基于非负矩阵分解的函数型聚类算法

函数型聚类分析算法涉及投影和聚类两个基本要素。通常,最优投影结果未必能够有效地保留类别信息,从而影响后续聚类效果。为此,本文梳理了函数型聚类的构成要素及运行过程;借助非负矩阵分解的聚类特性,提出了基于非负矩阵分解的函数型聚类算法,构建了“投影与聚类”并行的实现框架,并采用交替迭代方法更新求解,分析了算法的计算时间复杂度。针对随机模拟数据验证和语音识别数据的实例检验结果显示,该函数型聚类算法有助于提高聚类效果;针对北京市二氧化氮(NO2)污染物小时浓度数据的实例应用表明,该函数型聚类算法对空气质量监测点类型的区分能够充分识别站点布局的空间模式,具有良好的实际应用价值。     

Performance improvement in aircraft maintenance

This study examined the use of utility theory to improve the performance of an aircraft maintenance organization. A utility model was developed and used as an aid in decision making by one large maintenance organization. The results indicated that utility analyses may provide a viable approach to improving effectiveness in organizations having a large number of competing factors. The chief of maintenance provided the output results and the attribute measures. Three team leaders were also interviewed and their views confirmed the views of the chief.     

基于负载均衡的免疫遗传算法网格调度研究

将现有基本遗传算法对网格任务调度进行研究，在此基础上提出解决资源负载均衡问题的方案，并对算法的终止条件进行改进。针对改进后的遗传算法可能出现的问题，在算法初始阶段引入免疫原理，利用免疫机制保证种群多样性，使算法能在更快找到最优解的基础上达到一定的负载均衡，最后通过仿真环境对算法进行验证，结果证明改进后的算法有效。     

Genetic algorithms with path relinking for the minimum tardiness permutation flowshop problem

In this work three genetic algorithms are presented for the permutation flowshop scheduling problem with total tardiness minimisation criterion. The algorithms include advanced techniques like path relinking, local search and a procedure to control the diversity of the population. We also include a speed up procedure in order to reduce the computational effort needed for the local search technique, which results in large CPU time savings. A complete calibration of the different parameters and operators of the proposed algorithms by means of a design of experiments approach is also given. We carry out a comparative evaluation with the best methods that can be found in the literature for the total tardiness objective, and with adaptations of other state-of-the-art methods originally proposed for other objectives, mainly makespan. All the methods have been implemented with and without the speed up procedure in order to test its effect. The results show that the proposed algorithms are very effective, outperforming the remaining methods of the comparison by a considerable margin.     

A simulated annealing algorithm for job shop scheduling

In this paper, the job shop scheduling problem is considered with the objective of minimization of makespan time. We first reviewed the literature on job shop scheduling using meta-heuristics. Then a simulated annealing algorithm is presented for scheduling in a job shop. To create neighbourhoods, three perturbation schemes, viz. pairwise exchange, insertion, and random insertion are used, and the effect of them on the final schedule is also compared. The proposed simulated annealing algorithm is compared with existing genetic algorithms and the comparative results are presented. For comparative evaluation, a wide variety of data sets are used. The proposed algorithm is found to perform well for scheduling in the job shop.     

Locations and vehicle routing designs of physical distribution systems

The design of a physical distribution system (PDS) involves the determination of the number and locations of distribution centres, estimation of required vehicle numbers and design of vehicle routings. Due to the enormous numbers of possible combinatorial designs of the system, it is difficult to obtain an optimal design in acceptable computation effort on many occasions. In this paper, a new solution framework for the design of PDS by implementing the genetic algorithm (GA) is presented. With the characteristic of simultaneous optimization of a large population of configuration, the proposed methodology has been proved to be an extremely efficient optimizer. In the experimental simulation conducted in this paper, it also indicates this approach can provide a near-optimal solution to the design of PDS. To analyse the growth and decay of many schemas contained in the population, the effects of the operation of reproduction, crossover and mutation on the schema are studied. The simulation evaluation about the system performance and genetic parameters is presented along with the discussions at the end of this paper.     

Evaluation of Beta Generation Algorithms

In this article, we discuss the maximum likelihood estimates (MLEs) for the exponential and Weibull distributions by considering progressive Type-I interval censored data. For exponential distribution, the explicit expression of MLE of failure rate cannot be obtained when the intervals are not equal in length. The direct application of some numerical algorithms, such as the Newton–Raphson algorithm, is non-ideal because of the cumbersome second derivative. We apply some equivalent quantities to obtain the MLE of failure rate of exponential distribution. Based on the equivalent quantities and the Weibull-to-exponential transformation technique, we propose a new algorithm to obtain the MLEs for the parameters of progressive Type-I interval Weibull data. An example reanalysis and some simulation studies are carried out to illustrate the performance of the estimations using the new algorithm.     

Genetic Algorithm-Based Redundancy Optimization Problems in Fuzzy Framework

This article uses a genetic algorithm to solve the series parallel redundancy optimization problem which is in a fuzzy framework. Three nonlinear chance constrained programing models and three goal programing models are formulated based on possibility measure and credibility measure. A fuzzy simulation-based genetic algorithm is then employed to solve these kinds of fuzzy programing. Finally, numerical examples are also given.     

Estimating Probability Distributions by Observing Betting Practices

A bookmaker takes bets on a two-horse race, attempting to minimize expected loss over all possible outcomes of the race. Profits are controlled by manipulation of customer' betting behavior; in order to do this, we need some information about the probability distribution which describes how the customers will bet. We examine what information initial customer' betting behavior provides about this probability distribution, and consider how to use this to estimate the probability distribution for remaining customers.