Automated tight Lyapunov analysis for first-order methods

被引:0
作者
Upadhyaya, Manu [1 ]
Banert, Sebastian [1 ]
Taylor, Adrien B. [2 ,3 ,4 ,5 ]
Giselsson, Pontus [1 ]
机构
[1] Lund Univ, Dept Automat Control, Lund, Sweden
[2] INRIA, SIERRA Projet Team, Paris, France
[3] DI Ecole Normale Super, Paris, France
[4] CNRS, Paris, France
[5] PSL Res Univ, Paris, France
基金
欧洲研究理事会; 瑞典研究理事会;
关键词
Performance estimation; Convex optimization; First-order methods; Quadratic constraints; Lyapunov functions; Semidefinite programming; WORST-CASE PERFORMANCE; OPTIMIZATION ALGORITHMS; SPLITTING METHOD; DESIGN;
D O I
10.1007/s10107-024-02061-8
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
We present a methodology for establishing the existence of quadratic Lyapunov inequalities for a wide range of first-order methods used to solve convex optimization problems. In particular, we consider (i) classes of optimization problems of finite-sum form with (possibly strongly) convex and possibly smooth functional components, (ii) first-order methods that can be written as a linear system on state-space form in feedback interconnection with the subdifferentials of the functional components of the objective function, and (iii) quadratic Lyapunov inequalities that can be used to draw convergence conclusions. We present a necessary and sufficient condition for the existence of a quadratic Lyapunov inequality within a predefined class of Lyapunov inequalities, which amounts to solving a small-sized semidefinite program. We showcase our methodology on several first-order methods that fit the framework. Most notably, our methodology allows us to significantly extend the region of parameter choices that allow for duality gap convergence in the Chambolle-Pock method when the linear operator is the identity mapping.
引用
收藏
页码:133 / 170
页数:38
相关论文
共 44 条
[1]  
[Anonymous], 2015, A three-operator splitting scheme and its optimization applications
[2]   DATA-DRIVEN NONSMOOTH OPTIMIZATION [J].
Banert, Sebastian ;
Ringh, Axel ;
Adler, Jonas ;
Karlsson, Johan ;
Oktem, Ozan .
SIAM JOURNAL ON OPTIMIZATION, 2020, 30 (01) :102-131
[3]   Fixing and extending some recent results on the ADMM algorithm [J].
Banert, Sebastian ;
Bot, Radu Ioan ;
Csetnek, Ernoe Robert .
NUMERICAL ALGORITHMS, 2021, 86 (03) :1303-1325
[4]   Principled analyses and design of first-order methods with inexact proximal operators [J].
Barre, Mathieu ;
Taylor, Adrien B. ;
Bach, Francis .
MATHEMATICAL PROGRAMMING, 2023, 201 (1-2) :185-230
[5]  
Bauschke HH, 2011, CMS BOOKS MATH, P1, DOI 10.1007/978-1-4419-9467-7
[6]  
Beck A. :., 2017, MOS-SIAM SER OPTIMIZ, DOI [DOI 10.1137/1.9781611974997, 10.1137/1.9781611974997]
[7]  
Bot RI., 2010, CONJUGATE DUALITY CO, DOI DOI 10.1007/978-3-642-04900-2
[8]   On the ergodic convergence rates of a first-order primal-dual algorithm [J].
Chambolle, Antonin ;
Pock, Thomas .
MATHEMATICAL PROGRAMMING, 2016, 159 (1-2) :253-287
[9]   A First-Order Primal-Dual Algorithm for Convex Problems with Applications to Imaging [J].
Chambolle, Antonin ;
Pock, Thomas .
JOURNAL OF MATHEMATICAL IMAGING AND VISION, 2011, 40 (01) :120-145
[10]  
Chen C., 1999, Linear system theory and design, V3rd