Loading…

Convergence rates of damped inertial dynamics under geometric conditions

In this article a family of second order ODEs associated to inertial gradient descend is studied. These ODEs are widely used to build trajectories converging to a minimizer x* of a function F , possibly convex. This family includes the continuous version of the Nesterov inertial scheme and the conti...

Full description

Saved in:
Bibliographic Details
Published in:SIAM journal on optimization 2020-08, Vol.30 (3)
Main Authors: Sebbouh, Othmane, Dossal, Charles H, Rondepierre, Aude
Format: Article
Language:English
Subjects:
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page
container_issue 3
container_start_page
container_title SIAM journal on optimization
container_volume 30
creator Sebbouh, Othmane
Dossal, Charles H
Rondepierre, Aude
description In this article a family of second order ODEs associated to inertial gradient descend is studied. These ODEs are widely used to build trajectories converging to a minimizer x* of a function F , possibly convex. This family includes the continuous version of the Nesterov inertial scheme and the continuous heavy ball method. Several damping parameters, not necessarily vanishing, and a perturbation term g are thus considered. The damping parameter is linked to the inertia of the associated inertial scheme and the perturbation term g is linked to the error that can be done on the gradient of the function F. This article presents new asymptotic bounds on F(x(t)) − F (x*) where x is a solution of the ODE, when F is convex and satisfies local geometrical properties such as Łojasiewicz properties and under integrability conditions on g. Even if geometrical properties and perturbations were already studied for most ODEs of these families, it is the first time they are jointly studied. All these results give an insight on the behavior of these inertial and perturbed algorithms if F satisfies some Łojasiewicz properties especially in the setting of stochastic algorithms.
doi_str_mv 10.1137/19M1272767
format article
fullrecord <record><control><sourceid>hal</sourceid><recordid>TN_cdi_hal_primary_oai_HAL_hal_02173978v3</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>oai_HAL_hal_02173978v3</sourcerecordid><originalsourceid>FETCH-hal_primary_oai_HAL_hal_02173978v33</originalsourceid><addsrcrecordid>eNqVyrsKwjAYQOEMCtbL4hNkdajmok07SlE66OZeQvO3RpqkJLHQtxfBF3A68HEQ2lKyp5SLAy3ulAkmMjFDCSUnlmaMHxdoGcKLEJIXWZ6gqnR2BN-BbQB7GSFg12IlzQAKaws-atljNVlpdBPw2yrwuANnIHrd4MZZpaN2NqzRvJV9gM2vK7S7Xh5llT5lXw9eG-mn2kldV-db_TXCqOCFyEfO_3k_NINDyw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Convergence rates of damped inertial dynamics under geometric conditions</title><source>SIAM Journals Archive</source><creator>Sebbouh, Othmane ; Dossal, Charles H ; Rondepierre, Aude</creator><creatorcontrib>Sebbouh, Othmane ; Dossal, Charles H ; Rondepierre, Aude</creatorcontrib><description>In this article a family of second order ODEs associated to inertial gradient descend is studied. These ODEs are widely used to build trajectories converging to a minimizer x* of a function F , possibly convex. This family includes the continuous version of the Nesterov inertial scheme and the continuous heavy ball method. Several damping parameters, not necessarily vanishing, and a perturbation term g are thus considered. The damping parameter is linked to the inertia of the associated inertial scheme and the perturbation term g is linked to the error that can be done on the gradient of the function F. This article presents new asymptotic bounds on F(x(t)) − F (x*) where x is a solution of the ODE, when F is convex and satisfies local geometrical properties such as Łojasiewicz properties and under integrability conditions on g. Even if geometrical properties and perturbations were already studied for most ODEs of these families, it is the first time they are jointly studied. All these results give an insight on the behavior of these inertial and perturbed algorithms if F satisfies some Łojasiewicz properties especially in the setting of stochastic algorithms.</description><identifier>ISSN: 1052-6234</identifier><identifier>DOI: 10.1137/19M1272767</identifier><language>eng</language><publisher>Society for Industrial and Applied Mathematics</publisher><subject>Mathematics ; Optimization and Control</subject><ispartof>SIAM journal on optimization, 2020-08, Vol.30 (3)</ispartof><rights>Distributed under a Creative Commons Attribution 4.0 International License</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><orcidid>0000-0001-7366-5677 ; 0000-0001-7366-5677</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>230,314,780,784,885,27924,27925</link.rule.ids><backlink>$$Uhttps://hal.science/hal-02173978$$DView record in HAL$$Hfree_for_read</backlink></links><search><creatorcontrib>Sebbouh, Othmane</creatorcontrib><creatorcontrib>Dossal, Charles H</creatorcontrib><creatorcontrib>Rondepierre, Aude</creatorcontrib><title>Convergence rates of damped inertial dynamics under geometric conditions</title><title>SIAM journal on optimization</title><description>In this article a family of second order ODEs associated to inertial gradient descend is studied. These ODEs are widely used to build trajectories converging to a minimizer x* of a function F , possibly convex. This family includes the continuous version of the Nesterov inertial scheme and the continuous heavy ball method. Several damping parameters, not necessarily vanishing, and a perturbation term g are thus considered. The damping parameter is linked to the inertia of the associated inertial scheme and the perturbation term g is linked to the error that can be done on the gradient of the function F. This article presents new asymptotic bounds on F(x(t)) − F (x*) where x is a solution of the ODE, when F is convex and satisfies local geometrical properties such as Łojasiewicz properties and under integrability conditions on g. Even if geometrical properties and perturbations were already studied for most ODEs of these families, it is the first time they are jointly studied. All these results give an insight on the behavior of these inertial and perturbed algorithms if F satisfies some Łojasiewicz properties especially in the setting of stochastic algorithms.</description><subject>Mathematics</subject><subject>Optimization and Control</subject><issn>1052-6234</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><recordid>eNqVyrsKwjAYQOEMCtbL4hNkdajmok07SlE66OZeQvO3RpqkJLHQtxfBF3A68HEQ2lKyp5SLAy3ulAkmMjFDCSUnlmaMHxdoGcKLEJIXWZ6gqnR2BN-BbQB7GSFg12IlzQAKaws-atljNVlpdBPw2yrwuANnIHrd4MZZpaN2NqzRvJV9gM2vK7S7Xh5llT5lXw9eG-mn2kldV-db_TXCqOCFyEfO_3k_NINDyw</recordid><startdate>202008</startdate><enddate>202008</enddate><creator>Sebbouh, Othmane</creator><creator>Dossal, Charles H</creator><creator>Rondepierre, Aude</creator><general>Society for Industrial and Applied Mathematics</general><scope>1XC</scope><scope>VOOES</scope><orcidid>https://orcid.org/0000-0001-7366-5677</orcidid><orcidid>https://orcid.org/0000-0001-7366-5677</orcidid></search><sort><creationdate>202008</creationdate><title>Convergence rates of damped inertial dynamics under geometric conditions</title><author>Sebbouh, Othmane ; Dossal, Charles H ; Rondepierre, Aude</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-hal_primary_oai_HAL_hal_02173978v33</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Mathematics</topic><topic>Optimization and Control</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Sebbouh, Othmane</creatorcontrib><creatorcontrib>Dossal, Charles H</creatorcontrib><creatorcontrib>Rondepierre, Aude</creatorcontrib><collection>Hyper Article en Ligne (HAL)</collection><collection>Hyper Article en Ligne (HAL) (Open Access)</collection><jtitle>SIAM journal on optimization</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Sebbouh, Othmane</au><au>Dossal, Charles H</au><au>Rondepierre, Aude</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Convergence rates of damped inertial dynamics under geometric conditions</atitle><jtitle>SIAM journal on optimization</jtitle><date>2020-08</date><risdate>2020</risdate><volume>30</volume><issue>3</issue><issn>1052-6234</issn><abstract>In this article a family of second order ODEs associated to inertial gradient descend is studied. These ODEs are widely used to build trajectories converging to a minimizer x* of a function F , possibly convex. This family includes the continuous version of the Nesterov inertial scheme and the continuous heavy ball method. Several damping parameters, not necessarily vanishing, and a perturbation term g are thus considered. The damping parameter is linked to the inertia of the associated inertial scheme and the perturbation term g is linked to the error that can be done on the gradient of the function F. This article presents new asymptotic bounds on F(x(t)) − F (x*) where x is a solution of the ODE, when F is convex and satisfies local geometrical properties such as Łojasiewicz properties and under integrability conditions on g. Even if geometrical properties and perturbations were already studied for most ODEs of these families, it is the first time they are jointly studied. All these results give an insight on the behavior of these inertial and perturbed algorithms if F satisfies some Łojasiewicz properties especially in the setting of stochastic algorithms.</abstract><pub>Society for Industrial and Applied Mathematics</pub><doi>10.1137/19M1272767</doi><orcidid>https://orcid.org/0000-0001-7366-5677</orcidid><orcidid>https://orcid.org/0000-0001-7366-5677</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 1052-6234
ispartof SIAM journal on optimization, 2020-08, Vol.30 (3)
issn 1052-6234
language eng
recordid cdi_hal_primary_oai_HAL_hal_02173978v3
source SIAM Journals Archive
subjects Mathematics
Optimization and Control
title Convergence rates of damped inertial dynamics under geometric conditions
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-01T14%3A45%3A51IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-hal&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Convergence%20rates%20of%20damped%20inertial%20dynamics%20under%20geometric%20conditions&rft.jtitle=SIAM%20journal%20on%20optimization&rft.au=Sebbouh,%20Othmane&rft.date=2020-08&rft.volume=30&rft.issue=3&rft.issn=1052-6234&rft_id=info:doi/10.1137/19M1272767&rft_dat=%3Chal%3Eoai_HAL_hal_02173978v3%3C/hal%3E%3Cgrp_id%3Ecdi_FETCH-hal_primary_oai_HAL_hal_02173978v33%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true