Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian

In this paper, we address the challenge of solving large-scale graph-structured nonlinear programs (gsNLPs) in a scalable manner. GsNLPs are problems in which the objective and constraint functions are associated with nodes on a graph and depend on the variables of adjacent nodes. This graph-structu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ni, Runxin, Na, Sen, Shin, Sungho, Anitescu, Mihai
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 Ni, Runxin
Na, Sen
Shin, Sungho
Anitescu, Mihai
description In this paper, we address the challenge of solving large-scale graph-structured nonlinear programs (gsNLPs) in a scalable manner. GsNLPs are problems in which the objective and constraint functions are associated with nodes on a graph and depend on the variables of adjacent nodes. This graph-structured formulation encompasses various specific instances, such as dynamic optimization, PDE-constrained optimization, multistage stochastic optimization, and general network optimization. By leveraging the sequential quadratic programming (SQP) framework, we propose a globally convergent overlapping graph decomposition method to solve large-scale gsNLPs under standard mild regularity conditions on the graph topology. In each iteration, we perform an overlapping graph decomposition to compute an approximate Newton direction in a parallel environment. Then, we select a suitable stepsize and update the primal-dual iterate by performing a backtracking line search on an exact augmented Lagrangian merit function. Built on the exponential decay of sensitivity of gsNLPs, we show that the approximate Newton direction is a descent direction of the augmented Lagrangian, which leads to global convergence with a local linear convergence rate. In particular, global convergence is achieved for sufficiently large overlaps, and the local linear convergence rate improves exponentially in terms of the overlap size. Our results match existing state-of-the-art guarantees established for dynamic programs (which simply correspond to linear graphs). We validate the theory on a semilinear elliptic PDE-constrained problem.
doi_str_mv 10.48550/arxiv.2402.17170
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2402_17170</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2402_17170</sourcerecordid><originalsourceid>FETCH-arxiv_primary_2402_171703</originalsourceid><addsrcrecordid>eNqFjjEOgkAQRbexMOoBrJwLiKAYbY2gFiZqtCcjrDAJu6zDgnh7wdhb_eTl5-UJMfZcx18vl-4MuaHamfvu3PFW3srtCxNQaZnulZUJXOWzktoS5nCpMGG0FMOZi5RRKdIpvMhmcKol52hMB_aMJoNAxoUyRUmWCg2oEwgbjC1sqlS1vtZ8xNahU0I9FL0H5qUc_XYgJrvwtj1Mv22RYVLI76hrjL6Ni_-PD8ZTSVM</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian</title><source>arXiv.org</source><creator>Ni, Runxin ; Na, Sen ; Shin, Sungho ; Anitescu, Mihai</creator><creatorcontrib>Ni, Runxin ; Na, Sen ; Shin, Sungho ; Anitescu, Mihai</creatorcontrib><description>In this paper, we address the challenge of solving large-scale graph-structured nonlinear programs (gsNLPs) in a scalable manner. GsNLPs are problems in which the objective and constraint functions are associated with nodes on a graph and depend on the variables of adjacent nodes. This graph-structured formulation encompasses various specific instances, such as dynamic optimization, PDE-constrained optimization, multistage stochastic optimization, and general network optimization. By leveraging the sequential quadratic programming (SQP) framework, we propose a globally convergent overlapping graph decomposition method to solve large-scale gsNLPs under standard mild regularity conditions on the graph topology. In each iteration, we perform an overlapping graph decomposition to compute an approximate Newton direction in a parallel environment. Then, we select a suitable stepsize and update the primal-dual iterate by performing a backtracking line search on an exact augmented Lagrangian merit function. Built on the exponential decay of sensitivity of gsNLPs, we show that the approximate Newton direction is a descent direction of the augmented Lagrangian, which leads to global convergence with a local linear convergence rate. In particular, global convergence is achieved for sufficiently large overlaps, and the local linear convergence rate improves exponentially in terms of the overlap size. Our results match existing state-of-the-art guarantees established for dynamic programs (which simply correspond to linear graphs). We validate the theory on a semilinear elliptic PDE-constrained problem.</description><identifier>DOI: 10.48550/arxiv.2402.17170</identifier><language>eng</language><subject>Mathematics - Optimization and Control</subject><creationdate>2024-02</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,781,886</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2402.17170$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2402.17170$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Ni, Runxin</creatorcontrib><creatorcontrib>Na, Sen</creatorcontrib><creatorcontrib>Shin, Sungho</creatorcontrib><creatorcontrib>Anitescu, Mihai</creatorcontrib><title>Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian</title><description>In this paper, we address the challenge of solving large-scale graph-structured nonlinear programs (gsNLPs) in a scalable manner. GsNLPs are problems in which the objective and constraint functions are associated with nodes on a graph and depend on the variables of adjacent nodes. This graph-structured formulation encompasses various specific instances, such as dynamic optimization, PDE-constrained optimization, multistage stochastic optimization, and general network optimization. By leveraging the sequential quadratic programming (SQP) framework, we propose a globally convergent overlapping graph decomposition method to solve large-scale gsNLPs under standard mild regularity conditions on the graph topology. In each iteration, we perform an overlapping graph decomposition to compute an approximate Newton direction in a parallel environment. Then, we select a suitable stepsize and update the primal-dual iterate by performing a backtracking line search on an exact augmented Lagrangian merit function. Built on the exponential decay of sensitivity of gsNLPs, we show that the approximate Newton direction is a descent direction of the augmented Lagrangian, which leads to global convergence with a local linear convergence rate. In particular, global convergence is achieved for sufficiently large overlaps, and the local linear convergence rate improves exponentially in terms of the overlap size. Our results match existing state-of-the-art guarantees established for dynamic programs (which simply correspond to linear graphs). We validate the theory on a semilinear elliptic PDE-constrained problem.</description><subject>Mathematics - Optimization and Control</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNqFjjEOgkAQRbexMOoBrJwLiKAYbY2gFiZqtCcjrDAJu6zDgnh7wdhb_eTl5-UJMfZcx18vl-4MuaHamfvu3PFW3srtCxNQaZnulZUJXOWzktoS5nCpMGG0FMOZi5RRKdIpvMhmcKol52hMB_aMJoNAxoUyRUmWCg2oEwgbjC1sqlS1vtZ8xNahU0I9FL0H5qUc_XYgJrvwtj1Mv22RYVLI76hrjL6Ni_-PD8ZTSVM</recordid><startdate>20240226</startdate><enddate>20240226</enddate><creator>Ni, Runxin</creator><creator>Na, Sen</creator><creator>Shin, Sungho</creator><creator>Anitescu, Mihai</creator><scope>AKZ</scope><scope>GOX</scope></search><sort><creationdate>20240226</creationdate><title>Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian</title><author>Ni, Runxin ; Na, Sen ; Shin, Sungho ; Anitescu, Mihai</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-arxiv_primary_2402_171703</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Mathematics - Optimization and Control</topic><toplevel>online_resources</toplevel><creatorcontrib>Ni, Runxin</creatorcontrib><creatorcontrib>Na, Sen</creatorcontrib><creatorcontrib>Shin, Sungho</creatorcontrib><creatorcontrib>Anitescu, Mihai</creatorcontrib><collection>arXiv Mathematics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Ni, Runxin</au><au>Na, Sen</au><au>Shin, Sungho</au><au>Anitescu, Mihai</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian</atitle><date>2024-02-26</date><risdate>2024</risdate><abstract>In this paper, we address the challenge of solving large-scale graph-structured nonlinear programs (gsNLPs) in a scalable manner. GsNLPs are problems in which the objective and constraint functions are associated with nodes on a graph and depend on the variables of adjacent nodes. This graph-structured formulation encompasses various specific instances, such as dynamic optimization, PDE-constrained optimization, multistage stochastic optimization, and general network optimization. By leveraging the sequential quadratic programming (SQP) framework, we propose a globally convergent overlapping graph decomposition method to solve large-scale gsNLPs under standard mild regularity conditions on the graph topology. In each iteration, we perform an overlapping graph decomposition to compute an approximate Newton direction in a parallel environment. Then, we select a suitable stepsize and update the primal-dual iterate by performing a backtracking line search on an exact augmented Lagrangian merit function. Built on the exponential decay of sensitivity of gsNLPs, we show that the approximate Newton direction is a descent direction of the augmented Lagrangian, which leads to global convergence with a local linear convergence rate. In particular, global convergence is achieved for sufficiently large overlaps, and the local linear convergence rate improves exponentially in terms of the overlap size. Our results match existing state-of-the-art guarantees established for dynamic programs (which simply correspond to linear graphs). We validate the theory on a semilinear elliptic PDE-constrained problem.</abstract><doi>10.48550/arxiv.2402.17170</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2402.17170
ispartof
issn
language eng
recordid cdi_arxiv_primary_2402_17170
source arXiv.org
subjects Mathematics - Optimization and Control
title Distributed Sequential Quadratic Programming with Overlapping Graph Decomposition and Exact Augmented Lagrangian
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-16T22%3A55%3A29IST&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=Distributed%20Sequential%20Quadratic%20Programming%20with%20Overlapping%20Graph%20Decomposition%20and%20Exact%20Augmented%20Lagrangian&rft.au=Ni,%20Runxin&rft.date=2024-02-26&rft_id=info:doi/10.48550/arxiv.2402.17170&rft_dat=%3Carxiv_GOX%3E2402_17170%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