Narrow Results

Compared to the single-strategy particle swarm optimization (PSO) algorithm, the multi-strategy PSO shows potential advantages in solving complex optimization problems. In this study, a novel framework of the multi-strategy co-evolutionary PSO (M-PSO) is first proposed in which a matrix parameter pool scheme is introduced. In the scheme, multiple strategies are taken into account in the matrix parameter pool and new hybrid strategies can be generated. Then, the convergence analysis is made and the convergence conditions are provided for the co-evolutionary PSO framework when some operators are specified. Subsequently, based on the PSO framework, a novel multi-strategy co-evolutionary PSO is developed using $Q$-learning which is a classical reinforcement learning technique. In the proposed M-PSO, both the parameter optimization by the orthogonal method and the convergence conditions are embedded to improve the performance of the algorithm. Finally, the experiments are conducted on two test suites, CEC2017 and CEC2019, and the results indicate that M-PSO outperforms several meta-heuristic algorithms on most of the test problems.

We study properties of a modified memory gradient method, including the global convergence and rate of convergence. Numerical results show that modified memory gradient methods are effective in solving large-scale minimization problems.

A continuous approach using NCP function for approximating the solution of the max-cut problem is proposed. The max-cut problem is relaxed into an equivalent nonlinearly constrained continuous optimization problem and a feasible direction method without line searches is presented for generating an optimal solution of the relaxed continuous optimization problem. The convergence of the algorithm is proved. Numerical experiments and comparisons on some max-cut test problems show that we can get the satisfactory solution of max-cut problems with less computation time. Furthermore, this is the first time that the feasible direction method is combined with NCP function for solving max-cut problem, and similar idea can be generalized to other combinatorial optimization problems.

In this paper, we discuss the Expected Residual Minimization (ERM) method, which is to minimize the expected residue of some merit function for box constrained stochastic variational inequality problems (BSVIPs). This method provides a deterministic model, which formulates BSVIPs as an optimization problem. We first study the conditions under which the level sets of the ERM problem are bounded. Then, we show that solutions of the ERM formulation are robust in the sense that they may have a minimum sensitivity with respect to random parameter variations in BSVIPs. Since the integrality involved in the ERM problem is difficult to compute generally, we then employ sample average approximation method to solve it. Finally, we show that the global optimal solutions and generalized KKT points of the approximate problems converge to their counterparts of the ERM problem. On the other hand, as an application, we consider the model of European natural gas market under price uncertainty. Preliminary numerical experiments indicate that the proposed approach is applicable.

In this paper, we use matrix multisplitting with weighting parameters as the preconditioner of A. The optimal weighting parameters are determined by the approaching theory, and the scale of approaching is defined by F-norm, 2-norm, and ∞-norm, respectively. Base on these three minimize models, three algorithms are presented and the convergence theories are established. Finally, numerical examples show that the preconditioner with the optimal weighting parameters, which are obtained from minimize F-norm and 2-norm models, can improve the condition number of A effectively. Besides, general weighting parameters are more effective than non-negative weighting parameters.

This paper concentrates on improving the convergence properties of the relaxation schemes introduced by Kadrani et al. and Kanzow and Schwartz for mathematical program with equilibrium constraints (MPEC) by weakening the original constraint qualifications. It has been known that MPEC relaxed constant positive-linear dependence (MPEC-RCPLD) is a class of extremely weak constraint qualifications for MPEC, which can be strictly implied by MPEC relaxed constant rank constraint qualification (MPEC-RCRCQ) and MPEC relaxed constant positive-linear dependence (MPEC-rCPLD), of course also by the MPEC constant positive-linear dependence (MPEC-CPLD). We show that any accumulation point of stationary points of these two approximation problems is M-stationarity under the MPEC-RCPLD constraint qualification, and further show that the accumulation point can even be S-stationarity coupled with the asymptotically weak nondegeneracy condition.

In this paper, we consider nonseparable convex minimization models with quadratic coupling terms arised in many practical applications. We use a majorized indefinite proximal alternating direction method of multipliers (iPADMM) to solve this model. The indefiniteness of proximal matrices allows the function we actually solved to be no longer the majorization of the original function in each subproblem. While the convergence still can be guaranteed and larger stepsize is permitted which can speed up convergence. For this model, we analyze the global convergence of majorized iPADMM with two different techniques and the sublinear convergence rate in the nonergodic sense. Numerical experiments illustrate the advantages of the indefinite proximal matrices over the positive definite or the semi-definite proximal matrices.