Linear Convergence of Forward-Backward Accelerated Algorithms without Knowledge of the Modulus of Strong Convexity

Abstract

SIAM Journal on Optimization, Volume 34, Issue 2, Page 2150-2168, June 2024.
Abstract. A significant milestone in modern gradient-based optimization was achieved with the development of Nesterov’s accelerated gradient descent (NAG) method. This forward-backward technique has been further advanced with the introduction of its proximal generalization, commonly known as the fast iterative shrinkage-thresholding algorithm (FISTA), which enjoys widespread application in image science and engineering. Nonetheless, it remains unclear whether both NAG and FISTA exhibit linear convergence for strongly convex functions. Remarkably, these algorithms demonstrate convergence without requiring any prior knowledge of strongly convex modulus, and this intriguing characteristic has been acknowledged as an open problem in the comprehensive review [A. Chambolle and T. Pock, Acta Numer., 25 (2016), pp. 161–319]. In this paper, we address this question by utilizing the high-resolution ordinary differential equation (ODE) framework. Expanding upon the established phase-space representation, we emphasize the distinctive approach employed in crafting the Lyapunov function, which involves a dynamically adapting coefficient of kinetic energy that evolves throughout the iterations. Furthermore, we highlight that the linear convergence of both NAG and FISTA is independent of the parameter [math]. Additionally, we demonstrate that the square of the proximal subgradient norm likewise advances toward linear convergence.