Loading…
Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution
•Monte Carlo algorithm is used to evaluate simultaneously the observable and its Jacobian.•Null-collision algorithm (NCA) are very successful when dealing with linear-physics in heterogeneous media.•We showed that NCA are sources of convergence difficulties to sensitivity-evaluations of optical para...
Saved in:
Published in: | Journal of computational physics 2020-07, Vol.413, p.109463-20/109463, Article 109463 |
---|---|
Main Authors: | , , , , , , , |
Format: | Article |
Language: | English |
Subjects: | |
Citations: | Items that this one cites Items that cite this one |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
cited_by | cdi_FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3 |
---|---|
cites | cdi_FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3 |
container_end_page | 20/109463 |
container_issue | |
container_start_page | 109463 |
container_title | Journal of computational physics |
container_volume | 413 |
creator | Tregan, J.-M. Blanco, S. Dauchet, J. El Hafi, M. Fournier, R. Ibarrart, L. Lapeyre, P. Villefranque, N. |
description | •Monte Carlo algorithm is used to evaluate simultaneously the observable and its Jacobian.•Null-collision algorithm (NCA) are very successful when dealing with linear-physics in heterogeneous media.•We showed that NCA are sources of convergence difficulties to sensitivity-evaluations of optical parameter.•We propose an alternative solution.
When a Monte Carlo algorithm is used to evaluate a physical observable A, it is possible to slightly modify the algorithm so that it evaluates simultaneously A and the derivatives ∂ςA of A with respect to each problem-parameter ς. The principle is the following: Monte Carlo considers A as the expectation of a random variable, this expectation is an integral, this integral can be derivated as function of the problem-parameter to give a new integral, and this new integral can in turn be evaluated using Monte Carlo. The two Monte Carlo computations (of A and ∂ςA) are simultaneous when they make use of the same random samples, i.e. when the two integrals have the exact same structure. It was proven theoretically that this was always possible, but nothing ensures that the two estimators have the same convergence properties: even when a large enough sample-size is used so that A is evaluated very accurately, the evaluation of ∂ςA using the same sample can remain inaccurate. We discuss here such a pathological example: null-collision algorithms are very successful when dealing with radiative transfer in heterogeneous media, but they are sources of convergence difficulties as soon as sensitivity-evaluations are considered. We analyse theoretically these convergence difficulties and propose an alternative solution. |
doi_str_mv | 10.1016/j.jcp.2020.109463 |
format | article |
fullrecord | <record><control><sourceid>proquest_hal_p</sourceid><recordid>TN_cdi_hal_primary_oai_HAL_hal_02546081v2</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S0021999120302370</els_id><sourcerecordid>2442333187</sourcerecordid><originalsourceid>FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3</originalsourceid><addsrcrecordid>eNp9kE9LwzAYh4MoOKcfwFvAk4fON3_WpXoaQ50w8aLnkKWJpsRmJm3Bb29KxaOn5Bee30veB6FLAgsCpLxpFo0-LCjQMVe8ZEdoli9Q0BUpj9EMgJKiqipyis5SagBALLmYIbUJ7WDiu2m1wS6l3iTsWlyb6AbVuSHHYPFzaDuDNyr6gNve-0IH711yoc1wZ96j8tiG-Nn73AltusVrnILvx3COTqzyyVz8nnP09nD_utkWu5fHp816V2gOtCsIJ4rXlQVeEVsKUIaVuqz03mhqLbdLxoATC6s90wRAaaH2fFlXQtSitGLP5uh6mvuhvDxE96nitwzKye16J8c3oEtegiADzezVxB5i-Mord7IJfWzz9yTlnDLGiFhlikyUjiGlaOzfWAJytC4bma3L0bqcrOfO3dQxedXBmSiTdqPc2kWjO1kH90_7B-CVijU</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2442333187</pqid></control><display><type>article</type><title>Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution</title><source>ScienceDirect Journals</source><creator>Tregan, J.-M. ; Blanco, S. ; Dauchet, J. ; El Hafi, M. ; Fournier, R. ; Ibarrart, L. ; Lapeyre, P. ; Villefranque, N.</creator><creatorcontrib>Tregan, J.-M. ; Blanco, S. ; Dauchet, J. ; El Hafi, M. ; Fournier, R. ; Ibarrart, L. ; Lapeyre, P. ; Villefranque, N.</creatorcontrib><description>•Monte Carlo algorithm is used to evaluate simultaneously the observable and its Jacobian.•Null-collision algorithm (NCA) are very successful when dealing with linear-physics in heterogeneous media.•We showed that NCA are sources of convergence difficulties to sensitivity-evaluations of optical parameter.•We propose an alternative solution.
When a Monte Carlo algorithm is used to evaluate a physical observable A, it is possible to slightly modify the algorithm so that it evaluates simultaneously A and the derivatives ∂ςA of A with respect to each problem-parameter ς. The principle is the following: Monte Carlo considers A as the expectation of a random variable, this expectation is an integral, this integral can be derivated as function of the problem-parameter to give a new integral, and this new integral can in turn be evaluated using Monte Carlo. The two Monte Carlo computations (of A and ∂ςA) are simultaneous when they make use of the same random samples, i.e. when the two integrals have the exact same structure. It was proven theoretically that this was always possible, but nothing ensures that the two estimators have the same convergence properties: even when a large enough sample-size is used so that A is evaluated very accurately, the evaluation of ∂ςA using the same sample can remain inaccurate. We discuss here such a pathological example: null-collision algorithms are very successful when dealing with radiative transfer in heterogeneous media, but they are sources of convergence difficulties as soon as sensitivity-evaluations are considered. We analyse theoretically these convergence difficulties and propose an alternative solution.</description><identifier>ISSN: 0021-9991</identifier><identifier>EISSN: 1090-2716</identifier><identifier>DOI: 10.1016/j.jcp.2020.109463</identifier><language>eng</language><publisher>Cambridge: Elsevier Inc</publisher><subject>Algorithms ; Computational physics ; Convergence ; Derivatives ; Direct derivatives ; Engineering Sciences ; Integral formulation ; Integrals ; Monte Carlo method ; Null-collision algorithm ; Parameters ; Radiative transfer ; Random variables ; Sensitivity</subject><ispartof>Journal of computational physics, 2020-07, Vol.413, p.109463-20/109463, Article 109463</ispartof><rights>2020 Elsevier Inc.</rights><rights>Copyright Elsevier Science Ltd. Jul 15, 2020</rights><rights>Distributed under a Creative Commons Attribution 4.0 International License</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3</citedby><cites>FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3</cites><orcidid>0000-0002-8191-8280 ; 0000-0002-4687-2635 ; 0000-0001-9675-9721</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>230,315,786,790,891,27957,27958</link.rule.ids><backlink>$$Uhttps://imt-mines-albi.hal.science/hal-02546081$$DView record in HAL$$Hfree_for_read</backlink></links><search><creatorcontrib>Tregan, J.-M.</creatorcontrib><creatorcontrib>Blanco, S.</creatorcontrib><creatorcontrib>Dauchet, J.</creatorcontrib><creatorcontrib>El Hafi, M.</creatorcontrib><creatorcontrib>Fournier, R.</creatorcontrib><creatorcontrib>Ibarrart, L.</creatorcontrib><creatorcontrib>Lapeyre, P.</creatorcontrib><creatorcontrib>Villefranque, N.</creatorcontrib><title>Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution</title><title>Journal of computational physics</title><description>•Monte Carlo algorithm is used to evaluate simultaneously the observable and its Jacobian.•Null-collision algorithm (NCA) are very successful when dealing with linear-physics in heterogeneous media.•We showed that NCA are sources of convergence difficulties to sensitivity-evaluations of optical parameter.•We propose an alternative solution.
When a Monte Carlo algorithm is used to evaluate a physical observable A, it is possible to slightly modify the algorithm so that it evaluates simultaneously A and the derivatives ∂ςA of A with respect to each problem-parameter ς. The principle is the following: Monte Carlo considers A as the expectation of a random variable, this expectation is an integral, this integral can be derivated as function of the problem-parameter to give a new integral, and this new integral can in turn be evaluated using Monte Carlo. The two Monte Carlo computations (of A and ∂ςA) are simultaneous when they make use of the same random samples, i.e. when the two integrals have the exact same structure. It was proven theoretically that this was always possible, but nothing ensures that the two estimators have the same convergence properties: even when a large enough sample-size is used so that A is evaluated very accurately, the evaluation of ∂ςA using the same sample can remain inaccurate. We discuss here such a pathological example: null-collision algorithms are very successful when dealing with radiative transfer in heterogeneous media, but they are sources of convergence difficulties as soon as sensitivity-evaluations are considered. We analyse theoretically these convergence difficulties and propose an alternative solution.</description><subject>Algorithms</subject><subject>Computational physics</subject><subject>Convergence</subject><subject>Derivatives</subject><subject>Direct derivatives</subject><subject>Engineering Sciences</subject><subject>Integral formulation</subject><subject>Integrals</subject><subject>Monte Carlo method</subject><subject>Null-collision algorithm</subject><subject>Parameters</subject><subject>Radiative transfer</subject><subject>Random variables</subject><subject>Sensitivity</subject><issn>0021-9991</issn><issn>1090-2716</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><recordid>eNp9kE9LwzAYh4MoOKcfwFvAk4fON3_WpXoaQ50w8aLnkKWJpsRmJm3Bb29KxaOn5Bee30veB6FLAgsCpLxpFo0-LCjQMVe8ZEdoli9Q0BUpj9EMgJKiqipyis5SagBALLmYIbUJ7WDiu2m1wS6l3iTsWlyb6AbVuSHHYPFzaDuDNyr6gNve-0IH711yoc1wZ96j8tiG-Nn73AltusVrnILvx3COTqzyyVz8nnP09nD_utkWu5fHp816V2gOtCsIJ4rXlQVeEVsKUIaVuqz03mhqLbdLxoATC6s90wRAaaH2fFlXQtSitGLP5uh6mvuhvDxE96nitwzKye16J8c3oEtegiADzezVxB5i-Mord7IJfWzz9yTlnDLGiFhlikyUjiGlaOzfWAJytC4bma3L0bqcrOfO3dQxedXBmSiTdqPc2kWjO1kH90_7B-CVijU</recordid><startdate>20200715</startdate><enddate>20200715</enddate><creator>Tregan, J.-M.</creator><creator>Blanco, S.</creator><creator>Dauchet, J.</creator><creator>El Hafi, M.</creator><creator>Fournier, R.</creator><creator>Ibarrart, L.</creator><creator>Lapeyre, P.</creator><creator>Villefranque, N.</creator><general>Elsevier Inc</general><general>Elsevier Science Ltd</general><general>Elsevier</general><scope>AAYXX</scope><scope>CITATION</scope><scope>7SC</scope><scope>7SP</scope><scope>7U5</scope><scope>8FD</scope><scope>JQ2</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><scope>1XC</scope><scope>VOOES</scope><orcidid>https://orcid.org/0000-0002-8191-8280</orcidid><orcidid>https://orcid.org/0000-0002-4687-2635</orcidid><orcidid>https://orcid.org/0000-0001-9675-9721</orcidid></search><sort><creationdate>20200715</creationdate><title>Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution</title><author>Tregan, J.-M. ; Blanco, S. ; Dauchet, J. ; El Hafi, M. ; Fournier, R. ; Ibarrart, L. ; Lapeyre, P. ; Villefranque, N.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Algorithms</topic><topic>Computational physics</topic><topic>Convergence</topic><topic>Derivatives</topic><topic>Direct derivatives</topic><topic>Engineering Sciences</topic><topic>Integral formulation</topic><topic>Integrals</topic><topic>Monte Carlo method</topic><topic>Null-collision algorithm</topic><topic>Parameters</topic><topic>Radiative transfer</topic><topic>Random variables</topic><topic>Sensitivity</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Tregan, J.-M.</creatorcontrib><creatorcontrib>Blanco, S.</creatorcontrib><creatorcontrib>Dauchet, J.</creatorcontrib><creatorcontrib>El Hafi, M.</creatorcontrib><creatorcontrib>Fournier, R.</creatorcontrib><creatorcontrib>Ibarrart, L.</creatorcontrib><creatorcontrib>Lapeyre, P.</creatorcontrib><creatorcontrib>Villefranque, N.</creatorcontrib><collection>CrossRef</collection><collection>Computer and Information Systems Abstracts</collection><collection>Electronics & Communications Abstracts</collection><collection>Solid State and Superconductivity Abstracts</collection><collection>Technology Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><collection>Hyper Article en Ligne (HAL)</collection><collection>Hyper Article en Ligne (HAL) (Open Access)</collection><jtitle>Journal of computational physics</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Tregan, J.-M.</au><au>Blanco, S.</au><au>Dauchet, J.</au><au>El Hafi, M.</au><au>Fournier, R.</au><au>Ibarrart, L.</au><au>Lapeyre, P.</au><au>Villefranque, N.</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution</atitle><jtitle>Journal of computational physics</jtitle><date>2020-07-15</date><risdate>2020</risdate><volume>413</volume><spage>109463</spage><epage>20/109463</epage><pages>109463-20/109463</pages><artnum>109463</artnum><issn>0021-9991</issn><eissn>1090-2716</eissn><abstract>•Monte Carlo algorithm is used to evaluate simultaneously the observable and its Jacobian.•Null-collision algorithm (NCA) are very successful when dealing with linear-physics in heterogeneous media.•We showed that NCA are sources of convergence difficulties to sensitivity-evaluations of optical parameter.•We propose an alternative solution.
When a Monte Carlo algorithm is used to evaluate a physical observable A, it is possible to slightly modify the algorithm so that it evaluates simultaneously A and the derivatives ∂ςA of A with respect to each problem-parameter ς. The principle is the following: Monte Carlo considers A as the expectation of a random variable, this expectation is an integral, this integral can be derivated as function of the problem-parameter to give a new integral, and this new integral can in turn be evaluated using Monte Carlo. The two Monte Carlo computations (of A and ∂ςA) are simultaneous when they make use of the same random samples, i.e. when the two integrals have the exact same structure. It was proven theoretically that this was always possible, but nothing ensures that the two estimators have the same convergence properties: even when a large enough sample-size is used so that A is evaluated very accurately, the evaluation of ∂ςA using the same sample can remain inaccurate. We discuss here such a pathological example: null-collision algorithms are very successful when dealing with radiative transfer in heterogeneous media, but they are sources of convergence difficulties as soon as sensitivity-evaluations are considered. We analyse theoretically these convergence difficulties and propose an alternative solution.</abstract><cop>Cambridge</cop><pub>Elsevier Inc</pub><doi>10.1016/j.jcp.2020.109463</doi><orcidid>https://orcid.org/0000-0002-8191-8280</orcidid><orcidid>https://orcid.org/0000-0002-4687-2635</orcidid><orcidid>https://orcid.org/0000-0001-9675-9721</orcidid><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | ISSN: 0021-9991 |
ispartof | Journal of computational physics, 2020-07, Vol.413, p.109463-20/109463, Article 109463 |
issn | 0021-9991 1090-2716 |
language | eng |
recordid | cdi_hal_primary_oai_HAL_hal_02546081v2 |
source | ScienceDirect Journals |
subjects | Algorithms Computational physics Convergence Derivatives Direct derivatives Engineering Sciences Integral formulation Integrals Monte Carlo method Null-collision algorithm Parameters Radiative transfer Random variables Sensitivity |
title | Convergence issues in derivatives of Monte Carlo null-collision integral formulations: A solution |
url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-09-21T21%3A32%3A45IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_hal_p&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Convergence%20issues%20in%20derivatives%20of%20Monte%20Carlo%20null-collision%20integral%20formulations:%20A%20solution&rft.jtitle=Journal%20of%20computational%20physics&rft.au=Tregan,%20J.-M.&rft.date=2020-07-15&rft.volume=413&rft.spage=109463&rft.epage=20/109463&rft.pages=109463-20/109463&rft.artnum=109463&rft.issn=0021-9991&rft.eissn=1090-2716&rft_id=info:doi/10.1016/j.jcp.2020.109463&rft_dat=%3Cproquest_hal_p%3E2442333187%3C/proquest_hal_p%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c402t-141a4d9f0491f680ae36c69cbec2ff4f533041f07b3c100ac8ab45d988d86f8b3%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2442333187&rft_id=info:pmid/&rfr_iscdi=true |