Communicating Likelihoods with Normalising Flows

Jack Y. Araz, Anja Beck, Méril Reboud, Michael Spannowsky, Danny van Dyk
2/13/2025
hep-phcs.LGhep-ex

No official code found. Best implementation: nabu-hep/nabu-hep (0 stars, MIT, last commit 2025-09-30).

Abstract

We present a machine-learning-based workflow to model an unbinned likelihood from its samples. A key advancement over existing approaches is the validation of the learned likelihood using rigorous statistical tests of the joint distribution, such as the Kolmogorov-Smirnov test of the joint distribution. Our method enables the reliable communication of experimental and phenomenological likelihoods for subsequent analyses. We demonstrate its effectiveness through three case studies in high-energy physics. To support broader adoption, we provide an open-source reference implementation, nabu.

View on arXivView PDF

Code Implementations(1)

nabu-hep/nabu-hepCommunity100%
MIT

Communicating Likelihoods with Normalising Flows

01PythonOct 29, 20241 years agoMIT
likelihoodnormalising-flows

Similar Papers

Flows for Flows: Morphing one Dataset into another with Maximum Likelihood Estimation

84% match2023

Testable Likelihoods for Beyond-the-Standard Model Fits

83% match2023

Contrastive Normalizing Flows for Uncertainty-Aware Parameter Estimation

80% match2025

A Step Toward Interpretability: Smearing the Likelihood

80% match2025

Normalizing Flow Regression for Bayesian Inference with Offline Likelihood Evaluations

79% match1 repo32025

Simulation-based inference for Precision Neutrino Physics through Neural Monte Carlo tuning

79% match1 repo12025

Cite this paper

@article{araz2025communicating,
  title  = {Communicating Likelihoods with Normalising Flows},
  author = {Jack Y. Araz and Anja Beck and Méril Reboud and Michael Spannowsky and Danny van Dyk},
  year   = {2025},
  eprint = {2502.09494},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2502.09494v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2502.09494.svg)](https://codeofpaper.com/paper/2502.09494)

Discussion