Saved in:
Bibliographic Details
Main Authors: Bieringer, Sebastian, Diefenbacher, Sascha, Kasieczka, Gregor, Trabs, Mathias
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.00838
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915026970869760
author Bieringer, Sebastian
Diefenbacher, Sascha
Kasieczka, Gregor
Trabs, Mathias
author_facet Bieringer, Sebastian
Diefenbacher, Sascha
Kasieczka, Gregor
Trabs, Mathias
contents Recently, combinations of generative and Bayesian machine learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.
format Preprint
id arxiv_https___arxiv_org_abs_2408_00838
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Calibrating Bayesian Generative Machine Learning for Bayesiamplification
Bieringer, Sebastian
Diefenbacher, Sascha
Kasieczka, Gregor
Trabs, Mathias
Machine Learning
Artificial Intelligence
High Energy Physics - Phenomenology
Recently, combinations of generative and Bayesian machine learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.
title Calibrating Bayesian Generative Machine Learning for Bayesiamplification
topic Machine Learning
Artificial Intelligence
High Energy Physics - Phenomenology
url https://arxiv.org/abs/2408.00838