AIセキュリティポータル K Program
Zeroth-Order primal-dual Alternating Projection Gradient Algorithms for Nonconvex Minimax Problems with Coupled linear Constraints
Share
Abstract
In this paper, we study zeroth-order algorithms for nonconvex minimax problems with coupled linear constraints under the deterministic and stochastic settings, which have attracted wide attention in machine learning, signal processing and many other fields in recent years, e.g., adversarial attacks in resource allocation problems and network flow problems etc. We propose two single-loop algorithms, namely the zero-order primal-dual alternating projected gradient (ZO-PDAPG) algorithm and the zero-order regularized momentum primal-dual projected gradient algorithm (ZO-RMPDPG), for solving deterministic and stochastic nonconvex-(strongly) concave minimax problems with coupled linear constraints. The iteration complexity of the two proposed algorithms to obtain an $\varepsilon$-stationary point are proved to be $\mathcal{O}(\varepsilon ^{-2})$ (resp. $\mathcal{O}(\varepsilon ^{-4})$) for solving nonconvex-strongly concave (resp. nonconvex-concave) minimax problems with coupled linear constraints under deterministic settings and $\tilde{\mathcal{O}}(\varepsilon ^{-3})$ (resp. $\tilde{\mathcal{O}}(\varepsilon ^{-6.5})$) under stochastic settings respectively. To the best of our knowledge, they are the first two zeroth-order algorithms with iterative complexity guarantees for solving nonconvex-(strongly) concave minimax problems with coupled linear constraints under the deterministic and stochastic settings.
Gradient-free methods with inexact oracle for convex-concave stochastic saddle-point problem
Beznosikov A, Sadiev A, Gasnikov A
Published: 2020
Accelerated schemes for a class of variational inequalities
Chen Y M, Lan G H, Ouyang Y Y
Published: 2017
ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
Pin-Yu Chen, Huan Zhang, Yash Sharma, Jinfeng Yi, Cho-Jui Hsieh
Published: 2017.8.14
Convergence analysis of saddle point problems in time varying wireless systems—Control theoretical approach
Chen J, Lau V K N
Published: 2011
Efficient methods for structured nonconvex-nonconcave min-max optimization
Diakonikolas J, Daskalakis C, Jordan M
Published: 2021
Convergence rates of two-time-scale gradient descent-ascent dynamics for solving nonconvex min-max problems
Doan T
Published: 2022
On approximate stationary points of the regularized mathematical program with complementarity constraints
Dussault J P, Haddou M, Kadrani A, et al.
Published: 2020
Scaleable input gradient regularization for adversarial robustness
Finlay C, Oberman A
Published: 2021
A variational inequality perspective on generative adversarial networks
Gidel G, Berard H, Vignoud G, Vincent P, Lacoste-Julien S
Published: 2019
The landscape of the proximal point method for nonconvex-nonconcave minimax optimization
Grimmer B, Lu H, Worah P, Mirrokni V
Published: 2023
Decentralized learning for wireless communications and networking
Giannakis G B, Ling Q, Mateos G, Schizas I D, Zhu H
Published: 2016
Accelerated Zeroth-Order and First-Order Momentum Methods from Mini to Minimax Optimization
Huang F, Gao S, Pei J
Published: 2022
An accelerated inexact proximal point method for solving nonconvex-concave min-max problems
Kong W, Monteiro R D C
Published: 2021
Share