Frank–Wolfe and friends: a journey into projection-free first-order optimization methods

Immanuel Bomze; Francesco Rinaldi; Damiano Zeffiro

doi:10.1007/s10288-021-00493-y

Frank–Wolfe and friends: a journey into projection-free first-order optimization methods

Immanuel Bomze, Francesco Rinaldi, Damiano Zeffiro

Publications: Contribution to journal › Review › Peer Reviewed

Abstract

Invented some 65 years ago in a seminal paper by Marguerite Straus-Frank and Philip Wolfe, the Frank-Wolfe method recently enjoys a remarkable revival, fuelled by the need of fast and reliable first-order optimization methods in Data Science and other relevant application areas. This review tries to explain the success of this approach by illustrating versatility and applicability in a wide range of contexts, combined with an account on recent progress in variants, improving on both the speed and efficiency of this surprisingly simple principle of first-order optimization.

Original language	English
Pages (from-to)	313-345
Number of pages	33
Journal	4 OR
Volume	19
Issue number	3
DOIs	https://doi.org/10.1007/s10288-021-00493-y
Publication status	Published - 6 Sep 2021

Austrian Fields of Science 2012

101015 Operations research

Keywords

ISOR
DSA (Data Science and Analytics)
Projection-free methods
Sparse optimization
Conditional gradient
DECOMPOSITION
Structured optimization
CONVERGENCE
CONDITIONAL GRADIENT ALGORITHMS
POINT
First-order methods

Access to Document

10.1007/s10288-021-00493-yLicence: CC BY 4.0

https://phaidra.univie.ac.at/o:1586496Licence: CC BY 4.0

Cite this

@article{d8e5ac9d80b841129c2d1ff124cda213,

title = "Frank–Wolfe and friends: a journey into projection-free first-order optimization methods",

abstract = "Invented some 65 years ago in a seminal paper by Marguerite Straus-Frank and Philip Wolfe, the Frank-Wolfe method recently enjoys a remarkable revival, fuelled by the need of fast and reliable first-order optimization methods in Data Science and other relevant application areas. This review tries to explain the success of this approach by illustrating versatility and applicability in a wide range of contexts, combined with an account on recent progress in variants, improving on both the speed and efficiency of this surprisingly simple principle of first-order optimization.",

keywords = "ISOR, DSA (Data Science and Analytics), Projection-free methods, Sparse optimization, Conditional gradient, DECOMPOSITION, Structured optimization, CONVERGENCE, CONDITIONAL GRADIENT ALGORITHMS, POINT, First-order methods",

author = "Immanuel Bomze and Francesco Rinaldi and Damiano Zeffiro",

note = "Publisher Copyright: {\textcopyright} 2021, The Author(s).",

year = "2021",

month = sep,

day = "6",

doi = "10.1007/s10288-021-00493-y",

language = "English",

volume = "19",

pages = "313--345",

journal = "4 OR",

issn = "1619-4500",

number = "3",

}

TY - JOUR

T1 - Frank–Wolfe and friends: a journey into projection-free first-order optimization methods

AU - Bomze, Immanuel

AU - Rinaldi, Francesco

AU - Zeffiro, Damiano

PY - 2021/9/6

Y1 - 2021/9/6

N2 - Invented some 65 years ago in a seminal paper by Marguerite Straus-Frank and Philip Wolfe, the Frank-Wolfe method recently enjoys a remarkable revival, fuelled by the need of fast and reliable first-order optimization methods in Data Science and other relevant application areas. This review tries to explain the success of this approach by illustrating versatility and applicability in a wide range of contexts, combined with an account on recent progress in variants, improving on both the speed and efficiency of this surprisingly simple principle of first-order optimization.

AB - Invented some 65 years ago in a seminal paper by Marguerite Straus-Frank and Philip Wolfe, the Frank-Wolfe method recently enjoys a remarkable revival, fuelled by the need of fast and reliable first-order optimization methods in Data Science and other relevant application areas. This review tries to explain the success of this approach by illustrating versatility and applicability in a wide range of contexts, combined with an account on recent progress in variants, improving on both the speed and efficiency of this surprisingly simple principle of first-order optimization.

KW - ISOR

KW - DSA (Data Science and Analytics)

KW - Projection-free methods

KW - Sparse optimization

KW - Conditional gradient

KW - DECOMPOSITION

KW - Structured optimization

KW - CONVERGENCE

KW - CONDITIONAL GRADIENT ALGORITHMS

KW - POINT

KW - First-order methods

UR - http://www.scopus.com/inward/record.url?scp=85114462441&partnerID=8YFLogxK

U2 - 10.1007/s10288-021-00493-y

DO - 10.1007/s10288-021-00493-y

M3 - Review

VL - 19

SP - 313

EP - 345

JO - 4 OR

JF - 4 OR

SN - 1619-4500

IS - 3

ER -

Frank–Wolfe and friends: a journey into projection-free first-order optimization methods

Abstract

Austrian Fields of Science 2012

Keywords

Access to Document

Other files and links

Cite this