1 July 2021 in Inverse Problems
Alternating direction method of multipliers (ADMM) is a popular first-order method owing to its simplicity and efficiency. However, similar to other proximal splitting methods, the performance of ADMM degrades significantly when the scale of optimization problems to solve becomes large. In this paper, we consider combining ADMM with a class of variance-reduced stochastic gradient estimators for solving large-scale non-convex and non-smooth optimization problems. Global convergence of the generated sequence is established under the additional assumption that the object function satisfies Kurdyka-Łojasiewicz property. Numerical experiments on graph-guided fused lasso and computed tomography are presented to demonstrate the performance of the proposed methods.
11 December 2022 in Journal of Scientific Computing
31 March 2023 in Journal of the Operations Research Society of China
23 October 2024 in SIAM Journal on Matrix Analysis and Applications
The low-rank matrix recovery problem often arises in various fields, including signal processing, machine learning, and imaging science. The Riemannian gradient descent (RGD) algorithm has proven to be an efficient algorithm for solving this problem. In this paper, we present a preconditioned Riemannian gradient descent (PRGD) for low-rank matrix recovery. The preconditioner, noted for its simplicity and computational efficiency, is constructed by weighting the [Formula: see text]th entry of the gradient matrix according to the norms of the [Formula: see text]th row and the [Formula: see text]th column. We establish the theoretical recovery guarantee for PRGD under the restricted isometry property assumption. Experimental results indicate that PRGD can accelerate RGD by up to tenfold in solving low-rank matrix recovery problems such as matrix completion.
1 December 2021 in Applied Mathematics and Computation
24 January 2025 in Inverse Problems
The inverse radiation therapy planning problem is crucial in achieving tumoricidal doses while minimizing radiation-induced normal-tissue toxicity. Compared to conventional radiation therapy (with conventional dose rates), flash proton radiation therapy (with ultra-high dose rates) can provide additional normal tissue sparing. However, this groundbreaking advancement in radiation oncology introduces a challenging nonconvex and nonsmooth optimization problem. In this paper, we propose a stochastic three-operator splitting (STOS) algorithm to address the flash proton radiation therapy problem. We establish the convergence and convergence rates of the STOS algorithm under the nonconvex framework for both unbiased gradient estimators and variance-reduced gradient estimators. These stochastic gradient estimators include the most popular ones, such as SGD, SAGA, SARAH, and SAG, among others. Experimental results demonstrate that the flash proton radiation therapy plans obtained by the STOS algorithm can effectively kill tumors while better protecting normal tissues.