Subgradient Methods for Saddle-Point Problems

被引:323
作者
Nedic, A. [1 ]
Ozdaglar, A. [2 ]
机构
[1] Univ Illinois, Dept Ind & Enterprise Syst Engn, Urbana, IL 61801 USA
[2] MIT, Dept Elect Engn & Comp Sci, Cambridge, MA 02139 USA
关键词
Saddle-point subgradient methods; Averaging; Approximate primal solutions; Primal-dual subgradient methods; Convergence rate; VARIATIONAL-INEQUALITIES; CONVEX-OPTIMIZATION; MONOTONE-OPERATORS; PRIMAL SOLUTIONS; GRADIENT METHOD; CONVERGENCE;
D O I
10.1007/s10957-009-9522-7
中图分类号
C93 [管理学]; O22 [运筹学];
学科分类号
070105 ; 12 ; 1201 ; 1202 ; 120202 ;
摘要
We study subgradient methods for computing the saddle points of a convex-concave function. Our motivation comes from networking applications where dual and primal-dual subgradient methods have attracted much attention in the design of decentralized network protocols. We first present a subgradient algorithm for generating approximate saddle points and provide per-iteration convergence rate estimates on the constructed solutions. We then focus on Lagrangian duality, where we consider a convex primal optimization problem and its Lagrangian dual problem, and generate approximate primal-dual optimal solutions as approximate saddle points of the Lagrangian function. We present a variation of our subgradient method under the Slater constraint qualification and provide stronger estimates on the convergence rate of the generated primal sequences. In particular, we provide bounds on the amount of feasibility violation and on the primal objective function values at the approximate solutions. Our algorithm is particularly well-suited for problems where the subgradient of the dual function cannot be evaluated easily (equivalently, the minimum of the Lagrangian function at a dual solution cannot be computed efficiently), thus impeding the use of dual subgradient methods.
引用
收藏
页码:205 / 228
页数:24
相关论文
共 32 条
[1]  
[Anonymous], 1985, NONDIFFERENTIABLE OP
[2]  
[Anonymous], 1958, Studies in Linear and Nonlinear Programming
[3]  
Arrow K. J., 1958, Stanford Mathematical Studies in the Social Sciences
[4]   Interior gradient and proximal methods for convex and conic optimization [J].
Auslender, A ;
Teboulle, M .
SIAM JOURNAL ON OPTIMIZATION, 2006, 16 (03) :697-725
[5]   Projected subgradient methods with non-Euclidean distances for non-differentiable convex minimization and variational inequalities [J].
Auslender, Alfred ;
Teboulle, Marc .
MATHEMATICAL PROGRAMMING, 2009, 120 (01) :27-48
[6]   Mirror descent and nonlinear projected subgradient methods for convex optimization [J].
Beck, A ;
Teboulle, M .
OPERATIONS RESEARCH LETTERS, 2003, 31 (03) :167-175
[7]  
Bertsekas D., 2003, Convex Analysis and Optimization
[8]  
Bertsekas D. P., 1999, Nonlinear programming
[9]   WEAK CONVERGENCE OF AN ERGODIC ITERATION FOR SOLUTION OF VARIATIONAL INEQUALITIES FOR MONOTONE OPERATORS IN HILBERT-SPACE [J].
BRUCK, RE .
JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS, 1977, 61 (01) :159-164
[10]   Layering as optimization decomposition: A mathematical theory of network architectures [J].
Chiang, Mung ;
Low, Steven H. ;
Calderbank, A. Robert ;
Doyle, John C. .
PROCEEDINGS OF THE IEEE, 2007, 95 (01) :255-312