Variance-Based Single-Call Proximal Extragradient Algorithms for Stochastic Mixed Variational Inequalities

In the study of stochastic variational inequalities, the extragradient algorithms attract much attention. However, such schemes require two evaluations of the expected mapping at each iteration in general. In this paper, we present several variance-based single-call proximal extragradient algorithms...

Full description

Saved in:
Bibliographic Details
Published in:Journal of optimization theory and applications Vol. 190; no. 2; pp. 393 - 427
Main Authors: Yang, Zhen-Ping, Lin, Gui-Hua
Format: Journal Article
Language:English
Published: New York Springer US 01.08.2021
Springer Nature B.V
Subjects:
Algorithms
Applications of Mathematics
Calculus of Variations and Optimal Control; Optimization
Convergence
Engineering
Inequalities
Iterative methods
Mapping
Mathematics
Mathematics and Statistics
Operations Research/Decision Theory
Optimization
Polynomials
Theory of Computation
Proximal extragradient algorithm
65K15
Stochastic mixed variational inequality
90C33
Variance reduction
Single-call
90C15
ISSN:0022-3239, 1573-2878
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the study of stochastic variational inequalities, the extragradient algorithms attract much attention. However, such schemes require two evaluations of the expected mapping at each iteration in general. In this paper, we present several variance-based single-call proximal extragradient algorithms for solving a class of stochastic mixed variational inequalities by aiming at alleviating the cost of an extragradient step. One salient feature of the proposed algorithms is that they require only one evaluation of the expected mapping at each iteration, and hence, the computation load may be significantly reduced. We show that the proposed algorithms can achieve sublinear ergodic convergence rate in terms of the restricted merit function. Furthermore, under the strongly Minty variational inequality condition, we derive some results related to convergence rate of the distance between iterates and solutions, the iteration and oracle complexities for the proposed algorithms when the sample size increases at a geometric or polynomial rate. Numerical experiments indicate that the proposed algorithms are quite competitive with some existing algorithms.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0022-3239
1573-2878
DOI:10.1007/s10957-021-01882-3