pyhf: pure-Python implementation of HistFactory with tensors and automatic differentiation

TL;DR

pyhf is a Python implementation of the HistFactory statistical model that leverages modern tensor libraries and auto-differentiation for flexible, efficient, and GPU-accelerated high-energy physics data analysis outside of ROOT.

Contribution

It introduces a pure-Python, tensor-compatible implementation of HistFactory with JSON serialization, enabling broader and more efficient statistical analysis in high-energy physics.

Findings

Supports TensorFlow, PyTorch, JAX for auto-differentiation and GPU use

Published 23 probability models from ATLAS analyses on HEPData

Enables analysis outside of ROOT framework

Abstract

The HistFactory p.d.f. template is per-se independent of its implementation in ROOT and it is useful to be able to run statistical analysis outside of the ROOT, RooFit, RooStats framework. pyhf is a pure-Python implementation of that statistical model for multi-bin histogram-based analysis and its interval estimation is based on the asymptotic formulas of "Asymptotic formulae for likelihood-based tests of new physics". pyhf supports modern computational graph libraries such as TensorFlow, PyTorch, and JAX in order to make use of features such as auto-differentiation and GPU acceleration. In addition, pyhf's JSON serialization specification for HistFactory models has been used to publish 23 full probability models from published ATLAS collaboration analyses to HEPData.