Abstract:The Performance Estimation Problem methodology makes it possible to determine the exact worst-case performance of an optimization method. In this work, we generalize this framework to first-order methods involving linear operators. This extension requires an explicit formulation of interpolation conditions for those linear operators. We consider the class of linear operators $\mathcal{M}:x \mapsto Mx$ where matrix $M$ has bounded singular values, and the class of linear operators where $M$ is symmetric and has bounded eigenvalues. We describe interpolation conditions for these classes, i.e. necessary and sufficient conditions that, given a list of pairs $\{(x_i,y_i)\}$, characterize the existence of a linear operator mapping $x_i$ to $y_i$ for all $i$. Using these conditions, we first identify the exact worst-case behavior of the gradient method applied to the composed objective $h\circ \mathcal{M}$, and observe that it always corresponds to $\mathcal{M}$ being a scaling operator. We then investigate the Chambolle-Pock method applied to $f+g\circ \mathcal{M}$, and improve the existing analysis to obtain a proof of the exact convergence rate of the primal-dual gap. In addition, we study how this method behaves on Lipschitz convex functions, and obtain a numerical convergence rate for the primal accuracy of the last iterate. We also show numerically that averaging iterates is beneficial in this setting.

Estimates related to the iterates of positive linear operators and their multidimensional analogues

The Equivalence Relations Between the Best Approximation and Linear Operators Approximation in a Besov Space

Properties of Positive Linear Operators Connected with Squared Fundamental Functions

Difference of Some Positive Linear Approximation Operators for Higher-Order Derivatives

Approximation of discontinuous functions by positive linear operators. A probabilistic approach

Initial l2×·· ·×l2 bounds for multilinear operators

Simultaneous approximation by operators of exponential type

Approximation on Variable Exponent Spaces by Linear Integral Operators

Initial 𝐿²×⋯×𝐿² Bounds for Multilinear Operators

Approximation of operators on real line

Geometric series of positive linear operators and inverse Voronovskaya theorem

Initial $L^2\times\cdots\times L^2 $ bounds for multilinear operators

Approximation properties of a certain modification of Durrmeyer operators

Direct Estimates of the Rate of Approximation by the Kantorovich Operator in Variable Exponent Lebesgue Spaces

Conditionally positive definiteness in operator theory

Analysis of Approximation by Linear Operators on VariableLρp(·)Spaces and Applications in Learning Theory

Approximation by α-Bernstein-Kantorovich Type Operators for Functions of One and Two Variables

Interpolation Conditions for Linear Operators and Applications to Performance Estimation Problems

Approximating Iterations for Nonexpansive and Maximal Monotone Operators

On Quantitative Approximation

INITIAL L2 x middot middot middot x L2 BOUNDS FOR MULTILINEAR OPERATORS