Global Convergence Results Research Articles

Global Well-Posedness and Convergence Results to a 3D Regularized Boussinesq System in Sobolev Spaces

We consider a regularized periodic three-dimensional Boussinesq system. For a mean free initial temperature, we use the coupling between the velocity and temperature to close the energy estimates independently of time. This allows proving the existence of a global in time unique weak solution. Also, we establish that this solution depends continuously on the initial data. Moreover, we prove that this solution converges to a Leray-Hopf weak solution of the three-dimensional Boussinesq system as the regularizing parameter vanishes.

PPO-Clip Attains Global Optimality: Towards Deeper Understandings of Clipping

Proximal Policy Optimization algorithm employing a clipped surrogate objective (PPO-Clip) is a prominent exemplar of the policy optimization methods. However, despite its remarkable empirical success, PPO-Clip lacks theoretical substantiation to date. In this paper, we contribute to the field by establishing the first global convergence results of a PPO-Clip variant in both tabular and neural function approximation settings. Our findings highlight the O(1/√T ) min-iterate convergence rate specifically in the context of neural function approximation. We tackle the inherent challenges in analyzing PPO-Clip through three central concepts: (i) We introduce a generalized version of the PPO-Clip objective, illuminated by its connection with the hinge loss. (ii) Employing entropic mirror descent, we establish asymptotic convergence for tabular PPO-Clip with direct policy parameterization. (iii) Inspired by the tabular analysis, we streamline convergence analysis by introducing a two-step policy improvement approach. This decouples policy search from complex neural policy parameterization using a regression-based update scheme. Furthermore, we gain deeper insights into the efficacy of PPO-Clip by interpreting these generalized objectives. Our theoretical findings also mark the first characterization of the influence of the clipping mechanism on PPO-Clip convergence. Importantly, the clipping range affects only the pre-constant of the convergence rate.

On the paper “Augmented Lagrangian algorithms for solving the continuous nonlinear resource allocation problem”

In the paper Torrealba et al. (2021) an augmented Lagrangian algorithm was proposed for resource allocation problems with the intriguing characteristic that instead of solving the box-constrained augmented Lagrangian subproblem, they propose projecting the solution of the unconstrained subproblem onto such box. A global convergence result for the quadratic case was provided, however, this is somewhat counterintuitive, as in usual augmented Lagrangian theory, this strategy can fail in solving the augmented Lagrangian subproblems. In this note we investigate further this algorithm and we show that the proposed method may indeed fail when the Hessian of the quadratic is not a multiple of the identity. In the paper, it is not clear enough that two different projections are being used: one for obtaining their convergence results and other in their implementation. However, despite the lack of theoretical convergence, their strategy works remarkably well in some classes of problems; thus, we propose a hybrid method which uses their idea as a starting point heuristics, switching to a standard augmented Lagrangian method under certain conditions. Our contribution consists in presenting an efficient way of determining when the heuristics is failing to improve the KKT residual of the problem, suggesting that the heuristic procedure should be abandoned. Numerical results are provided showing that this strategy is successful in accelerating the standard method.

An efficient modified residual-based algorithm for large scale symmetric nonlinear equations by approximating successive iterated gradients

In this paper, a new descent approximate modified residual algorithm is developed to solve a large scale system of nonlinear symmetric equations, where the basic strategy to improve its numerical performance is to approximately compute the gradients and the difference of gradients. The error bounds of this approximation are presented, and in virtue of this approximation, a conjugate gradient algorithm for solving large-scale optimization problems in the literature is extended to solve the system of nonlinear symmetric equations without needs of computing and storing the Jacobian matrices or their approximate matrices. It is proved that the obtained search directions in our developed algorithm are sufficiently decent with respect to the so-called approximate modified residues. Under mild assumptions, global and local convergence results of the developed algorithm are proved. Numerical tests indicate that the developed algorithm outperforms the other similar ones available in the literature.

Multi-step inertial strictly contractive PRSM algorithms for convex programming problems with applications

The Peaceman–Rachford splitting method (PRSM) has been widely studied in recent years because of its excellent numerical performance, especially with inertial acceleration. In this paper, we propose a multi-step inertial strictly contractive PRSM (shortly, MISCPRSM) for solving convex optimization problems, which has not been studied before. This paper aims to give an exhaustive analysis of the value relationship between multiple inertial parameters. The second subproblem utilizes an additional customized indefinite proximal term in order to obtain a closed-form solution. The global convergence and complexity result of the introduced method are analyzed by using variational inequality and basic inequalities. Finally, through simulation and computed tomography experiments, the effectiveness and robustness of the MISCPRSM are proved, and the detailed values of inertial parameters are given.

Scaled fixed point algorithm for computing the matrix square root

This paper addresses the numerical solution of the matrix square root problem. Two fixed point iterations are proposed by rearranging the nonlinear matrix equation $A - X^2 = 0$ and incorporating a positive scaling parameter. The proposals only need to compute one matrix inverse and at most two matrix multiplications per iteration. A global convergence result is established. The numerical comparisons versus some existing methods from the literature, on several test problems, demonstrate the efficiency and effectiveness of our proposals.

Two-step inertial forward-reflected-backward splitting based algorithm for nonconvex mixed variational inequalities

This paper presents a modification of a recently studied Malitsky–Tam forward-reflected-backward splitting method with two-step inertial extrapolation to solve nonconvex mixed variational inequalities. We give global convergence results and nonasymptotic O(1/k) convergence rate of our proposed method under some appropriate conditions and present some mathematical justifications.

A derivative‐free projection method for nonlinear equations with non‐Lipschitz operator: Application to LASSO problem

In this paper, we introduce a derivative‐free iterative method for finding the solutions of convex constrained nonlinear equations (CCNE) using the projection strategy. The new approach is free from gradient evaluations at each iteration. Also, the search direction generated by the proposed method satisfies the sufficient descent property, which is independent of the line search. Compared with traditional methods for solving CCNE that assumes Lipschitz continuity and monotonicity to establish the global convergence result, an advantage of our proposed method is that the global convergence result does not require the assumption of Lipschitz continuity. Moreover, the underlying operator is assumed to be pseudomonotone, which is a milder condition than monotonicity. As an applications, we solve the LASSO problem in compressed sensing. Numerical experiments illustrate the performances of our proposed algorithm and provide a comparison with related algorithms.

Convergence analysis of an ALF-based nonconvex splitting algorithm with SQP structure

<p style='text-indent:20px;'>In this paper, by combining the splitting method of augmented Lagrange function (ALF) with the sequential quadratic programming (SQP) approximation, a novel ALF-based splitting algorithm with SQP structure is proposed for multi-block linear constrained nonconvex separable optimization. The new algorithm uses ALF-based splitting idea to decompose the original problem into several small-scale subproblems. Meanwhile, the SQP approximation and Armijo-type line search are used to solve some subproblems with smoothness concurrently. Under the conventional weak hypothesis, the decreasing property of ALF as merit function is obtained. Furthermore, the global convergence, strong convergence and convergence rate results of the new algorithm in general sense are given.</p>

A penalty function-based greedy diffusion search algorithm for the optimization of constrained nonlinear dynamical processes with discrete-valued input

This paper considers an optimization problem of nonlinear dynamical processes with discrete-valued input and continuous-time inequality constraints. This problem requires to determine an order and some instants. Since the order is a discrete variable and the constraints are complex, classical optimal control approaches cannot be used directly to solve this problem. To tackle this issue, a penalty function-based greedy diffusion search algorithm is proposed based on a novel transformation for each discrete value, a novel second-order smoothing technique, a $ l_1 $ penalty function, and a novel greedy rule. In the novel greedy rule, there exist only two parameters, which can be set in advance. Thus, unlike existing hybrid approaches, no parameters is required adjusted in the penalty function-based greedy diffusion search algorithm. Following that, the global convergence results of the proposed algorithm are established. Finally, an energy management problem in an electrical vehicle is used to illustrate the effectiveness of the proposed algorithm. The numerical simulation results show that the algorithm proposed by this paper can obtain a better solution with less time-consuming, fewer iterations, and fewer errors compared with existing methods.

An inexact projected gradient method with rounding and lifting by nonlinear programming for solving rank-one semidefinite relaxation of polynomial optimization

We consider solving high-order and tight semidefinite programming (SDP) relaxations of nonconvex polynomial optimization problems (POPs) that often admit degenerate rank-one optimal solutions. Instead of solving the SDP alone, we propose a new algorithmic framework that blends local search using the nonconvex POP into global descent using the convex SDP. In particular, we first design a globally convergent inexact projected gradient method (iPGM) for solving the SDP that serves as the backbone of our framework. We then accelerate iPGM by taking long, but safeguarded, rank-one steps generated by fast nonlinear programming algorithms. We prove that the new framework is still globally convergent for solving the SDP. To solve the iPGM subproblem of projecting a given point onto the feasible set of the SDP, we design a two-phase algorithm with phase one using a symmetric Gauss–Seidel based accelerated proximal gradient method (sGS-APG) to generate a good initial point, and phase two using a modified limited-memory BFGS (L-BFGS) method to obtain an accurate solution. We analyze the convergence for both phases and establish a novel global convergence result for the modified L-BFGS that does not require the objective function to be twice continuously differentiable. We conduct numerical experiments for solving second-order SDP relaxations arising from a diverse set of POPs. Our framework demonstrates state-of-the-art efficiency, scalability, and robustness in solving degenerate SDPs to high accuracy, even in the presence of millions of equality constraints.

Extra parameters for conjugate gradient method for removing impulse noise images

In most applications, denoising a photograph is essential for most image processing processes. In this study, a conjugate gradient method for impulse noise reduction is presented, which is based on a two-phase system. The noise candidates are found with an adaptive (center-weighted) median filter, and then restored with the CG method, which entails the minimization of an edge-preserving regularization functional. The fact that the search direction created at each iteration is descending is one of the most appealing characteristics of the suggested technique. Its global convergence result might be obtained if a strong Wolfe line search is used. The spectral conjugate gradient approach for impulse noise reduction is demonstrated through numerical experiments.

A New Class of Optimization Methods Based on Coefficient Conjugate Gradient

The coefficient conjugate serves as the foundation for many conjugate gradient methods. The quadratic model is used to derive a novel coefficient conjugate in this study. Its global convergence result might be produced under Wolfe line search circumstances. The conjugate gradient method’s performance for unconstrained optimization problems is demonstrated through numerical tests.

A new smoothing method for nonlinear complementarity problems involving P0-function

In this paper, we present a family of smoothing methods to solve nonlinear complementarity problems (NCPs) involving P0-function. Several regularization or approximation techniques like Fisher-Burmeister’s method, interior-point methods (IPMs) approaches, or smoothing methods already exist. All the corresponding methods solve a sequence of nonlinear systems of equations and depend on parameters that are difficult to drive to zero. The main novelty of our approach is to consider the smoothing parameters as variables that converge by themselves to zero. We do not need any complicated updating strategy, and then obtain nonparametric algorithms. We prove some global and local convergence results and present several numerical experiments, comparisons, and applications that show the efficiency of our approach.

Adaptive parameter based matrix splitting iteration method for the large and sparse linear systems

For the large and sparse linear systems, we utilize the efficient splittings of the system matrix and introduce an intermediate variable. The main contribution of this paper is that the adaptive parameter based matrix splitting iteration method is constructed from the view of numerical optimization and hybrid procedure to solve the derived equation instead. The novel method adopts the prediction and correction two-step iteration and the global convergence results are established under some suitable conditions. Compared with some well-known methods, numerical experiments show the efficiency and effectiveness of the new proposal with application to the three-dimensional convection-diffusion equation and the image restoration problems.

Bregman Finito/MISO for Nonconvex Regularized Finite Sum Minimization without Lipschitz Gradient Continuity

We introduce two algorithms for nonconvex regularized finite sum minimization, where typical Lipschitz differentiability assumptions are relaxed to the notion of relative smoothness. The first one is a Bregman extension of Finito/MISO, studied for fully nonconvex problems when the sampling is random, or under convexity of the nonsmooth term when it is essentially cyclic. The second algorithm is a low-memory variant, in the spirit of SVRG and SARAH, that also allows for fully nonconvex formulations. Our analysis is made remarkably simple by employing a Bregman Moreau envelope as Lyapunov function. In the randomized case, linear convergence is established when the cost function is strongly convex, yet with no convexity requirements on the individual functions in the sum. For the essentially cyclic and low-memory variants, global and linear convergence results are established when the cost function satisfies the Kurdyka-\L ojasiewicz property.

Higher order difference equations with homogeneous governing functions nonincreasing in each variable with unbounded solutions

By using a comparison method and some difference inequalities we show that the following higher order difference equation xn+k=1f(xn+k−1,…,xn),n∈N,\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ x_{n+k}=\\frac{1}{f(x_{n+k-1},\\ldots ,x_{n})},\\quad n\\in{\\mathbb{N}},$$\\end{document} where kin{mathbb{N}}, f:[0,+infty )^{k}to [0,+infty ) is a homogeneous function of order strictly bigger than one, which is nondecreasing in each variable and satisfies some additional conditions, has unbounded solutions, presenting a large class of such equations. The class can be used as a useful counterexample in dealing with the boundedness character of solutions to some difference equations. Some analyses related to such equations and a global convergence result are also given.

A study of Liu-Storey conjugate gradient methods for vector optimization

This work presents a study of Liu-Storey (LS) nonlinear conjugate gradient (CG) methods to solve vector optimization problems. Three variants of the LS-CG method originally designed to solve single-objective problems are extended to the vector setting. The first algorithm restricts the LS conjugate parameter to be nonnegative and use a sufficiently accurate line search satisfying the (vector) standard Wolfe conditions. The second algorithm combines a modification in the LS conjugate parameter with a line search satisfying the (vector) strong Wolfe conditions. The third algorithm consists of a combination of the LS conjugate parameter with a new Armijo-type line search (to be proposed here for the vector setting). Global convergence results and numerical experiments are presented.

A Liu-Storey-type conjugate gradient method for unconstrained minimization problem with application in motion control

Conjugate gradient methods have played a vital role in finding the minimizers of large-scale unconstrained optimization problems due to the simplicity of their iteration, convergence properties and their low memory requirements. Based on the Liu-Storey conjugate gradient method, in this paper, we present a Liu-Storey type method for finding the minimizers of large-scale unconstrained optimization problems. The direction of the proposed method is constructed in such a way that the sufficient descent condition is satisfied. Furthermore, we establish the global convergence result of the method under the standard Wolfe and Armijo-like line searches. Numerical findings indicate that our presented approach is efficient and robust in solving large-scale test problems. In addition, an application of the method is explored.

Distributed 3D Navigation of Swarms of Non-Holonomic UAVs for Coverage of Unsteady Environmental Boundaries

A team of non-holonomic constant-speed under-actuated unmanned aerial vehicles (UAVs) with lower-limited turning radii travel in 3D. The space hosts an unknown and unpredictably varying scalar environmental field. A space direction is given; this direction and the coordinate along it are conditionally termed as the “vertical” and “altitude”, respectively. All UAVs should arrive at the moving and deforming isosurface where the field assumes a given value. They also should evenly distribute themselves over a pre-specified range of the “altitudes” and repeatedly encircle the entirety of the isosurface while remaining on it, each at its own altitude. Every UAV measures only the field intensity at the current location and both the Euclidean and altitudinal distances to the objects (including the top and bottom of the altitudinal range) within a finite range of visibility and has access to its own speed and the vertical direction. The UAVs carry no communication facilities, are anonymous to one another, and cannot play distinct roles in the team. A distributed control law is presented that solves this mission under minimal and partly inevitable assumptions. This law is justified by a mathematically rigorous global convergence result; computer simulation tests confirm its performance.