浅论雷达装备"双三化""三新"与"三性"

雷达装备的"三性"是装备战斗力最重要的组成部分,是使用方关注的焦点."双三化"和"三新"是"三性"的基础和动力.文中从雷达装备的"双三化"、"三新"和"三性"概念入手,深入分析了"双三化"与"三新"的相互关系,对雷达装备技术跨代的推动作用,与雷达装备"三性"的内在联系及影响,三者目前的发展动态及未来发展方向.     

基于频繁模式树的普遍化关联规则挖掘

提出了基于频繁模式树的普遍化关联规则挖掘算法 MGAR- FP,充分利用频繁模式树的性质 ,避免大量候选模式的生成和频繁模式匹配 ,提高了挖掘的效率和速度 .实验表明 ,算法是有效的 ,比传统的普遍化关联规则挖掘算法Cum ulate快     

电力系统复杂故障通用算法的研究

我们用补偿法原理开发出电力系统复杂故障的通用计算软件。本文着重讨论其核心部分－故障后补偿导纳矩阵和故障前端口阻抗矩阵的形成，并推导出了有关计算公式。此程序可灵活地修改运行方式，对故障类型、故障特殊相、故障过渡电阻处理简便，易于编程。可满足电力系统继电保护整定计算和事故分析的需要。     

基于样本个体差异性的深度神经网络训练方法

深度神经网络目前在许多任务中的表现已经达到甚至超越了人类的水平,但是其泛化能力和人类相比还是相去甚远.如何提高网络的泛化性,一直是重要的研究方向之一.围绕这个方向开展的大量卓有成效的研究,从扩展增强训练数据、通过正则化抑制模型复杂度、优化训练策略等角度,提出了很多行之有效的方法.这些方法对于训练数据集来说都是某种全局性质的策略,每一个样本数据都会被平等的对待.但是,每一个样本数据由于其携带的信息量、噪声等的不同,在训练过程中,对模型的拟合性能和泛化性能的影响也应该是有差异性的.针对是否一些样本在反复的迭代训练中更倾向于使得模型过度拟合,如何找到这些样本,是否可以通过对不同的样本采用差异化的抗过拟合策略使得模型获得更好的泛化性能等问题,提出了一种依据样本数据的差异性来训练深度神经网络的方法,首先使用预训练模型对每一个训练样本进行评估,判断每个样本对该模型的拟合效果;然后依据评估结果将训练集分为易使得模型过拟合的样本和普通的样本两个子集;最后,再使用两个子集的数据对模型进行交替训练,过程中对易使得模型过拟合的子集采用更强有力的抗过拟合策略.通过在不同的数据集上对多种深度模型进行的一系列实验...     

A novel method for fabricating of high-performance multi-seeding GdBCO bulk superconductors

A novel multi-seeding (NMS) method is proposed for improving the performances of GdBCO samples by optimizing the shapes and the space configurations of the single-domain GdBCO grains. Three multi-seeding GdBCO bulk superconductors with dimensions of 60 × 20 × 14 mm³ were fabricated by a Gd + 011 infiltration growth method with different seeding techniques: samples S1 and S2 by the traditional multi-seeding method (TMS) with two and three seeds, respectively; and sample S3 by the NMS method with three seeds. The results show that among the three, sample S3 has both of the largest levitation force, 111 N, and the largest magnetic field coupling ratio, 68.57%, which are clearly higher than the levitation force of 100 N and 90 N, and the magnetic field coupling ratios of 34.25% and 11.33% for the samples S1 and S2, respectively. These results indicate that the NSM method is a useful way to improve the mechanical strength, homogeneity, and performances of multi-seeding GdBCO samples.     

基于遗传模拟退火算法优化的BP神经网络

在研究标准BP神经网络的基础上,针对其存在的收敛速度慢、且容易陷入局部极小值等问题进行分析,设计实现一种采用数值优化的方法来改进BP网络性能的新的混合神经网络模型.通过引入遗传模拟退火算法扩大了网络的权值更新空间,把得到最优权值赋予BP神经网络,从而使优化后的神经网络具有泛化性好,不易陷入局部极小值等优点.与标准BP神...     

基于支持向量回归模型的电力系统谐波分析新方法

当前电力系统中的谐波问题日益严重,对谐波的准确检测和分析是抑制谐波畸变的重要依据。将基于改进的SMO算法的支持向量回归模型应用于电力系统谐波的检测,该算法克服了常规算法计算规模大和建模复杂的困难,通过引入一个中间因子,将原来问题的计算规模减半,并利用迭代算法求解中间因子,使得该算法简单可行。对三相桥式整流电路交流侧产生的特征谐波和非特征谐波电流进行了分析,仿真结果通过与FFT算法和ADALINE神经网络的检测分析结果对比,表明该方法无论是在理想情况下还是在考虑了各种影响因素的情况下,都具有很高的检测精度,可以满足电力系统的谐波分析的要求。该方法的不足之处是计算量会随着输入量分辨率的提高而增大。     

基于多维约束空间的自动制图综合质量评估模型

为解决现阶段自动制图综合结果难以量化和比较的问题，提出一种基于多维约束空间的制图综合质量评估模型．首先对综合结果所得的各约束值进行归一化和无量纲处理，然后在单个一维约束坐标轴的基础上建立虚拟的多维约束空间，使综合结果在同一尺度范围内进行讨论，最后把欧氏距离扩展到多维约束空间用来度量制图综合质量以进行质量优劣的比较．使用该方法对普通建筑物的不同综合结果进行了评估，计算了面积、相邻点之间距离等5个建筑物约束条件所得制图综合结果的综合评价值，按照该综合评价值的排序获取的质量评估结果与实际情况完全符合．结果证明，该方法原理正确，计算简单，是评价和比较多个制图综合结果的有效方法．     

Enhanced piecewise regression based on deterministic annealing

Regression is one of the important problems in statistical learning theory. This paper proves the global convergence of the piecewise regression algorithm based on deterministic annealing and continuity of global minimum of free energy w.r.t temperature, and derives a new simplified formula to compute the initial critical temperature. A new enhanced plecewise regression algorithm by using ＂migration of prototypes＂ is proposed to eliminate ＂empty cell＂ in the annealing process. Numerical experiments on several benchmark datasets show that the new algorithm can remove redundancy and improve generalization of the piecewise regression model.     

Acquiring Recursive and Iterative Concepts with Explanation-Based Learning

In explanation-based learning, a specific problem's solution is generalized into a form that can be later used to solve conceptually similar problems. Most research in explanation-based learning involves relaxing constraints on the variables in the explanation of a specific example, rather than generalizing the graphical structure of the explanation itself. However, this precludes the acquisition of concepts where an iterative or recursive process is implicitly represented in the explanation by a fixed number of applications. This paper presents an algorithm that generalizes explanation structures and reports empirical results that demonstrate the value of acquiring recursive and iterative concepts. The BAGGER2 algorithm learns recursive and iterative concepts, integrates results from multiple examples, and extracts useful subconcepts during generalization. On problems where learning a recursive rule is not appropriate, the system produces the same result as standard explanation-based methods. Applying the learned recursive rules only requires a minor extension to a PROLOG-like problem solver, namely, the ability to explicitly call a specific rule. Empirical studies demonstrate that generalizing the structure of explanations helps avoid the recently reported negative effects of learning.