Splitting forward-backward penalty scheme for constrained variational problems

MaRDI QIDQ3186418zbMATH OpenFDO

Authors Marc-Olivier Czarnecki, Nahla Noun, Juan Peypouquet

Publication date 9 August 2016

Full work available at URL https://arxiv.org/abs/1408.0974, http://www.heldermann.de/JCA/JCA23/JCA232/jca23020.htm

zbMATH Keywords

constrained convex optimizationforward-backward algorithmhierarchical optimizationpenalization methods

Mathematics Subject Classification ID

Applications of functional analysis in optimization, convex analysis, mathematical programming, economics (46N10) Variational inequalities (49J40) Numerical mathematical programming methods (65K05) Numerical optimization and variational techniques (65K10)

Abstract: We study a forward backward splitting algorithm that solves the variational inequality �egin{equation*} A x +

abla Phi(x)+ N_C (x) i 0 end{equation*} where $H$ is a real Hilbert space, $A:HightrightarrowsH$ is a maximal monotone operator, $Phi:HomathbbR$ is a smooth convex function, and $N_C$ is the outward normal cone to a closed convex set $CsubsetH$. The constraint set $C$ is represented as the intersection of the sets of minima of two convex penalization function $Ps{i}_1:HomathbbR$ and $Ps{i}_2:HomathbbRcup+infty$. The function $Ps{i}_1$ is smooth, the function $Ps{i}_2$ is proper and lower semicontinuous. Given a sequence $$ of penalization parameters which tends to infinity, and a sequence of positive time steps $(lambd{a}_n)$, the algorithm ${\displaystyle }$ left{�egin{array}{rcl} x_1 & in & H,\ x_{n+1} & = & (I+lambda_n A+lambda_n�eta_npartialPsi_2)^{-1}(x_n-lambda_n ablaPhi(x_n)-lambda_n�eta_n

ablaPsi_1(x_n)), ngeq 1. end{array} ight. ${\displaystyle }$ performs forward steps on the smooth parts and backward steps on the other parts. Under suitable assumptions, we obtain weak ergodic convergence of the sequence $(x_n)$ to a solution of the variational inequality. Convergence is strong when either $A$ is strongly monotone or $Phi$ is strongly convex. We also obtain weak convergence of the whole sequence $(x_n)$ when $A$ is the subdifferential of a proper lower-semicontinuous convex function. This provides a unified setting for several classical and more recent results, in the line of historical research on continuous and discrete gradient-like systems.

Recommendations

• Coupling Forward-Backward with Penalty Schemes and Parallel Splitting for Constrained Variational Inequalities
• scientific article; zbMATH DE number 7436312
• scientific article; zbMATH DE number 1749789
• Applications of a Splitting Algorithm to Decomposition in Convex Programming and Variational Inequalities
• A penalization-gradient algorithm for variational inequalities

Cited in

(9)

• Iterative regularization via dual diagonal descent
• Modified forward-backward splitting method for variational inclusions
• Generalized forward-backward splitting with penalization for monotone inclusion problems
• Coupling Forward-Backward with Penalty Schemes and Parallel Splitting for Constrained Variational Inequalities
• A sequential constraint method for solving variational inequality over the intersection of fixed point sets
• Deep unfolding of a proximal interior point method for image restoration
• A penalization-gradient algorithm for variational inequalities
• Accelerated iterative regularization via dual diagonal descent
• Asymptotic behavior of penalty dynamics for constrained variational inequalities

This page was built for publication: Splitting forward-backward penalty scheme for constrained variational problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3186418)