Generalized self-concordant analysis of Frank–Wolfe algorithms

dc.bibliographicCitation.date2023
dc.bibliographicCitation.firstPage255
dc.bibliographicCitation.lastPage323
dc.bibliographicCitation.volume198
dc.contributor.authorDvurechensky, Pavel
dc.contributor.authorSafin, Kamil
dc.contributor.authorShtern, Shimrit
dc.contributor.authorStaudigl, Mathias
dc.date.accessioned2022-06-23T08:53:50Z
dc.date.available2022-06-23T08:53:50Z
dc.date.issued2022
dc.description.abstractProjection-free optimization via different variants of the Frank–Wolfe method has become one of the cornerstones of large scale optimization for machine learning and computational statistics. Numerous applications within these fields involve the minimization of functions with self-concordance like properties. Such generalized self-concordant functions do not necessarily feature a Lipschitz continuous gradient, nor are they strongly convex, making them a challenging class of functions for first-order methods. Indeed, in a number of applications, such as inverse covariance estimation or distance-weighted discrimination problems in binary classification, the loss is given by a generalized self-concordant function having potentially unbounded curvature. For such problems projection-free minimization methods have no theoretical convergence guarantee. This paper closes this apparent gap in the literature by developing provably convergent Frank–Wolfe algorithms with standard O(1/k) convergence rate guarantees. Based on these new insights, we show how these sublinearly convergent methods can be accelerated to yield linearly convergent projection-free methods, by either relying on the availability of a local liner minimization oracle, or a suitable modification of the away-step Frank–Wolfe method.eng
dc.description.versionpublishedVersioneng
dc.identifier.urihttps://oa.tib.eu/renate/handle/123456789/9126
dc.identifier.urihttps://doi.org/10.34657/8164
dc.language.isoengeng
dc.publisherBerlin ; Heidelberg : Springer
dc.relation.doihttps://doi.org/10.1007/s10107-022-01771-1
dc.relation.essn1436-4646
dc.relation.ispartofseriesMathematical programming : Series A, Series B 198 (2023)
dc.rights.licenseCC BY 4.0 Unported
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/
dc.subjectConvex programmingeng
dc.subjectFrank–Wolfe algorithmeng
dc.subjectGeneralized self-concordant functionseng
dc.subject.ddc510
dc.titleGeneralized self-concordant analysis of Frank–Wolfe algorithmseng
dc.typearticleeng
dc.typeTexteng
dcterms.bibliographicCitation.journalTitleMathematical programming : Series A, Series B
tib.accessRightsopenAccesseng
wgl.contributorWIASger
wgl.subjectMathematikger
wgl.subjectInformatikger
wgl.typeZeitschriftenartikelger
Files
Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Generalized_self-concordant.pdf
Size:
1.23 MB
Format:
Adobe Portable Document Format
Description: