LATIN: A new view and an extension to wave propagation in nonlinear media

The LATIN (acronym of LArge Time INcrement) method was originally devised as a non‐incremental procedure for the solution of quasi‐static problems in continuum mechanics with material nonlinearity. In contrast to standard incremental methods like Newton and modified Newton, LATIN is an iterative procedure applied to the entire loading path. In each LATIN iteration, two problems are solved: a local problem, which is nonlinear but algebraic and miniature, and a global problem, which involves the entire loading process but is linear. The convergence of these iterations, which has been shown to occur for a large class of nonlinear problems, provides an approximate solution to the original problem. In this paper, the LATIN method is presented from a different viewpoint, taking advantage of the causality principle. In this new view, LATIN is an incremental method, and the LATIN iterations are performed within each load step, similarly to the way that Newton iterations are performed. The advantages of the new approach are discussed. In addition, LATIN is extended for the solution of time‐dependent wave problems. As a relatively simple model for illustrating the new formulation, lateral wave propagation in a flat membrane made of a nonlinear material is considered. Numerical examples demonstrate the performance of the scheme, in conjunction with finite element discretization in space and the Newmark trapezoidal algorithm in time. Copyright © 2017 John Wiley & Sons, Ltd.     

基于增量局部加权学习的查询模板自适应基数估计

基数估计是基于代价查询优化的关键步骤,已经被研究了近40年.传统方法如基于直方图的方法在一些假设如属性相互独立、相交的表满足包含原则等成立时能基本满足准确性要求.然而,在真实运行环境中这些假设往往不再成立,可能导致基数估计严重错误进而造成查询延迟.近年来,随着数据的增多和新硬件的发展,使用机器学习方法来提高基数估计的质量成为了可能.由于基于代价的查询优化主要根据查询中子执行计划的估计代价来选择最优的查询执行计划,因此,有一些最近的工作针对一些关键的子执行计划模板建立相应的局部学习模型,取得了不错的进展.但是,这些局部模型主要用于查询(查询空间)分布和数据(数据库数据)分布不变的场景,而在真实运行环境中,它们往往不断地发生变化,限制了这些估计技术的有效性.在本文中,我们针对子执行计划模板在查询分布和数据分布不断变化的环境下提出了一种使用增量的局部加权学习进行自适应基数估计的方法.具体地说,首先抽取子执行计划的语义和统计特征使之能代表当前查询和数据的特性,然后使用增量的局部加权学习模型根据查询分布和数据分布的变化进行自适应的学习,实现基数估计.最后,通过对比实验验证了本文方法的有效性.     

Online anomaly detection for multi‐source VMware using a distributed streaming framework

Anomaly detection refers to the identification of patterns in a dataset that do not conform to expected patterns. Such non‐conformant patterns typically correspond to samples of interest and are assigned to different labels in different domains, such as outliers, anomalies, exceptions, and malware. A daunting challenge is to detect anomalies in rapid voluminous streams of data. This paper presents a novel, generic real‐time distributed anomaly detection framework for multi‐source stream data. As a case study, we investigate anomaly detection for a multi‐source VMware‐based cloud data center, which maintains a large number of virtual machines (VMs). This framework continuously monitors VMware performance stream data related to CPU statistics (e.g., load and usage). It collects data simultaneously from all of the VMs connected to the network and notifies the resource manager to reschedule its CPU resources dynamically when it identifies any abnormal behavior from its collected data. A semi‐supervised clustering technique is used to build a model from benign training data only. During testing, if a data instance deviates significantly from the model, then it is flagged as an anomaly. Effective anomaly detection in this case demands a distributed framework with high throughput and low latency. Distributed streaming frameworks like Apache Storm, Apache Spark, S4, and others are designed for a lower data processing time and a higher throughput than standard centralized frameworks. We have experimentally compared the average processing latency of a tuple during clustering and prediction in both Spark and Storm and demonstrated that Spark processes a tuple much quicker than storm on average. Copyright © 2016 John Wiley & Sons, Ltd.     

Exploiting user metadata for energy-aware node allocation in a cloud storage system

Cloud computing has gained popularity in recent years delivering various services as cost-effective platforms. However, the increasing energy consumption needs to be addressed in order to preserve the cost-effectiveness of these systems. In this work, we target the storage infrastructure in a cloud system and introduce several energy efficient storage node allocation methods by exploiting the metadata heterogeneity of cloud users. Our proposed methods preserve load balance on demand and switch inactive nodes into low-energy modes to save energy. We provide a mathematical model to estimate the outcome of proposed methods and conduct theoretical and simulative analyses using real-world workloads.     

变长增量型极限学习机及其泛化性能研究*

极限学习机（ELM）在训练过程中无需调整隐层节点参数,因其高效的训练方式被广泛应用于分类和回归,然而极限学习机也面临着结构选择与过拟合等严重等问题。为了解决此问题,针对隐层节点增量数目对收敛速度以及训练时间的影响进行了研究,提出一种利用网络输出误差的变化率控制网络增长速度的变长增量型极限学习机算法(VI-ELM)。通过对多个数据集进行回归和分类问题分析实验,结果表明,本文提出的方法能够以更高效的训练方式获得良好的泛化性能。     

中心修正增量主成分分析及其在文本分类中的应用

增量式学习模型是挖掘大规模文本流数据的一种有效的数据处理技术。无偏协方差无关增量主成分分析(Candid Covariance-free Incremental Principal Component Analysis, CCIPCA)是一种增量主成分分析模型,具有收敛速度快和降维效果好的特点。但是,CCIPCA模型要求训练数据是已经中心化或中心向量固定的。在实际的应用中,CCIPCA往往采用一种近似的中心化算法对新样本进行处理,而不会对历史数据进行中心化修正。针对这一问题,该文提出了一种中心修正增量主成分分析模型(Centred Incremental Principal Component Analysis, CIPCA)。CIPCA算法不仅对新样本进行中心化处理,而且会对历史数据进行准确的中心化修正。在文本流数据上的实验结果表明,CIPCA算法的收敛速度和分类性能明显优于CCIPCA算法,特别是在原始数据的内在模型不稳定的情况下,新算法的优势更为明显。     

改进的细菌觅食优化算法及其应用研究

为了解决离散域问题,提出了改进的细菌觅食优化算法,并用该算法以减少分布式发电单元总的功率损耗,提高径向配电系统的电压分布。该算法旨在改进细菌觅食优化算法的性能。实验结果表明,提出的改进的细菌觅食优化算法得出的结果优于细菌觅食优化算法。最后将电力系统的12路、34路总线径向分布系统组成的11和33节进行仿真比较,证明了所提出的优化方法的可实现性和方便性。     

Ncut聚类与增量支持向量机的SMB智能建模

针对当前SMB （simulated moving bed）难以实时在线测得输出组分纯度的现状,结合Ncut （normalized cut）聚类及增量学习支持向量机的方法建立达到周期性稳定状态时系统的智能模型。采用Ncut方法对离线采集的数据样本进行聚类,得到样本的聚类结果;将聚类后的样本数据按反复记忆增强机制输入向量机进行增强‐增量学习训练;将原始测试样本输入到训练好的模型中进行检验。检验结果表明,采用该模型可以获得更好的模型适应度和检验精度,仿真结果验证了该方法的有效性。     

核函数自适应的Mean Shift目标跟踪算法

为解决Mean Shift算法无法对核函数带宽进行自适应更新的缺陷,提出目标质心的分布散度与多级正方形匹配结合的核函数带宽的更新算法。利用目标质心点的分布散度和增量试探法计算几个目标的可能变化尺度,采用多级正方形匹配计算各回字形区域间的Bhattacharyya距离预测目标的尺度变化趋势,对该趋势下的几个目标尺度进行Bhattacharyya距离对比,Bhattacharyya距离最大者为当前核函数的带宽,即目标的尺度。该策略减少了背景噪声的干扰以及每次计算目标收敛区域时的冗余像素的干扰。实验结果表明,该策略优于增量试探法和传统的核函数带宽不变化的方法,在时间代价上略低于增量试探法。