FrankWolfe.jl: A High-Performance and Flexible Toolbox for Frank-Wolfe Algorithms and Conditional Gradients

被引：4

作者：

Besancon, Mathieu ^{[1
]}

Carderera, Alejandro ^{[1
,2
]}

Pokutta, Sebastian ^{[1
,3
]}

机构：

[1] Zuse Inst Berlin, D-14195 Berlin, Germany

[2] Georgia Inst Technol, Atlanta, GA 30308 USA

[3] Tech Univ Berlin, D-10623 Berlin, Germany

来源：

INFORMS JOURNAL ON COMPUTING | 2022年 / 34卷 / 05期

关键词：

first-order methods; optimization software; nonlinear programming;

D O I：

10.1287/ijoc.2022.1191

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

We present FrankWolfe.jl, an open-source implementation of several popular Frank-Wolfe and conditional gradients variants for first-order constrained optimization. The package is designed with flexibility and high performance in mind, allowing for easy extension and relying on few assumptions regarding the user-provided functions. It supports Julia's unique multiple dispatch feature, and it interfaces smoothly with generic linear optimization formulations using MathOptInterface.jl.

引用

页码：2611 / 2620

页数：10

共 28 条

[1]

[Anonymous], 2002, Matrix rank minimization with applications

[2]

Antonello N, 2018, PROXIMAL GRADIENT AL

[3] Mixed-Projection Conic Optimization: A New Paradigm for Modeling Rank Constraints [J].