relaxed

Provides differentiable ("relaxed") versions of common operations in high-energy physics. Based on jax. Where possible, function APIs try to mimic their commonly used counterparts, e.g. fitting and hypothesis testing in pyhf.

Currently implemented:

• relaxed.hist: histograms via kernel density estimation
• fitting routines:
  • relaxed.mle.fit: global MLE fit
  • relaxed.mle.fixed_poi_fit: constrained fit given a value of a parameter of interest
• relaxed.infer.hypotest: hypothesis test using the profile likelihood as a test statistic
• relaxed.fisher_info: the fisher information matrix (of a pyhf-type model)
• relaxed.cramer_rao_uncert: inverts the fisher information matrix to provide uncertainties valid through the Cramér-Rao bound
• relaxed.gaussianity: an experimental metric that quantifies the mean-squared difference of a likelihood function with respect to its gaussian approximation (covariance calculated using the Cramér-Rao bound above)

Will implement:

• smooth cuts via sigmoid
• your favourite idea here!
  • we're maintaining a list of desired differentiable operations in list_of_operations.md (thanks to @cranmer) -- feel free to take inspiration or contribute :)

install

pip install relaxed

For use with pyhf, e.g. in a neos-type workflow, it is temporarily recommended to install pyhf using a specific branch that is designed to be differentiable with respect to model construction:

pip install git+http://github.com/scikit-hep/pyhf.git@make_difffable_model_ctor

We plan to merge this into pyhf when it's stable, and will then drop this instruction :)