一个新的填充函数及其在数据拟合问题中的应用

填充函数法是求解全局优化问题的有效方法之一,针对无约束优化问题,提出一个新的连续可微的无参数填充函数,证明其相关性质并给出相应的算法,数值实验结果表明该算法是有效可行的。同时用此填充函数对切削温度实验数据这一拟合实例进行求解,与已有的最小二乘法和遗传算法的求解结果相比较,拟合效果较好。     

A modified EM algorithm for mixture models based on Bregman divergence

The EM algorithm is a sophisticated method for estimating statistical models with hidden variables based on the Kullback–Leibler divergence. A natural extension of the Kullback–Leibler divergence is given by a class of Bregman divergences, which in general enjoy robustness to contamination data in statistical inference. In this paper, a modification of the EM algorithm based on the Bregman divergence is proposed for estimating finite mixture models. The proposed algorithm is geometrically interpreted as a sequence of projections induced from the Bregman divergence. Since a rigorous algorithm includes a nonlinear optimization procedure, two simplification methods for reducing computational difficulty are also discussed from a geometrical viewpoint. Numerical experiments on a toy problem are carried out to confirm appropriateness of the simplifications.     

Optimal Noncollapsing Space-Filling Designs for Irregular Experimental Regions

Space-filling and noncollapsing are two important properties in designing computer experiments. We study how the noncollapsing, space-filling designs for irregular experimental regions can be generated efficiently by the proposed metaheuristic methods. We solve this optimal design problem using variants of the discrete particle swarm optimization (DPSO) approaches. Numerical results, including an application in data center thermal management, are used to illustrate the performances of the proposed algorithms. Based on these numerical results, we assert that the most efficient approach is to reformulate the target optimal design problem as a constrained optimization problem and then use a modified DPSO to solve the constrained optimization problem.     

Optimization algorithm for learning consistent belief rule-base from examples

A belief rule-based inference approach and its corresponding optimization algorithm deal with a rule-base with a belief structure called a belief rule base (BRB) that forms a basis in the inference mechanism. In this paper, a new learning method is proposed based on the given sample data for optimally generating a consistent BRB. The focus is given on the consistency of BRB knowing that the consistency conditions are often violated if the system is generated from real world data. The measurement of BRB inconsistency is incorporated in the objective function of the optimization algorithm. This process is formulated as a non-linear constraint optimization problem and solved using the optimization tool provided in MATLAB. A numerical example is demonstrated the effectiveness of the proposed algorithm.     

An artificial neural network meta-model for constrained simulation optimization

This paper presents artificial neural network (ANN) meta-models for expensive continuous simulation optimization (SO) with stochastic constraints. These meta-models are used within a sequential experimental design to approximate the objective function and the stochastic constraints. To capture the non-linear nature of the ANN, the SO problem is iteratively approximated via non-linear programming problems whose (near) optimal solutions obtain estimates of the global optima. Following the optimization step, a cutting plane-relaxation scheme is invoked to drop uninformative estimates of the global optima from the experimental design. This approximation is iterated until a terminating condition is met. To study the robustness and efficiency of the proposed algorithm, a realistic inventory model is used; the results are compared with those of the OptQuest optimization package. These numerical results indicate that the proposed meta-model-based algorithm performs quite competitively while requiring slightly fewer simulation observations.     

A multi-objective memetic algorithm based on decomposition for big optimization problems

When solving multi-objective optimization problems (MOPs) with big data, traditional multi-objective evolutionary algorithms (MOEAs) meet challenges because they demand high computational costs that cannot satisfy the demands of online data processing involving optimization. The gradient heuristic optimization methods show great potential in solving large scale numerical optimization problems with acceptable computational costs. However, some intrinsic limitations make them unsuitable for searching for the Pareto fronts. It is believed that the combination of these two types of methods can deal with big MOPs with less computational cost. The main contribution of this paper is that a multi-objective memetic algorithm based on decomposition for big optimization problems (MOMA/D-BigOpt) is proposed and a gradient-based local search operator is embedded in MOMA/D-BigOpt. In the experiments, MOMA/D-BigOpt is tested on the multi-objective big optimization problems with thousands of variables. We also combine the local search operator with other widely used MOEAs to verify its effectiveness. The experimental results show that the proposed algorithm outperforms MOEAs without the gradient heuristic local search operator.     

An iterative numerical algorithm for modeling a class of Wiener nonlinear systems

This letter presents an iterative estimation algorithm for modeling a class of output nonlinear systems. The basic idea is to derive an estimation model and to solve an optimization problem using the gradient search. The proposed iterative numerical algorithm can estimate the parameters of a class of Wiener nonlinear systems from input–output measurement data. The proposed algorithm has faster convergence rates compared with the stochastic gradient algorithm. The numerical simulation results indicate that the proposed algorithm works well.     

A simplex-based labelling algorithm for the linear fractional assignment problem

This paper constructs an algorithm to solve the fractional assignment problem. Algorithms that are currently used are mostly based on parametric approaches and must solve a sequence of optimization procedures. They also neglect the difficulties caused by degeneracy. The proposed algorithm performs optimization once and overcomes degeneracy. The main features of the algorithm are an effective initial heuristic approach, a simple labelling procedure and an implicit primal-dual schema. A numerical example is presented and demonstrates that the proposed algorithm is easy to apply. Computational results are compared with those from other developed methods. The results show that the proposed algorithm is efficient.     

An effective adaptive trust region algorithm for nonsmooth minimization

In this paper, an adaptive trust region algorithm that uses Moreau–Yosida regularization is proposed for solving nonsmooth unconstrained optimization problems. The proposed algorithm combines a modified secant equation with the BFGS update formula and an adaptive trust region radius, and the new trust region radius utilizes not only the function information but also the gradient information. The global convergence and the local superlinear convergence of the proposed algorithm are proven under suitable conditions. Finally, the preliminary results from comparing the proposed algorithm with some existing algorithms using numerical experiments reveal that the proposed algorithm is quite promising for solving nonsmooth unconstrained optimization problems.     

球面上$\ell_1$正则优化的随机临近梯度方法

本文研究球面上的$\ell_1$正则优化问题,其目标函数由一般光滑函数项和非光滑$\ell_1$正则项构成,且假设光滑函数的随机梯度可由随机一阶oracle估计.这类优化问题被广泛应用在机器学习,图像、信号处理和统计等领域.根据流形临近梯度法和随机梯度估计技术,提出一种球面随机临近梯度算法.基于非光滑函数的全局隐函数定理,分析了子问题解关于参数的Lipschtiz连续性,进而证明了算法的全局收敛性.在基于随机数据集和实际数据集的球面$\ell_1$正则二次规划问题、有限和SPCA问题和球面$\ell_1$正则逻辑回归问题上数值实验结果显示所提出的算法与流形临近梯度法、黎曼随机临近梯度法相比CPU时间上具有一定的优越性.