pyPESTO v0.3.1

• Visualize:
  • Parameter plot w/ hier. pars, noise estimation for splines (#1061)
• Sampling:
  • AdaptiveMetropolis failure fix for bounded priors (#1065)
• Ensembles
  • Speed up Ensemble from History (#1063)
• PEtab support:
  • Support for petab 0.2.x (#1073)
  • Remove PetabImporterPysb #1082)
• Objective
  • AggregatedObjective: objective-specific kwargs for call_unprocessed (#1068)
• Select
  • Use predecessor stored in file (#1059)
  • support petab-select version 0.1.8 (#1070)
• Examples
  • Synthetic data: update for libpetab-python v0.2.0 (#1060)
  • Fix error in sampling_diagnostics which led to test failure(#1092)
• General
  • Test fixes (#1064)
  • Fix numpy DeprecationWarnings (#1076)
  • GHA: Fix deprecation warnings (#1075)
  • Fixed bug on existing file and no overwrite (#1046)
  • Fix error in bound checking (#1081)

pyPESTO v0.3.0

New functionalities compared to 0.2.0:

• New supported data types for parameter estimation:
  • ordinal data
  • censored data
  • unbounded parameter optimization
• New optimization approaches:
  • Hierarchical optimization
  • Spline approximation
• New optimizers: CMA-ES, Enhanced Scatter Search, Fides, NLopt, SACESS, SciPy Differential Evolution
• New samplers: Emcee, Dynesty, Pymc v4
• New Objectives: Aesara objective, Julia objective, Jax objective
• Ensemble analysis
• Model selection
• Predictions
• Hdf5 Storage

Not supported functionalities and versions compared to 0.2.0:

• Removed Python 3.8 and older support
• Pymc (v3)
• Removed Theano objective
• Changed parameter indexing from boolean to int in profiling routines

pyPESTO v0.2.17

• Optimize:
  • Parameter estimation from ordinal data (#971)
  • Parameter estimation from nonlinear-monotone data using spline approximation (#1028)
  • Parameter estimation using censored data (#1041)
  • Fix optimizer start point handling. (#1027)
  • Add option to summary to print full or reduced vectors. (#1040, #1045)
• Sampling:
  • Dynesty sampler parallelization: changed the nested loglikelihood function to a class method (#1037)
  • Dynesty sampler docs (#1039)
• Engine
  • Allow custom multiprocessing context (#1032)
• General
  • Updated example notebooks (#1050, #1026, #1051, #1056)
  • Refactor docs (#1052)
  • Update Dockerfile (#1034)
  • proper bound handling for x_guesses (#1029)
  • Updated to flake8 standards (#1042, #1049)
  • Removed Python 3.8 support according to NEP29 (#1056)

pyPESTO v0.2.16

• Optimize:
  • sacess optimizer (#988, #997)
  • Warn only once if using ineffiecient objective settings (#996)
  • Hierarchical Optimization (#1006)
  • Fix cma documentation (#987)
• Petab
  • Improvement to create_startpoint_method() (#1018)
• Sampling:
  • Dynesty sampler (#1002)
  • Fix test/sample/test_sample.py::test_samples_cis failures (#1004)
• Visualization:
  • Fix misuse of start indices in waterfall plot (#1000)
  • Fix large function values in clustering for visualizations (#999)
  • parameter correlation diverging color scheme (#1009)
  • Optimization Parameter scatter plot (#1015)
• Profiling:
  • added option to profile the whole parameter bounds. (#1014)
• General
  • Add CODEOWNERS (#1001)
  • Add list of publications using pypesto (#1008)
  • allow passing results to init of pypesto.Result (#998)
  • Updated flake8 to ignore Error B028 from bugbear until support for python 3.8 runs out. (#1005)
  • black update (#1010)
  • Doc typo fixes (#995)
  • Doc: Install amici on RTD (#1016)
  • Add getting_started notebook (#1023)
  • remove alernative formats build (#1022)

pyPESTO v0.2.15

• Optimize:
  • Add an Enhanced Scatter Search optimizer (#941, #972)
  • Cooperative enhanced scatter search (#954)
  • Hierarchical optimization (#952, #975 )
  • Allow scipy optimizer to use fun with integrated grad (#979)
• Sampling:
  • Remove fixed parameters from pymc sampling (#951)
  • emcee sampler: initialize walkers near optimum (#961)
  • dynesty Sampler (#963)
  • Fix pymc>=5 aesara/pytensor issues (#983)
• Visualization:
  • Multi-result waterfall plot (#966)
  • Model fit visualization: use problem.objective to simulate, instead of AMICI directly (#969)
  • Unfix matplotlib version (#977)
  • Plot measurements in sampling_prediction_trajectories (#976)
• Objective definition:
  • Support for jax objectives (#986)
• General
  • Fix license_file SetuptoolsDeprecationWarning (#965)
  • Remove benchmark-models-petab requirement (#964)
  • Github Actions(#958, #989 )
  • Fix typehint for problem.x_priors_defs (#962)
  • Fix tox4-related issues (#981)
  • Fix AMICI deprecation warning (#956)
  • Add pypesto.visualize.model_fit to API doc (#991)
  • Exclude numpy==1.24.0 (#993)

pyPESTO v0.2.14

• Ensembles:
  • Save and load weights and sigmay (#876)
  • Define relative cutoff (#855)
• PEtab:
  • Pass problem kwargs via petab importer (#874)
  • Use benchmark-models-petab instead of manual download (#915)
  • Use fake RData in in prediction_to_petab_measurement_df (#925)
• Optimize:
  • Fides: Include message according to exitflag (#878)
• Sampling:
  • Added Pymc v4 Sampler (#818, #944, #948)
• Visualization:
  • Fix waterfall plot limits for non-offsetted log-plots (#891)
  • Plot unflattened model fit from flattened PEtab problems (#914)
  • Added the offset value to waterfall plot for better intuitive understanding (#910, #945)
  • Visualize parameter correlation (#888)
• History and storage:
  • Fix history-result reconstruction mismatch (#902)
  • Move history to own module (#903)
  • Remove chi2, schi2 except for history convenience function (#904)
  • Clean up history hierarchy (#908)
  • Fix read_result with history (#907)
  • Improve hdf5 history file lock (#909, #921)
  • Fix message in check_overwrite (#894)
  • Deactivate automatic saving (#930, #932)
  • Allow problem=None in read_result_from_file (#936)
  • Remove superfluous get_or_create_group (#937)
  • Extract read_history_from_file from read_result_from_file (#939)
  • Select: use model ID in save postprocessor filename, by default (#943)
• Select:
  • Clean up use of minimize_options in model problem (#918)
  • User-supplied method to produce pyPESTO problem (#884)
  • Report, and binary model ID post-processors (#900)
  • Move method.py functionalities to ui.py in petab_select (#919)
• Objective and Result:
  • Julia objective (#927)
  • Fix set of keys to aggregate results in aggregated objective (#883)
  • Nicer OptimizeResult.summary (#895, #916, #935, #942, )
  • Fix disjoint IDs check in OptimizerResult.append (#922)
  • Fix OptimizeResult pickling (#953)
• General:
  • Remove version from CITATION.cff (#887)
  • Fix CI and docs (#892, #893)
  • Literal typehints for mode (#899)
  • Fix pandas deprecation warning (#896)
  • Document NEP 29 (time-window based python support) (#905)
  • Fix get_for_key deprecation warning (#906)
  • Fix multiple warnings from existing AMICI model (#912)
  • Fix warning from AMICI fixed overrides (#912)
  • Fix flaky test CRFunModeHistoryTest.test_trace_all (#917)
  • Fix novel B024 ABC without abstract methods (#923)
  • Improve API docs and add overview notebook (#911)
  • Fix typos (#926)
  • Fix julia tests (#929, #933)
  • Fix flaky test_mpipoolengine (#938)
  • More informative test IDs in test_optimize (#940)
  • Speed-up import via lazy imports (#946)

pyPESTO v0.2.13

• Ensembles:
  • Added standard deviation to ensemble prediction plots (#853)
• Storage
  • Distinguish between scalar and vector values in Hdf5History._get_hdf5_entries (#856)
  • Fix hdf5 history overwrite (#861)
  • Updated optimization storage format. Made attributes explicit. (#863)
  • Added problem to result from read_results_from_file (#862)
• General
  • Various additions to Optimize(r)Result summary method (#859, #865, #866, #867)
  • Fixed optimizer history fval offset (#834)
  • Updated the profile, minimize, sample and added overwrite as argument. (#864)
  • Fixed y-labels in pypesto.visualize.optimizer_history (#869)
  • Created show_bounds, to display proper sampling scatter plots. (#868)
  • Enabled saving messages and exit flags in hdf5 history in case of finished run (#873)
  • Select: use objective function evaluation time as optimization time for models with no estimated parameters (#872)
  • removed checking for equality and checking for np.allclose in `test_aesara (#877)

pyPESTO v0.2.12

• AMICI:
  • Update to renamed steady state sensitivity modes (#843)
  • Set amici.Solver.setReturnDataReportingMode (#835)
  • Optimize pypesto/objective/amici_util.py::par_index_slices(#845)
  • Remove Solver.getPreequilibration (#830)
  • fix n_res size for error output with parameter dependent sigma (#812)
  • PetabImporter: Auto-regenerate AMICI models in case of version mismatch (#848 )
• Pymc3
  • Disable Pymc3 Sampler tests (#831 )
• Visualizations:
  • Waterfall zoom (#808 )
  • Reverse opacities of colors in prediction trajectories plots(#838)
  • Model fit plots (#850)
• OptimizeResult:
  • Summary method (#816 )
  • Append method for OptimizeResult (#815 )
  • added getattr function to OptimizeResult (#802 )
• General:
  • disable progress bar in tests (#799 )
  • Make Fides work with objectives, that do not have a hessian (#807 )
  • removed ftol in favor of tol (#803 )
  • Fix pyPESTO Select test; Update to stable black version (#810 )
  • Fix id assignment in case of large number of starts (#825 )
  • Temporarily fix jinja2 version (#826)
  • Upgrade black to be compatible with latest click (#829)
  • Fix wrong link in doc/example/hdf5_storage.ipynb (#827)
  • Mark test/base/test_prior.py::test_mode as flaky (#833)
  • Custom methods for autosave filenames (#822)
  • fix saving ensemble predictions to hdf5 (#840 )
  • Upgrad nbQA to 1.3.1 (#846)
  • removed constantParameters for constant_parameters in notebook (#852)

pyPESTO 0.2.11

• Model selection (#397):

  • Automated model selection with forward/backward/brute force methods and
    AIC/AICc/BIC criteria
  • Much functionality (methods, criteria, model space, problem
    specification) via PEtab Select <https://github.com/PEtab-dev/petab_select>
  • Plotting routines
  • Example notebook <https://github.com/ICB-DCM/pyPESTO/blob/main/doc/example/model_selection.ipynb>
  • Model calibration postprocessors
  • Select first model that improves on predecessor model
  • Use previous MLE as startpoint
  • Tests

• AMICI:

  • Maintain model settings when pickling for multiprocessing (#747)

• General:

  • Apply nbqa black and isort to auto-format all notebooks via
    pre-commit hook (#794)
  • Apply black formatting via pre-commit hook (#796)
  • Require Python >= 3.8 (#795)
  • Fix various warnings (#778)
  • Minor fixes (#792)

pyPESTO 0.2.10

• AMICI:

  • Make AMICI objective report only what is being asked for (#777)

• Optimization:

  • (Breaking) Refactor startpoint generation with clear assignments;
    allow checking gradients (#769)
  • (Breaking) Prioritize history vs optimize result (#775)

• Storage:

  • Fix loading empty history and result generation from multiple
    histories (#764)
  • Fix autosave function for single-core (#770)
  • Fix potential autosave overwriting and typehints (#772)
  • Allow loading of partial results from history file (#783)

• CI:

  • Compile AMICI models without gradients in test suite (#774)

• General:

  • (Breaking) Create result sub-module; shift storage+result related
    functionality (#784)
  • Fix finite difference constant mode (#786)
  • Refactor ensemble module (#788)
  • Introduce general C constants file (#788)
  • Apply isort for automatic imports formatting (#785)
  • Reduce run log output (#789)
  • Various minor fixes (#765, #766, #768, #771)