Convergence of the forward-backward algorithm: beyond the worst-case with the help of geometry

被引:13
作者
Garrigos, Guillaume [1 ]
Rosasco, Lorenzo [2 ,3 ]
Villa, Silvia [4 ]
机构
[1] Univ Paris & Sorbonne Univ, Lab Probabilites Stat & Modelisat, CNRS, F-75013 Paris, France
[2] Univ Genoa, MaLGa Ctr, DIBRIS, Via Dodecaneso 35, I-16146 Genoa, Italy
[3] MIT, CBMM, Ist Italiano Tecnol, I-16146 Genoa, Italy
[4] Univ Genoa, MaLGa Ctr, Dipartimento Matemat, Via Dodecaneso 35, I-16146 Genoa, Italy
基金
欧洲研究理事会; 欧盟地平线“2020”;
关键词
Forward Backward algorithm; Convergence rates; Conditioned functions; Lojasiewicz property; Inverse problems; Source condition; ERROR-BOUNDS; DESCENT METHODS; LINEAR CONVERGENCE; THRESHOLDING ALGORITHM; METRIC REGULARITY; CONVEX; SYSTEMS; OPTIMIZATION; STABILITY; NONCONVEX;
D O I
10.1007/s10107-022-01809-4
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
We provide a comprehensive study of the convergence of the forward-backward algorithm under suitable geometric conditions, such as conditioning or Lojasiewicz properties. These geometrical notions are usually local by nature, and may fail to describe the fine geometry of objective functions relevant in inverse problems and signal processing, that have a nice behaviour on manifolds, or sets open with respect to a weak topology. Motivated by this observation, we revisit those geometric notions over arbitrary sets. In turn, this allows us to present several new results as well as collect in a unified view a variety of results scattered in the literature. Our contributions include the analysis of infinite dimensional convex minimization problems, showing the first Lojasiewicz inequality for a quadratic function associated to a compact operator, and the derivation of new linear rates for problems arising from inverse problems with low-complexity priors. Our approach allows to establish unexpected connections between geometry and a priori conditions in inverse problems, such as source conditions, or restricted isometry properties.
引用
收藏
页码:937 / 996
页数:60
相关论文
共 106 条
  • [41] Engl HW., 2000, REGULARIZATION INVER
  • [42] SENSITIVITY ANALYSIS FOR MIRROR-STRATIFIABLE CONVEX FUNCTIONS
    Fadili, Jalal
    Malick, Jerome
    Peyre, Gabriel
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2018, 28 (04) : 2975 - 3000
  • [43] Federer H., 1959, T AM MATH SOC, V93, P418, DOI 10.1090/S0002-9947-1959-0110078-1
  • [44] FINITE TERMINATION OF THE PROXIMAL POINT ALGORITHM
    FERRIS, MC
    [J]. MATHEMATICAL PROGRAMMING, 1991, 50 (03) : 359 - 366
  • [45] Foucart S., 2017, Bull. Amer. Math., V54, P151, DOI 10.1007/978-0-8176-4948-7
  • [46] Splitting Methods with Variable Metric for Kurdyka-Aojasiewicz Functions and General Convergence Rates
    Frankel, Pierre
    Garrigos, Guillaume
    Peypouquet, Juan
    [J]. JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS, 2015, 165 (03) : 874 - 900
  • [47] Garrigos G., 2015, THESIS
  • [48] Thresholding gradient methods in Hilbert spaces: support identification and linear convergence
    Garrigos, Guillaume
    Rosasco, Lorenzo
    Villa, Silvia
    [J]. ESAIM-CONTROL OPTIMISATION AND CALCULUS OF VARIATIONS, 2020, 26
  • [49] Goldstein AA, 1962, Numerische Mathematik, V4, P146, DOI 10.1007/BF01386306
  • [50] Groetsch CW., 1977, GEN INVERSES LINEAR