All notable changes to this project will be documented in this file.
- Feature X
- Feature Y
- Refactored module A to improve performance
- Bug in module B that caused a crash
- Raised required amici version to 0.26.1
- Ignoring sampler warnings caused by parameters outside the parameter limits
- Limited the parameter range and increased the solver accuracy of the Corona model to avoid warnings
- Seed numpy random number generator for reproducibility of pytest runs
- Raised the required python version to 3.10
- Removed upper limit on the version number of dependencies
- Updated amici to newest version to fix compilation with Intel MKL
- Added
PyQt6
as dev dependency to allow developers to view plots in the development environment
- Removed the redundant function
evaluate_sample
fromsampling.py
- Switched from
importlib.resources.path
toimportlib.resources.files
due to deprecation - Removed argument name in call to
files(...)
that is changing frompackage
toanchor
- Removed unused kwargs passed to the plotting function
contour
that generated a warning intest_heat_model()
- Bug that changed the model attribute
central_param
duringinference
- Basic plotting function for sample results
- Users can now check the models they want to use with a basic functionality check and a quick inference check on artificial data.
- Added code usage examples to documentation.
- Fixed deprecated jax.config import
Breaking changes: The stock model is removed from the available example models! You can copy the model code and the data files from a previous version and import it as custom model, if you need it.
- Removed the stock model, because it does not fulfill the requirements for the EPI method
- Minor dependency updates
- Added link to changelog on pypi
Breaking changes: SBML support is now only available when install eulerpi with the "sbml" extra!
- Updated the dependencies to the latest possible versions
- Made the sbml model support optional using the extra
sbml
- Fixed poetry.lock containing faulty version 1.6.8 of debugpy
- Fixed github actions failing due to missing
sudo apt update
- Added support for for sbml models to select parameters and species used in the inference
- Added support for evaluating sbml models at multiple time points
- Switched from using parameter names to using parameter ids in the sbml model
- Change spatial discretization of heat model example to proper second order central differences
- Change the initial walker position generation of the emcee sampler
- Added a new example model, a 2d heat conduction equation
- Added a function to model to specify more complex parameter domains
- Removed old functions from corona model
- Temporarily removed stock and sbml models from test examples
- Updated tutorial to be consistent with current state of the project
- Fixed bug in dense grid generation causing parameter limits to be applied inconsistently when using slices
- Bug in result manager causing burn-in and thinning to be performed on the wrong samples.
- Bug in result manager that caused density evals to be saved as data samples for non-full slices.
- Added data normalization to improve numerics and performance in the inference method.
- Added possiblity to use no PCA transformation or completely custom transformation for inference.
- More tests to validate EPI
- Fixed grid results beeing unusable due to numerical issues
- Fixed sbml models ignoring parameter_names argument
- Fix sparse grid returning meaningless data
- Fix menten data shape
- Fix SBMLModel data_dim: Return states count instead of observables count
- Validating the shapes of the inference results
- Now using default emcee move policy during MCMC sampling to improve convergence speed
- Changed argument calc_walker_acceptance_bool to get_walker_acceptance
- Changed meaning of argument num_burn_in_samples to refer to the number of samples burned per chain, not in total
- Fixed SwigPyObject pickling
- Fixed dead lock for larger data sets and models
- ResultManager is saving meta data for inference
- ResultManager uses saved meta data to load inference results
- Renamed epipy to eulerpi
- Updated installation instructions
- Updated dependencies
- removed
__pycache__
,build
and.so
from pypi
- fixed epi logo not showing on pypi
- Initial release