基于γ-散度的稳健有序误标记logistic回归

有序多分类方法已经得到了广泛研究。传统的有序多分类方法假设样本的类别标签是不存在误标记的。但是由于实际数据复杂以及人工经验有限,获得标记完全正确的样本是不现实的,因此,传统的方法就存在局限性。提出一种基于γ-散度的有序误标记logistic回归方法,在处理存在误标记的有序多分类问题时具有很强的稳健性,也就是说,当某一样本被错误标记时它对参数估计的权重小于其被正确标记时的权重。该方法通过最小化γ-散度构建模型,利用梯度下降算法求解模型,不仅具有很强的稳健性而且在模型中可以忽略误标记概率。模拟研究和真实数据分析都说明该有序误标记logistic回归方法在处理存在误标记的有序分类问题时效果很好。     

C/C++的编程风格与强壮性的探讨

阐述为了使开发出来的程序具有更高的可读性、可维护性和强壮性,需要约束程序的编写;指出某些不良方法存在的弊病;提出一种比较规范的有利于程序开发的编程风格,并讨论了如何提高程序的强壮性。     

基于扩张状态观测器的鲁棒转子磁链估计研究

提出了新型的基于线性扩张状态观测器的感应电动机转子磁链估计模型。以两个独立的线性扩张状态观测器观测定子电流,用分别被扩张的两个状态构建闭环转子磁链观测器,根据Lyapunov稳定性理论推导满足转子磁链观测器渐进稳定的反馈增益矩阵和转速估计自适应律。利用Matlab进行了仿真,分析了电机参数变化对磁链观测和转速估计精度的影响,表明了提出的模型对电机参数变化具有强鲁棒性。     

The Effect of Bilingualism on the Measured Intelligence of Elementary School Children of Italian Parentage

The recency of the community college movement has prevented large scale organization and dissemination of knowledge concerning the processes and steps required to establish a community college. There is, therefore, a need for means to incorporate contemporary evidence into the curriculum of prospective administrators of higher education. The purpose of this study was to compare Higher Education Administration students’ peceptions of a logical sequence of events for establishing a community college with a sequence of events for establishing a community college as determined by practice. The Edwards’ matched-pair technique was used to rank the perceptions of the students. A sequence of twenty-nine events was selected from the literature. Each event was matched with every other, thus providing 406 decision situations. From these decisions an ordering was obtained. Using the Spearman rho correlation technique, a coefficient of 0.82 was found between the sequence of events taken from the literature and rank ordering of these events as determined by the Edwards’ matched-pair technique.     

带有倒位变异的差分进化算法

针对高维复杂函数的优化问题,提出一种带有倒位变异的差分进化算法.当个体适应度值连续几代不变时,对前一代的最优个体进行倒位变异,以增强种群的多样性,使其跳出局部最优.数值实验结果表明:该算法全局搜索能力强,收敛速度快,且鲁棒性好.     

大型数据库交叉型数据鲁棒性挖掘模型仿真

在大型流媒体数据库数据集中,交叉性数据的鲁棒性挖掘结构建立是实现对数据库差异属性分类和数据访问的基础。传统方法对大型数据库中的交叉性数据的鲁棒性挖掘结构建模采用基于遗传算法的数据集聚调度方法,存在较大的路径损耗,数据挖掘的鲁棒性不好。提出改进的基于局部离群点检测遗传进化的大型数据库交叉型数据挖掘模型,构建基于遗传算法的大型流媒体数据库挖掘结构,进行大型流媒体数据库中交叉型数据信息流特征预处理,结合交叉性型数据的离群因子概念,调整流媒体数据调度的位置变换策略,实现交叉性数据的鲁棒性挖掘算法改进。仿真实验结果表明,该算法能有效数据挖掘的a最大匹配率和局部离群点检测性能,保证了数据挖掘的鲁棒性,各项参数指标优于传统方法,展示了较好的应用价值。     

Cross-sector collaboration in upper secondary school vocational education: experiences from two industrial towns in Sweden and Norway

This study explores industry–education collaboration on vocational education and training (VET) in upper secondary schools in Sweden and Norway, with particular attention to the initiatives, organisation and operational management, and aspects of robustness and lock-in effects. The case studies include two upper secondary schools situated in manufacturing based towns, which are similar in size and industrial structure, and have the dual system of education. Pupils come from these towns and other places in the surrounding region. The research design is based on a mixed-method approach, including interviews and other sources of information from schools and industry. This covered organisational and operational levels, corporate motives and arrangements, industrial composition, labour market conditions, and other regional and national characteristics. The results demonstrated how shared goals of improving the reputation and quality of VET and meeting industries’ needs for skills formed in two distinct organisational models. These findings raise concerns about the robustness of these collaborations, since how changes occur in companies’ economic situations may affect their ability and motivation to participate in VET training. The authors conclude that the arrangements have had success but may create a lock-in-situation regarding the companies’ ability to restructure and develop new skills for new sectors.     

鲁棒性分析的方法论意义

本文在考察鲁棒性分析概念及其本质的基础上,对其方法论特征与功能进行了探讨。     

蓝细菌体系中外噪音维持的生物钟

采用数值模拟方法 ,研究了外部色噪音对蓝细菌生物钟体系的影响。结果表明,外噪音可以在体系的确定性稳态区诱导产生振荡和相干共振;在确定性振荡区,外噪音对振荡起到了破坏性作用,但是振荡的周期并没有随着外噪音的增加而明显变化,说明体系对外噪音具有鲁棒性;而在右侧分叉点附近的确定性振荡区,也发现了外噪音诱导相干共振现象。另外,研究了噪音相关时间的影响,发现其与噪音强度起相反的作用。     

一种神经网络预测控制在超临界主汽温度中的应用研究

传统PID控制难以在非线性、迟延、时变和具有扰动特质的超临界主汽温度控制系统中达到满意的控制效果.因此,提出了一种采用多步预测、滚动优化和反馈校正的神经网络预测控制系统.以某超临界电厂主汽温度为研究对象,MATLAB仿真结果表明：不同的工况建立的主汽温度神经网络动态模型,能够很好地预测对象的动态特性,取得了优于传统PID的控制效果.