Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems

Solving high-dimensional partial differential equations is a recurrent challenge in economics, science and engineering. In recent years, a great number of computational approaches have been developed, most of them relying on a combination of Monte Carlo sampling and deep learning based approximation...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Nüsken, Nikolas, Richter, Lorenz
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 Nüsken, Nikolas
Richter, Lorenz
description Solving high-dimensional partial differential equations is a recurrent challenge in economics, science and engineering. In recent years, a great number of computational approaches have been developed, most of them relying on a combination of Monte Carlo sampling and deep learning based approximation. For elliptic and parabolic problems, existing methods can broadly be classified into those resting on reformulations in terms of $\textit{backward stochastic differential equations}$ (BSDEs) and those aiming to minimize a regression-type $L^2$-error ($\textit{physics-informed neural networks}$, PINNs). In this paper, we review the literature and suggest a methodology based on the novel $\textit{diffusion loss}$ that interpolates between BSDEs and PINNs. Our contribution opens the door towards a unified understanding of numerical approaches for high-dimensional PDEs, as well as for implementations that combine the strengths of BSDEs and PINNs. The diffusion loss furthermore bears close similarities to $\textit{(least squares) temporal difference}$ objectives found in reinforcement learning. We also discuss eigenvalue problems and perform extensive numerical studies, including calculations of the ground state for nonlinear Schr\"odinger operators and committor functions relevant in molecular dynamics.
doi_str_mv 10.48550/arxiv.2112.03749
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2112_03749</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2112_03749</sourcerecordid><originalsourceid>FETCH-LOGICAL-a679-44745d7a2d3fbfc38217e2b2bf7d8c1d53bd71ce56d5b246151f0567f1e400bc3</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwAr_QIKfccoOSoFKVUGi--g6vq4iuY7lpAX-njZlNRrpaDSHkDvOSlVrzR4g_3THUnAuSiaNml-T3SqOmFMfYOzijlocvxEjff56WQ4UoqOfq81meKQOMdGAkOMZ832mGEKXxq6dqAQZbB9OzfaH6CD_0iOEA9KUextwP9yQKw9hwNv_nJHt63K7eC_WH2-rxdO6gMrMC6WM0s6AcNJb38pacIPCCuuNq1vutLTO8BZ15bQVquKae6Yr4zkqxmwrZ-T-MjuZNil3-9OX5mzcTMbyD0DYUfA</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems</title><source>arXiv.org</source><creator>Nüsken, Nikolas ; Richter, Lorenz</creator><creatorcontrib>Nüsken, Nikolas ; Richter, Lorenz</creatorcontrib><description>Solving high-dimensional partial differential equations is a recurrent challenge in economics, science and engineering. In recent years, a great number of computational approaches have been developed, most of them relying on a combination of Monte Carlo sampling and deep learning based approximation. For elliptic and parabolic problems, existing methods can broadly be classified into those resting on reformulations in terms of $\textit{backward stochastic differential equations}$ (BSDEs) and those aiming to minimize a regression-type $L^2$-error ($\textit{physics-informed neural networks}$, PINNs). In this paper, we review the literature and suggest a methodology based on the novel $\textit{diffusion loss}$ that interpolates between BSDEs and PINNs. Our contribution opens the door towards a unified understanding of numerical approaches for high-dimensional PDEs, as well as for implementations that combine the strengths of BSDEs and PINNs. The diffusion loss furthermore bears close similarities to $\textit{(least squares) temporal difference}$ objectives found in reinforcement learning. We also discuss eigenvalue problems and perform extensive numerical studies, including calculations of the ground state for nonlinear Schr\"odinger operators and committor functions relevant in molecular dynamics.</description><identifier>DOI: 10.48550/arxiv.2112.03749</identifier><language>eng</language><subject>Computer Science - Numerical Analysis ; Mathematics - Numerical Analysis ; Mathematics - Probability ; Statistics - Machine Learning</subject><creationdate>2021-12</creationdate><rights>http://creativecommons.org/licenses/by-nc-sa/4.0</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/2112.03749$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2112.03749$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Nüsken, Nikolas</creatorcontrib><creatorcontrib>Richter, Lorenz</creatorcontrib><title>Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems</title><description>Solving high-dimensional partial differential equations is a recurrent challenge in economics, science and engineering. In recent years, a great number of computational approaches have been developed, most of them relying on a combination of Monte Carlo sampling and deep learning based approximation. For elliptic and parabolic problems, existing methods can broadly be classified into those resting on reformulations in terms of $\textit{backward stochastic differential equations}$ (BSDEs) and those aiming to minimize a regression-type $L^2$-error ($\textit{physics-informed neural networks}$, PINNs). In this paper, we review the literature and suggest a methodology based on the novel $\textit{diffusion loss}$ that interpolates between BSDEs and PINNs. Our contribution opens the door towards a unified understanding of numerical approaches for high-dimensional PDEs, as well as for implementations that combine the strengths of BSDEs and PINNs. The diffusion loss furthermore bears close similarities to $\textit{(least squares) temporal difference}$ objectives found in reinforcement learning. We also discuss eigenvalue problems and perform extensive numerical studies, including calculations of the ground state for nonlinear Schr\"odinger operators and committor functions relevant in molecular dynamics.</description><subject>Computer Science - Numerical Analysis</subject><subject>Mathematics - Numerical Analysis</subject><subject>Mathematics - Probability</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwAr_QIKfccoOSoFKVUGi--g6vq4iuY7lpAX-njZlNRrpaDSHkDvOSlVrzR4g_3THUnAuSiaNml-T3SqOmFMfYOzijlocvxEjff56WQ4UoqOfq81meKQOMdGAkOMZ832mGEKXxq6dqAQZbB9OzfaH6CD_0iOEA9KUextwP9yQKw9hwNv_nJHt63K7eC_WH2-rxdO6gMrMC6WM0s6AcNJb38pacIPCCuuNq1vutLTO8BZ15bQVquKae6Yr4zkqxmwrZ-T-MjuZNil3-9OX5mzcTMbyD0DYUfA</recordid><startdate>20211207</startdate><enddate>20211207</enddate><creator>Nüsken, Nikolas</creator><creator>Richter, Lorenz</creator><scope>AKY</scope><scope>AKZ</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20211207</creationdate><title>Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems</title><author>Nüsken, Nikolas ; Richter, Lorenz</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a679-44745d7a2d3fbfc38217e2b2bf7d8c1d53bd71ce56d5b246151f0567f1e400bc3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Numerical Analysis</topic><topic>Mathematics - Numerical Analysis</topic><topic>Mathematics - Probability</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Nüsken, Nikolas</creatorcontrib><creatorcontrib>Richter, Lorenz</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Mathematics</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Nüsken, Nikolas</au><au>Richter, Lorenz</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems</atitle><date>2021-12-07</date><risdate>2021</risdate><abstract>Solving high-dimensional partial differential equations is a recurrent challenge in economics, science and engineering. In recent years, a great number of computational approaches have been developed, most of them relying on a combination of Monte Carlo sampling and deep learning based approximation. For elliptic and parabolic problems, existing methods can broadly be classified into those resting on reformulations in terms of $\textit{backward stochastic differential equations}$ (BSDEs) and those aiming to minimize a regression-type $L^2$-error ($\textit{physics-informed neural networks}$, PINNs). In this paper, we review the literature and suggest a methodology based on the novel $\textit{diffusion loss}$ that interpolates between BSDEs and PINNs. Our contribution opens the door towards a unified understanding of numerical approaches for high-dimensional PDEs, as well as for implementations that combine the strengths of BSDEs and PINNs. The diffusion loss furthermore bears close similarities to $\textit{(least squares) temporal difference}$ objectives found in reinforcement learning. We also discuss eigenvalue problems and perform extensive numerical studies, including calculations of the ground state for nonlinear Schr\"odinger operators and committor functions relevant in molecular dynamics.</abstract><doi>10.48550/arxiv.2112.03749</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2112.03749
ispartof
issn
language eng
recordid cdi_arxiv_primary_2112_03749
source arXiv.org
subjects Computer Science - Numerical Analysis
Mathematics - Numerical Analysis
Mathematics - Probability
Statistics - Machine Learning
title Interpolating between BSDEs and PINNs: deep learning for elliptic and parabolic boundary value problems
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-30T12%3A57%3A39IST&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=Interpolating%20between%20BSDEs%20and%20PINNs:%20deep%20learning%20for%20elliptic%20and%20parabolic%20boundary%20value%20problems&rft.au=N%C3%BCsken,%20Nikolas&rft.date=2021-12-07&rft_id=info:doi/10.48550/arxiv.2112.03749&rft_dat=%3Carxiv_GOX%3E2112_03749%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