Discriminative versus generative approaches to simulation-based inference

Most of the fundamental, emergent, and phenomenological parameters of particle and nuclear physics are determined through parametric template fits. Simulations are used to populate histograms which are then matched to data. This approach is inherently lossy, since histograms are binned and low-dimen...

Full description

Saved in:
Bibliographic Details
Main Authors: Benjamin Sluijter, Sascha Diefenbacher, Wahid Bhimji, Benjamin Nachman
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:Machine Learning: Science and Technology
Subjects:
Online Access:https://doi.org/10.1088/2632-2153/adf68b
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849339121972019200
author Benjamin Sluijter
Sascha Diefenbacher
Wahid Bhimji
Benjamin Nachman
author_facet Benjamin Sluijter
Sascha Diefenbacher
Wahid Bhimji
Benjamin Nachman
author_sort Benjamin Sluijter
collection DOAJ
description Most of the fundamental, emergent, and phenomenological parameters of particle and nuclear physics are determined through parametric template fits. Simulations are used to populate histograms which are then matched to data. This approach is inherently lossy, since histograms are binned and low-dimensional. Deep learning has enabled unbinned and high-dimensional parameter estimation through neural likelihood(-ratio) estimation. We compare two approaches for neural simulation-based inference (NSBI): one based on discriminative learning (classification) and one based on generative modeling. These two approaches are directly evaluated on the same datasets, with a similar level of hyperparameter optimization in both cases. In addition to a Gaussian dataset, we study NSBI using a Higgs boson dataset from the FAIR Universe Challenge. We find that both the direct likelihood and likelihood ratio estimation are able to effectively extract parameters with reasonable uncertainties. For the numerical examples and within the set of hyperparameters studied, we found that the likelihood ratio method is more accurate and/or precise. Both methods have a significant spread from the network training and would require ensembling or other mitigation strategies in practice.
format Article
id doaj-art-2403e5b1dae1463993932faa75b6ae71
institution Kabale University
issn 2632-2153
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series Machine Learning: Science and Technology
spelling doaj-art-2403e5b1dae1463993932faa75b6ae712025-08-20T03:44:13ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016303050310.1088/2632-2153/adf68bDiscriminative versus generative approaches to simulation-based inferenceBenjamin Sluijter0https://orcid.org/0009-0009-6830-9016Sascha Diefenbacher1https://orcid.org/0000-0003-4308-6804Wahid Bhimji2https://orcid.org/0000-0002-6213-8617Benjamin Nachman3https://orcid.org/0000-0003-1024-0932Leiden Institute of Physics, Universiteit Leiden , Leiden, RA 2300, The Netherlands; Physics Division, Lawrence Berkeley National Laboratory , Berkeley, CA 94720, United States of AmericaPhysics Division, Lawrence Berkeley National Laboratory , Berkeley, CA 94720, United States of AmericaNational Energy Research Scientific Computing Center, Berkeley Lab , Berkeley, CA 94720, United States of AmericaPhysics Division, Lawrence Berkeley National Laboratory , Berkeley, CA 94720, United States of America; Fundamental Physics Directorate, SLAC National Accelerator Laboratory , Menlo Park, CA 94025, United States of America; Department of Particle Physics and Astrophysics, Stanford University , Stanford, CA 94305, United States of AmericaMost of the fundamental, emergent, and phenomenological parameters of particle and nuclear physics are determined through parametric template fits. Simulations are used to populate histograms which are then matched to data. This approach is inherently lossy, since histograms are binned and low-dimensional. Deep learning has enabled unbinned and high-dimensional parameter estimation through neural likelihood(-ratio) estimation. We compare two approaches for neural simulation-based inference (NSBI): one based on discriminative learning (classification) and one based on generative modeling. These two approaches are directly evaluated on the same datasets, with a similar level of hyperparameter optimization in both cases. In addition to a Gaussian dataset, we study NSBI using a Higgs boson dataset from the FAIR Universe Challenge. We find that both the direct likelihood and likelihood ratio estimation are able to effectively extract parameters with reasonable uncertainties. For the numerical examples and within the set of hyperparameters studied, we found that the likelihood ratio method is more accurate and/or precise. Both methods have a significant spread from the network training and would require ensembling or other mitigation strategies in practice.https://doi.org/10.1088/2632-2153/adf68bmachine learninguncertainty quantificationlikelihood estimationparticle physicshigh energy physics
spellingShingle Benjamin Sluijter
Sascha Diefenbacher
Wahid Bhimji
Benjamin Nachman
Discriminative versus generative approaches to simulation-based inference
Machine Learning: Science and Technology
machine learning
uncertainty quantification
likelihood estimation
particle physics
high energy physics
title Discriminative versus generative approaches to simulation-based inference
title_full Discriminative versus generative approaches to simulation-based inference
title_fullStr Discriminative versus generative approaches to simulation-based inference
title_full_unstemmed Discriminative versus generative approaches to simulation-based inference
title_short Discriminative versus generative approaches to simulation-based inference
title_sort discriminative versus generative approaches to simulation based inference
topic machine learning
uncertainty quantification
likelihood estimation
particle physics
high energy physics
url https://doi.org/10.1088/2632-2153/adf68b
work_keys_str_mv AT benjaminsluijter discriminativeversusgenerativeapproachestosimulationbasedinference
AT saschadiefenbacher discriminativeversusgenerativeapproachestosimulationbasedinference
AT wahidbhimji discriminativeversusgenerativeapproachestosimulationbasedinference
AT benjaminnachman discriminativeversusgenerativeapproachestosimulationbasedinference