silence as one option to mute tensorflow warnings and other log messages (#7)

No significant changes.

No significant changes.

No significant changes.

No significant changes.

No significant changes.