Geometric characterizations of Lipschitz stability for convex optimization problems

DOI10.1137/24M1637532MaRDI QIDQ6974659zbMATH OpenFDO

Authors T. T. A. Nghia

Publication date 5 June 2025

Published in SIAM Journal on Optimization (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1137/24M1637532

zbMATH Keywords

convex optimizationgroup Lasso problemsLipschitz stabilitynuclear norm minimization problemssecond-order analysistilt stabilityvariational analysis and nonsmooth optimization

Mathematics Subject Classification ID

Nonsmooth analysis (49J52) Set-valued and variational analysis (49J53) Sensitivity, stability, well-posedness (49K40) Convex programming (90C25) Sensitivity, stability, parametric optimization (90C31)

Review text

Tilt and Lipschitz stability are investigated for unconstrained convex optimization problems in a finite dimensional setting. A characterization of tilt stability is obtained via the second subderivative. When the objective function is the sum of a convex and a twice continuously differentiable function, a geometric characterization is obtained making use of a suitable quadratic growth condition and a relative approximation of the subdifferential mapping of the convex term. Applications are considered to the regularized optimization problem related to signal recovering from noisy observations. In this case, a sufficient condition is found such that the solution mapping is single-valued and locally Lipschitz continuous. When the convex term is the \(l_1 / l_2\) norm, the condition is shown to be necessary, and single-valuedness is equivalent to Lipschitz continuity. Applications are also considered to characterize tilt stability of nuclear norm minimization problems, and to prove again equivalence between single-valuedness and Lipschitz continuity of the solution mapping. Several relevant examples are discussed in detail.

Cites work

• A Generalized Newton Method for Subgradient Systems
• A unified approach to error bounds for structured convex optimization problems
• Augmented Lagrangians and hidden convexity in sufficient conditions for local optimality
• Characterization of metric regularity of subdifferentials
• Characterization of the subdifferential of some matrix norms
• Characterization of tilt stability via subgradient graphical derivative with applications to nonlinear programming
• Characterizations of tilt-stable minimizers in second-order cone programming
• Complete characterizations of tilt stability in nonlinear programming under weakest qualification conditions
• Convex Analysis
• Differentiating Nonsmooth Solutions to Parametric Monotone Inclusion Problems
• Exact matrix completion via convex optimization
• From error bounds to the complexity of first-order descent methods for convex functions
• Full stability in finite-dimensional optimization
• Generalized Newton algorithms for tilt-stable minimizers in nonsmooth optimization
• High-dimensional statistics. A non-asymptotic viewpoint
• scientific article; zbMATH DE number 176973 (Why is no real title available?)
• scientific article; zbMATH DE number 1502618 (Why is no real title available?)
• Implicit Functions and Solution Mappings
• Isolated calmness of solution mappings and exact recovery conditions for nuclear norm optimization problems
• LASSO Reloaded: A Variational Analysis Perspective with Applications to Compressed Sensing
• Model Selection and Estimation in Regression with Grouped Variables
• On (local) analysis of multifunctions via subspaces contained in graphs of generalized derivatives
• On a globally convergent semismooth^* Newton method in nonsmooth nonconvex optimization
• On a Semismooth* Newton Method for Solving Generalized Equations
• On Lipschitzian properties of implicit multifunctions
• Partial Smoothness, Tilt Stability, and Generalized Hessians
• Quadratic growth and critical point stability of semi-algebraic functions
• Quadratic Growth and Strong Metric Subregularity of the Subdifferential via Subgradient Graphical Derivative
• Quadratic growth conditions for convex matrix optimization problems associated with spectral functions
• Regularization and Variable Selection Via the Elastic Net
• Second-order characterizations of tilt stability with applications to nonlinear programming
• Second-order growth, tilt stability, and metric regularity of the subdifferential
• Second-Order Subdifferential Calculus with Applications to Tilt Stability in Optimization
• Sharp, strong and unique minimizers for low complexity robust recovery
• Simple bounds for recovering low-complexity models
• Solution uniqueness of convex optimization problems via the radial cone
• Some continuity properties of polyhedral multifunctions
• Strongly Regular Generalized Equations
• The convex geometry of linear inverse problems
• The degrees of freedom of partly smooth regularizers
• The Lasso problem and uniqueness
• The smooth-Lasso and other \(\ell _{1}+\ell _{2}\)-penalized methods
• The Strong Second-Order Sufficient Condition and Constraint Nondegeneracy in Nonlinear Semidefinite Programming and Their Implications
• Tilt Stability of a Local Minimum
• Tilt stability, uniform quadratic growth, and strong metric regularity of the subdifferential
• Upper-Lipschitz multifunctions and inverse subdifferentials
• Variational Analysis

Cited in

(3)

• Tilt stability of Ky-Fan -norm composite optimization
• The geometry and well-posedness of sparse regularized linear regression
• Tilt stability on Riemannian manifolds with application to convergence analysis of generalized Riemannian Newton method

This page was built for publication: Geometric characterizations of Lipschitz stability for convex optimization problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6974659)