Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains

Non-Gaussian distributions in cosmology are commonly evaluated with Monte Carlo Markov-chain methods, as the Fisher-matrix formalism is restricted to the Gaussian case. The Metropolis-Hastings algorithm will provide samples from the posterior distribution after a burn-in period, and the correspondin...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Röver, Lennart, von Campe, Heinrich, Herzog, Maximilian Philipp, Kuntz, Rebecca Maria, Schäfer, Björn Malte
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Röver, Lennart
von Campe, Heinrich
Herzog, Maximilian Philipp
Kuntz, Rebecca Maria
Schäfer, Björn Malte
description Non-Gaussian distributions in cosmology are commonly evaluated with Monte Carlo Markov-chain methods, as the Fisher-matrix formalism is restricted to the Gaussian case. The Metropolis-Hastings algorithm will provide samples from the posterior distribution after a burn-in period, and the corresponding convergence is usually quantified with the Gelman-Rubin criterion. In this paper, we investigate the convergence of the Metropolis-Hastings algorithm by drawing analogies to statistical Hamiltonian systems in thermal equilibrium for which a canonical partition sum exists. Specifically, we quantify virialisation, equipartition and thermalisation of Hamiltonian Monte Carlo Markov-chains for a toy-model and for the likelihood evaluation for a simple dark energy model constructed from supernova data. We follow the convergence of these criteria to the values expected in thermal equilibrium, in comparison to the Gelman-Rubin criterion. We find that there is a much larger class of physically motivated convergence criteria with clearly defined target values indicating convergence. As a numerical tool, we employ physics-informed neural networks for speeding up the sampling process.
doi_str_mv 10.48550/arxiv.2305.07061
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2305_07061</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2305_07061</sourcerecordid><originalsourceid>FETCH-LOGICAL-a671-6403604104b0057a50c923933e10d6854d4f8836b706810fae3d4597ddddb1ed3</originalsourceid><addsrcrecordid>eNotz71OwzAUBWAvDKjwAEz4BRKu658kbKiCglQEQ_foxnaI1dSObDeib08JnOWc6UgfIXcMSlFLCQ8Yv91crjnIEipQ7JrkT4zZZRc87U9eLwOnKQbUA82B-uCLLZ5Scujp6A52dEMIJj3SaTgnp3Ecz_QYspsxW0N18LONX9ZrS3V02UaHtA-RvmM8hLnQAzqfbshVj2Oyt_-9IvuX5_3mtdh9bN82T7sCVcUKJYArEAxEByArlKCbNW84twyMqqUwoq9rrrqLpGbQo-VGyKYyl3TMGr4i93-3C7udojtiPLe__Hbh8x9JcVdu</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains</title><source>arXiv.org</source><creator>Röver, Lennart ; von Campe, Heinrich ; Herzog, Maximilian Philipp ; Kuntz, Rebecca Maria ; Schäfer, Björn Malte</creator><creatorcontrib>Röver, Lennart ; von Campe, Heinrich ; Herzog, Maximilian Philipp ; Kuntz, Rebecca Maria ; Schäfer, Björn Malte</creatorcontrib><description>Non-Gaussian distributions in cosmology are commonly evaluated with Monte Carlo Markov-chain methods, as the Fisher-matrix formalism is restricted to the Gaussian case. The Metropolis-Hastings algorithm will provide samples from the posterior distribution after a burn-in period, and the corresponding convergence is usually quantified with the Gelman-Rubin criterion. In this paper, we investigate the convergence of the Metropolis-Hastings algorithm by drawing analogies to statistical Hamiltonian systems in thermal equilibrium for which a canonical partition sum exists. Specifically, we quantify virialisation, equipartition and thermalisation of Hamiltonian Monte Carlo Markov-chains for a toy-model and for the likelihood evaluation for a simple dark energy model constructed from supernova data. We follow the convergence of these criteria to the values expected in thermal equilibrium, in comparison to the Gelman-Rubin criterion. We find that there is a much larger class of physically motivated convergence criteria with clearly defined target values indicating convergence. As a numerical tool, we employ physics-informed neural networks for speeding up the sampling process.</description><identifier>DOI: 10.48550/arxiv.2305.07061</identifier><language>eng</language><subject>Physics - Cosmology and Nongalactic Astrophysics ; Physics - Data Analysis, Statistics and Probability ; Physics - Instrumentation and Methods for Astrophysics</subject><creationdate>2023-05</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2305.07061$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2305.07061$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Röver, Lennart</creatorcontrib><creatorcontrib>von Campe, Heinrich</creatorcontrib><creatorcontrib>Herzog, Maximilian Philipp</creatorcontrib><creatorcontrib>Kuntz, Rebecca Maria</creatorcontrib><creatorcontrib>Schäfer, Björn Malte</creatorcontrib><title>Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains</title><description>Non-Gaussian distributions in cosmology are commonly evaluated with Monte Carlo Markov-chain methods, as the Fisher-matrix formalism is restricted to the Gaussian case. The Metropolis-Hastings algorithm will provide samples from the posterior distribution after a burn-in period, and the corresponding convergence is usually quantified with the Gelman-Rubin criterion. In this paper, we investigate the convergence of the Metropolis-Hastings algorithm by drawing analogies to statistical Hamiltonian systems in thermal equilibrium for which a canonical partition sum exists. Specifically, we quantify virialisation, equipartition and thermalisation of Hamiltonian Monte Carlo Markov-chains for a toy-model and for the likelihood evaluation for a simple dark energy model constructed from supernova data. We follow the convergence of these criteria to the values expected in thermal equilibrium, in comparison to the Gelman-Rubin criterion. We find that there is a much larger class of physically motivated convergence criteria with clearly defined target values indicating convergence. As a numerical tool, we employ physics-informed neural networks for speeding up the sampling process.</description><subject>Physics - Cosmology and Nongalactic Astrophysics</subject><subject>Physics - Data Analysis, Statistics and Probability</subject><subject>Physics - Instrumentation and Methods for Astrophysics</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz71OwzAUBWAvDKjwAEz4BRKu658kbKiCglQEQ_foxnaI1dSObDeib08JnOWc6UgfIXcMSlFLCQ8Yv91crjnIEipQ7JrkT4zZZRc87U9eLwOnKQbUA82B-uCLLZ5Scujp6A52dEMIJj3SaTgnp3Ecz_QYspsxW0N18LONX9ZrS3V02UaHtA-RvmM8hLnQAzqfbshVj2Oyt_-9IvuX5_3mtdh9bN82T7sCVcUKJYArEAxEByArlKCbNW84twyMqqUwoq9rrrqLpGbQo-VGyKYyl3TMGr4i93-3C7udojtiPLe__Hbh8x9JcVdu</recordid><startdate>20230511</startdate><enddate>20230511</enddate><creator>Röver, Lennart</creator><creator>von Campe, Heinrich</creator><creator>Herzog, Maximilian Philipp</creator><creator>Kuntz, Rebecca Maria</creator><creator>Schäfer, Björn Malte</creator><scope>GOX</scope></search><sort><creationdate>20230511</creationdate><title>Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains</title><author>Röver, Lennart ; von Campe, Heinrich ; Herzog, Maximilian Philipp ; Kuntz, Rebecca Maria ; Schäfer, Björn Malte</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a671-6403604104b0057a50c923933e10d6854d4f8836b706810fae3d4597ddddb1ed3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Physics - Cosmology and Nongalactic Astrophysics</topic><topic>Physics - Data Analysis, Statistics and Probability</topic><topic>Physics - Instrumentation and Methods for Astrophysics</topic><toplevel>online_resources</toplevel><creatorcontrib>Röver, Lennart</creatorcontrib><creatorcontrib>von Campe, Heinrich</creatorcontrib><creatorcontrib>Herzog, Maximilian Philipp</creatorcontrib><creatorcontrib>Kuntz, Rebecca Maria</creatorcontrib><creatorcontrib>Schäfer, Björn Malte</creatorcontrib><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Röver, Lennart</au><au>von Campe, Heinrich</au><au>Herzog, Maximilian Philipp</au><au>Kuntz, Rebecca Maria</au><au>Schäfer, Björn Malte</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains</atitle><date>2023-05-11</date><risdate>2023</risdate><abstract>Non-Gaussian distributions in cosmology are commonly evaluated with Monte Carlo Markov-chain methods, as the Fisher-matrix formalism is restricted to the Gaussian case. The Metropolis-Hastings algorithm will provide samples from the posterior distribution after a burn-in period, and the corresponding convergence is usually quantified with the Gelman-Rubin criterion. In this paper, we investigate the convergence of the Metropolis-Hastings algorithm by drawing analogies to statistical Hamiltonian systems in thermal equilibrium for which a canonical partition sum exists. Specifically, we quantify virialisation, equipartition and thermalisation of Hamiltonian Monte Carlo Markov-chains for a toy-model and for the likelihood evaluation for a simple dark energy model constructed from supernova data. We follow the convergence of these criteria to the values expected in thermal equilibrium, in comparison to the Gelman-Rubin criterion. We find that there is a much larger class of physically motivated convergence criteria with clearly defined target values indicating convergence. As a numerical tool, we employ physics-informed neural networks for speeding up the sampling process.</abstract><doi>10.48550/arxiv.2305.07061</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2305.07061
ispartof
issn
language eng
recordid cdi_arxiv_primary_2305_07061
source arXiv.org
subjects Physics - Cosmology and Nongalactic Astrophysics
Physics - Data Analysis, Statistics and Probability
Physics - Instrumentation and Methods for Astrophysics
title Partition function approach to non-Gaussian likelihoods: physically motivated convergence criteria for Markov-chains
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-23T16%3A41%3A05IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Partition%20function%20approach%20to%20non-Gaussian%20likelihoods:%20physically%20motivated%20convergence%20criteria%20for%20Markov-chains&rft.au=R%C3%B6ver,%20Lennart&rft.date=2023-05-11&rft_id=info:doi/10.48550/arxiv.2305.07061&rft_dat=%3Carxiv_GOX%3E2305_07061%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true