Calibration and optimization tools for summer2 https://github.com/monash-emu/summer2
Estival provides a simple API for using summer2 CompartmentalModels with a variety of optimization frameworks, including
- pymc
- nevergrad
- 0.2.2
Add logprior/logposterior to BayesianCompartmentalModel - 0.2.3
Include tensorflow-probability(jax) for more (and better tested) stats modules - 0.2.4 Bugfix (vector priors were not exported to pymc correctly) Add Epoch support to allow DatetimeIndex targets
- 0.2.5 Bugfix for BinomialTarget (wasn't indexing modelled data)
- 0.2.6 Bugfix (reference index for models without date returned incorrect type)
- 0.3.0 Note - breaking changes! Remove old AuTuMN MCMC implementation Move nevergrad/pymc -> wrappers Expand likelihood output tools Include parallelism framework
- 0.3.1 Bugfix (submodules not properly exported)
- 0.3.2 SampleIterator tools (better support for shaped priors) Attempted map_parallel bugfix
- 0.3.3 Requirements fix (update summerepi2)
- 0.3.4 Swap modelled/target data in Normal and TruncatedNormal targets (incorrect results previously)
- 0.3.5 Add sampling utils Add gamma prior Support multiple targets for each derived output
- 0.3.6 Minor bugfix to 0.3.5
- 0.3.7 Reimplement BetaPrior, and get_series and finite_bounds for priors
- 0.3.8 Fix BetaPrior.to_pymc, add testing
- 0.3.9 Bugfix (BetaPrior .from_ method injected arrays into params)
- 0.4.0 Improved sampling tools
- 0.4.1 Extend sampling tools, more options for map_parallel
- 0.4.2b Experimental release using expanded transform for uniform priors
- 0.4.3 Better exec_mode defaults for parallel helper functions
- 0.4.4 Fix issues with xarray converting array parameters
- 0.4.5 Fix nevergrad wrapper issue with infinite support priors
- 0.4.8 Improved ergonomics and sample type support
- 0.4.9 Add BetaTarget
- 0.5.0 Correct loc and iloc methods for SampleIterator
- 0.5.1 Make wrapper libraries (pymc/nevergrad) optional extras