parPE 0.4.3

Library

• Update to current PEtab
• Save more metadata in simulator result file (#259)
• More informative output for hierarchical optimization (Closes #210) (#255)
• Update to AMICI v0.10.21 and petab 0.1.7 (#261)
• Fix: Handle paths with blanks in build scripts (Fixes #246)
• Remove old example models lucarelli and raia (Closes #215)
• Misc fixups
• Cleanup build system, e.g. replace redundant CMake option BUILD_TESTS by auto-provided BUILD_TESTING

Documentation

• Update documentation, now created with sphinx/breeze/exhale and available at https://readthedocs.org/projects/parpe/ (#249) (#251)

CI

• Use Github actions instead of shippable (#256)
• Set up sonarcloud for static code analysis (#254)
• Various fixes (#258) (#236)
• Better integration of google tests in ctest using gtest_discover_tests. Raise CMake requirement. (#253)
• Run PEtab test suite (#244)
• Run python package tests via CTest
• Tests with steadystate example to pytest

parPE 0.4.2

• Improve parallel scaling by better message handling
• Remove unreliable auto-detection of launch via mpi launcher, use --mpi instead
• Allow setting integration retries / tolerance relaxation via environment variables PARPE_NUM_SIMULATION_TRIALS and PARPE_INTEGRATION_TOLERANCE_RELAXATION_FACTOR
• CMake: enable use of PACKAGE_ROOT
• Minor fixes

parPE 0.4.1

• Update AMICI and disentangle parallel AMICI output during optimization and simulation (#204)
• Autodeploy to dockerhub using github action (#205)
• Improvements minibatch (#206) (https://www.biorxiv.org/content/10.1101/859884v1)
• implemented changes from supermuc with fixes in minibatch rescue interceptor and line search algorithm
• Use PEtab parameter prior for starting point sampling (#194)

Minor:

• Explicitly enable oversubscription since we are using OpenMPI
• Disentangle parallel AMICI output during optimization (Closes #202)
• Update list of references
• Fix model naming
• Fix docker MPI issues
  • due to lack of SYS_PTRACE permissions (disable Vader) (Closes #185)
  • Disable Infiniband for OpenMPI
  • enable tcp via loopback
• Set RelWithDebInfo as default build type
• Update AMICI to v0.10.16 and PEtab v0.0.0a17 (#209)
• Disable saving states and observables during optimization, as the current format does not support varying number of timepoints and crashes
• Allow creation of virtual environment in the absence of ensurepip
• Fix uninitialized usage warnings
• Add Zenodo badge (Closes #189)

parPE 0.4.0

• Enabled Zenodo integration
• Update AMICI to 0.10.13
• Setup docker image for use with charliecloud (#182)
• Cleanup misc/generateHDF5DataFileFromText.py, remove obsolete warning
• Formatting
• Update reference; add funding section to README
• Add parPE logo (#195)
• Show only interesting lines from /proc/self/status
• Update FAQ
• Disable building tests by default
• CLI option to show supported optimizers (Closes #169)
• Enable CLI starting point index selection for OptimizationApplication
• Add script to read final parameter vectors from parPE (multi-start) optimization result file and write to TSV file
• Add AMICI version and script invocation to generated hdf5 file
• Add gradient check to snakefile
• Fix: Ensure rdata->numstepsB has elements before access (segfault with forward sensitivities)
• Fix parameter indexing for generating example data
• Add doc/README.md
• Document HDF5 output file format (Closes #175)
• Add worfklow figure
• Add ../doc/petab_model_import.md describing PEtab model import (Closes #168)
• Pass H5File as const&, not as Id which will implicitely call H5File which requires HDF5 mutex locking.
• Extend, split, update steadystate example notebooks (#171)
• Ensure hdf5 mutex lock when using AMICI hdf5 helper functions (Closes #167)
• Keep fixed parameters out of /parameters/parameterNames in HDF5 file, which should only contain optimization parameters
• Update readme: Add coinhsl instructions
• Fix HDF5 file generation after PEtab update: do not add AMICI fixed parameters to optmization parameters
• Add scaffold for generation of LoadLeveler job files
• Add snakemake scaffold for PEtab...analysis
• Update PEtab parameter mapping to allow for condition specific scaling etc.
• Update benchmark collection scripts, add Snakefile
• Add basic data analysis Snakefile
• Update plotting functions and example notebook
• Use condition-specific parameter scales (#161)
• Remove obsolete test files
• Remove cpputest remnants
• Update Ipopt install script to 3.12.12, cleanup
• Remove work-around for non-threadsafe adjoints
• Gradient check tolerance relative to objective function values to reduce number of test failures (Closes #80)
• Reenable CERES parameter_tolerance option
• PEtab: Add support for non-constant parameter overrides in condition table (Closes #158); support non-gaussian noise
• Use HDF5 C++ interface; cleanup.

v0.3.2

Changes to previous version:

• Extended and updates documentation
• Extended and updates example notebooks
• Fixed parallel HDF5 file access bugs
• Fixed outdated paths

v0.3.1

• PEtab: Add support for non-constant parameter overrides in condition table (Closes #158); support non-gaussian noise
• Updated minibatch optimizer
• Updated AMICI (now with sundials 4.xx)
• Support for shared-memory parallel adjoint sensitivity analysis
• Replace cpputest by googletest
• Add tests
• Various bugfixes
• Refactoring, cleanup

v0.3.0

Change pipeline to use PEtab data format.

Added support for preequilibration, replicates, numeric parameter overrides, fixed parameters.

Fixed various smaller issues. Cleanup. Documentation.

parPE 0.2.0

• Added mini-batch optimizer
• ...
• Updated AMICI to 0.8.2
• Automated generation of example code (now requires AMICI Python package)
• Major cleanup