A Modified Popov Algorithm for Non-Monotone and Non-Lipschitzian Stochastic Variational Inequalities

In this paper, we propose a modified Popov algorithm with variable sample-size for solving non-monotone and non-Lipschitzian stochastic variational inequality problems. It is inspired by an improved Popov algorithm for solving deterministic variational inequality problems, proposed by Malitsky and S...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Journal of optimization theory and applications Ročník 205; číslo 3; s. 54
Hlavní autoři: Li, Jun-zuo, Xiao, Yi-bin
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York Springer US 01.06.2025
Springer Nature B.V
Témata:
Algorithms
Applications of Mathematics
Calculus of Variations and Optimal Control; Optimization
Constraints
Convergence
Engineering
Half spaces
Inequalities
Mathematics
Mathematics and Statistics
Operations Research/Decision Theory
Optimization
Theory of Computation
Stochastic variational inequality
65K15
90C33
Stochastic approximation
Modified Popov method
Variable sample-size
62L20
ISSN:0022-3239, 1573-2878
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper, we propose a modified Popov algorithm with variable sample-size for solving non-monotone and non-Lipschitzian stochastic variational inequality problems. It is inspired by an improved Popov algorithm for solving deterministic variational inequality problems, proposed by Malitsky and Semenov (Malitsky and Semenov in Cybern. Syst. Anal. 50:271–277, 2014), and a stochastic Popov method for solving stochastic variational inequality problems, developed by Vankov et al. (Last iterate convergence of Popov method for non-monotone stochastic variational inequalities, 2023). In contrast to the stochastic Popov method, the proposed algorithm incorporates a variable sample-size strategy and, crucially, conducts a projection onto a half-space followed by a projection onto the constraint set in each iteration, rather than two projections onto the constraint set. These modifications have the potential to reduce computational cost, particularly when the computation of projection onto the constraint set is expensive, and thus improve performance. Subsequently, we discuss the almost sure convergence of the algorithm, its sublinear and linear convergence rate, and the oracle complexity. Finally, we present numerical experiments to demonstrate the competitiveness of the algorithm and further apply it to solve a signal estimation problem.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0022-3239
1573-2878
DOI:10.1007/s10957-025-02676-7