基于多层grams的在线支持向量机的中文垃圾邮件过滤

该文提出一种多层grams特征抽取方法来提升基于在线支持向量模型的垃圾邮件过滤器。基于在线支持向量机模型的垃圾邮件过滤器在大规模垃圾邮件数据集已取得了很好的过滤效果,但与逻辑回归模型相比,计算性能的耗时是巨大的,很难被工业界所运用。该文提出的多层grams特征抽取方法能够有效减少特征数,抽取更精准有效的特征,大幅降低模型的运行时间,同时提升过滤器的过滤效果。实验表明,该方法使得在线支持向量机模型的运行时间从10337s减少到3784s,同时模型(1-ROCA)%降低了一半。     

一种基于数据流环境感知的共享过滤算法

为了有效过滤数据流中的有害信息,通常在数据流上注册大量查询,同时构建过滤器来计算这些查询.在多媒体流环境中,查询和过滤器常常是一种“多对多”的连接,也就是说,对于单个过滤器的计算可能会同时给出多个查询的结果.在这种情况下,如何排序所有的过滤器来获得最小的过滤代价变得非常重要.对于过滤器的排序一般依赖于3个指标:过滤器本身的执行代价c、过滤器连接的查询数目p以及过滤器对于随机样本判断为真的概率s.目前基于贪心的排序算法虽然在一定程度上给出了近似最优的结果,但是还存在以下两个问题:1)指标s只是简单依据经验值设定,不能随着流的变化而自适应变化;2)将3个指标融合成一个代价函数进行排序,而没有深入分析各个指标之间的关系.考虑到以上方法存在的不足,提出一个层次排序算法(adaptive hierarchal ordering,AHO)来高效地过滤多媒体数据流.该算法首先依据过滤器的指标c和p进行分类,然后再在每个类别上按照s进行二次排序.在真实多媒体流环境中的过滤结果证明:AHO可以在不降低准确度的情况下,自适应调整过滤器顺序,其性能优于已有的贪心排序算法.     

BPF数据包过滤器的分析与研究

随着网络的迅速发展,信息量的增大,大量的无效数据降低了网络监控的效率,对网络数据包的捕获和过滤变得尤为必要了.BPF数据包过滤器以其简洁的结构在计算机内核中对数据报进行第1层过滤,减轻了用户层工作的负担并且提高了抓包率.研究了BPF的工作原理和过滤模型并且提出了在Windows系统下如何运用BPF过滤器开发包过滤系统.BPF数据包过滤器在网络监控方面有巨大的研究和使用价值.     

串行端口过滤器的研究与实现

针对在不影响系统现有驱动程序接口的前提下实现串行端口传输数据过滤功能的目的,通过对Win32驱动程序模型的驱动程序分层和IRP结构的分析,结合驱动程序处理IRP结构数据的方法,得出了在功能驱动程序的上层增加过滤器可以过滤串行端口传输数据的结论,并介绍了串行端口过滤器驱动程序的开发方法。     

一种优化多重过滤的序列查询算法

序列数据一类重要的数据类型,在文本、Web访问日志文件、生物数据库等应用中普遍存在,对其进行相似性查询是一种获取有用信息的重要手段.在大型序列数据库中进行高效相似性查询的关键因素之一就是查询算法的过滤能力,即设计能快速过滤与查询序列不相关序列集的过滤器十分重要.提出了结合序列距离的度量性质和序列自身特征的多重过滤算法SSQ_MF,SSQ_MF使用了长度过滤器、前缀过滤器和基于参考集的过滤器,使得算法过滤能力较基于单一过滤器算法进一步增强.此外,设计了有关数据结构对查询数据库的一些统计信息进行了预计算和保存,有效估计了各过滤器的过滤集大小,并构建了一个由过滤集大小确定的最优过滤顺序模型,使得算法的过滤代价最低.实验结果表明,算法SSQ_MF的查询性能优于单一过滤器算法和随机过滤顺序的多过滤器算法.     

轧制力传感器的强度验算和安装研究

针对国内热轧带钢连轧机组生产线使用的轧制力传感器在线使用周期缩短、非正常下线逐渐增加的现象,从轧制力传感器的强度验算和安装着手进行分析。主要介绍了磁弹性轧制力传感器压头的环形、盘形、矩形三种结构安装类型,给出了压头的强度及固定夹板双头螺栓的强度验算,并提出了相应了改进方案和措施。探讨了压头的机械安装结构原理。最后分析得出,压头的机械安装问题同样是保证压头不局部受力和大的侧向力影响,使压头的受力面局部应力集中而损坏压头。该研究的目的是减少技术大修,节约备件修复费用和提高设备上机安全运行时间,最大程度提高压头在线使用寿命。     

无线传感器网络中Skyline节点连续查询算法

作为多目标决策的重要手段之一,Skyline节点查询在传感器网络应用中发挥着非常重要的作用.文中深入地分析了Skyline节点查询的性质,提出了基于过滤的Skyline节点连续查询算法(FIlter based Skyline moniToringalgorithm,FIST).FIST算法共包括自底向上、自顶向下和混合3种过滤方式,均通过在传感器节点设置本地或全局过滤器来避免不必要的数据传输,进而节约传感器节点的能量.自底向上过滤方式通过缓存先前Skyline结果作为本地过滤器来避免数据重复传输,而自顶向下过滤则通过设置超立方体作为全局过滤器来避免数据反复更新.由于两者各有利弊,因而提出了混合过滤方式,通过为节点选择合适的过滤器来扬长避短.大量仿真实验的结果表明,FIST算法能有效地减少Skyline节点连续查询过程中传感器节点的通信代价,进而降低传感器网络的能量消耗.     

自洁式超声波在线过滤器的研制

我国地表水浊度高,极大地影响基于光度法的在线分析仪器正常运行;以自主创新的在线采水过滤头为核心,提出一种新型的自洁式超声波在线过滤器,由超声波电源、采水和反冲洗模块、嵌入式控制模块等组成,具备定时定速过滤采水、超声波反冲清洗等功能;实验结果表明,过滤器能够连续地自动滤除水中的悬浮物质,并且不影响目标物的化学形态,对同一水样的浊度去除率维持在85%左右。     

分层法求解网络最大流的研究

网络最大流问题是经典的组合优化问题,随着网络规模的增加,提高算法效率成为解决问题的关键.为了降低求解大规模网络最大流的计算量,针对单源单汇网络提出基于网络分层的最大流问题求解新方法.分层法首先构造原有向网络对应的层次网络,接着在构造出的层次网络中计算各相邻结点层之间的最大流,以此为基础最终获得整个网络最大流的快速估算.分层法有效降低了计算的复杂性,为在大规模网络中快速获取最大流的求解提供了方便,并给出了一个解决最大流问题的新思路.不同网络上测试的实验结果显示,最大流的近似解误差可控制在1%左右,而平均运行时间仅为经典算法(FordFulkerson算法)运行时间的11%,最好情况下的运行时间仅为经典算法运行时间的2%,是two-phase capacity scaling改进算法运行时间的25%,表明分层方法的有效性.     

智能化分段线性恒流LED驱动电源设计

传统的开关型LED驱动电源中含有大电解电容和高频变压器,导致LED驱动电源的体积庞大且使用寿命较短。分段式线性恒流驱动电源可以避免使用大电解电容和高频变压器。设计了一种新型的分段线性恒流LED驱动电源,利用整流之后的高压脉动直流电压的变化,自适应地控制LED灯珠分阶段恒流工作。除整流桥和采样电阻外,整个驱动电路可实现单芯片集成,有效缩小了体积、延长了寿命。电路中还设计了智能拓展端口,可实现智能化控制。基于华虹宏力0.5μm 700 V BCD工艺对电路进行了仿真验证,在0~311 V周期脉动高电压输入条件下,驱动芯片分四阶段恒流工作,输出最大恒定电流可达97.17 m A,在恒流阶段,电流的瞬态精度误差仅为0.031%。仿真结果表明,该LED驱动电路各指标参数均满足预期要求。     

Simulation of the filtration behaviour of dust filters

Periodically regenerable cake-forming filters have assumed a commercially dominant role in air purification due to their excellent dust collection capabilities. These filters are used to separate particles from gases with high dust concentrations, whereby the separation arises as the dust-gas mixture passes through the filter medium and the particles are retained. As soon as a certain maximum pressure drop over the filter medium is reached the filter medium is regenerated. In order to optimally design such filters, there are three important parameters for the economical evaluation of the filter: the correlating pressure drop increase of the filter during filtration which is mainly influenced by the compressibility of the dust cake, the dust separation efficiency especially at the beginning of the filtration and the rise of the residual pressure drop of the filter after several cycles of filtration and regeneration. In this work a 2-dimensional computer simulation model for the compressible dust cake build-up on widely used non-woven fabrics (needle-felts) is introduced. First this model is used to investigate the mechanism causing compression inside dust filter cakes and to clarify different compression phenomena. Second it is shown that this simulation program is capable of describing the particle penetration inside and through the filter medium, the clogging of the filter medium and the growing of the dust filter cake out of the filter medium, whereby the corresponding pressure drop and the number of particles in the clean gas as an function of the filtration time show good qualitative agreement with known experimental data.     

一个基于Naive Bayesian垃圾邮件过滤器的改进

近几年来,垃圾邮件成为互联网的公害之一。现有的反垃圾邮件技术中,基于统计方法的Naive Bayesian分类算法在垃圾邮件过滤中有很好的效果。文中简单介绍了Naive Bayesian分类算法,提出了一种旨在提高垃圾邮件过滤精确率的改进方案,并给出了实验结果。     

超链接数据模型的数据操作

文章在Web数据和关系数据库数据的统一模型———超链接数据模型定义的基础上,给出了超链接数据模型的数据操作定义,包括并、差、交、积、筛选、连接、链接等操作。     

基于LD算法的SQL注入攻击过滤方法研究

如何有效过滤SQL注入攻击是当前Web安全领域亟需解决的关键问题之一。结合关键字过滤与序列比对过滤技术,提出一种基于LD算法的SQL注入攻击过滤（SQLIAF）方法。首先利用黑名单技术,从IP角度对非法用户进行过滤,进而缩小用于过滤的流量规模。其次对用户输入进行关键字检测：若不含关键字,利用LD算法序列比对的方法对非法输入进行过滤;若包含关键字则区分用户请求方式,采用直接加入黑名单的方式或利用LD算法进行处理,以解决传统关键字过滤方法对正常请求的误报问题。实验结果表明,与传统关键字过滤方法和规则匹配过滤方法相比,该方法能有效过滤SQL注入攻击,且误报率及漏报率更低、过滤速度更快。     

社交网络中模块关系树的相似性算法的研究

提出一种基于模块关系树的分析方法,考虑每个实体与用户之间的兴趣、住址和共同好友等相关因素,制定不同的关系树,然后根据路径长度计算各因素的相关度值,最后综合每个实体模块,从而筛选出关系最密切的实体。实验结果证明,该算法能过滤掉大量无关信息,有效找出最相关的实体,提高了搜索结果的准确率。     

New CBR adaptation method combining with problem–solution relational analysis for mechanical design

Case based reasoning (CBR) methodology is proved to be a promising methodology on determining the parameter values of new mechanical product by adapting previously successful solutions to current problems. Compared with the sophisticated case retrieval technique, the case adaptation under K-nearest neighbour is still a bottleneck problem in CBR researches, which needs to be resolved urgently. According to the characteristics of parametric machinery design (PMD), i.e., less data and many parameters, this paper employs weighted mean (WM) as a basic model, and presents a new CBR adaptation method for PMD by integrating with problem–solution (PS) relational information. In our proposed adaptation method, prior to adapting the similar cases, the grey relational analysis (GRA) is utilized to investigate the PS relational information hidden in K retrieved cases, and the proposed method is called as GRA-WM. Different from classical WM method, the weighting factor of retrieved case for each solution element adaptation is calculated by multiplying similarity matrix (SM) and relational matrix (RM), and the adapted solution values of new mechanical product are subsequently obtained by calculating the weighted average of solution values of K similar cases. A case study on the power transformer design is given to prove the industrial applicability of GRA-WM. Moreover, the empirical comparisons between GRA-WM and other adaptation methods are carried out to validate its superiority. The empirical results indicate that GRA-WM can offer an acceptable adaptation proposal in application of CBR for mechanical design.     

Coupling of local visualization and numerical approach for particle microfiltration optimization

The performance of the microfiltration process is controlled by the filter fouling due to the accumulation of solid matter forming a cake layer on the membrane surface. The objective of this work is to study the cake build up and growth at the particle level and to establish correlations with microfiltration performance measured at the process scale. A theoretical model coupling Navier–Stokes equation, convective/diffusion particle transport and deposit formation is developed to simulate a sequence of microfiltration in a confined geometry (Comsol Multyphysics^®). This model is used to make predictive simulations of cake growth during the filtration of diluted particles in the range of size of microorganism (5 μm). In the same time a specific filtration micro-system including an optical chamber and a microsieve (Aquamarijn^®) filtration membrane is designed in order to perform an experimental approach allowing in situ 3D-visualization of a deposit of model particles (polystyrene fluorescent microspheres) using Confocal laser scanning microscopy (CLSM). Based on image analysis, the cake building and properties (particle arrangement, thickness) are analyzed. These experimental data will be further used to improve the filtration model in order to obtain a predictive tool for process optimization.     

Outer joins and filters for instantiating objects from relationaldatabases through views

One of the approaches for integrating object-oriented programs with databases is to instantiate objects from relational databases by evaluating view queries. In that approach, it is often necessary to evaluate some joins of the query by left outer joins to prevent information loss caused by the tuples discarded by inner joins. It is also necessary to filter some relations with selection conditions to prevent the retrieval of unwanted nulls. The system should automatically prescribe joins as inner or left outer joins and generate the filters, rather than letting them be specified manually for every view definition. We develop such a mechanism in this paper. We first develop a rigorous system model to facilitate the mapping between an object-oriented model and the relational model. The system model provides a well-defined context for developing a simple mechanism. The mechanism requires only one piece of information from users: null options on an object attribute. The semantics of these options are mapped to non-null constraints on the query result. Then the system prescribes joins and generates filters accordingly. We also address reducing the number of left outer joins and the filters so that the query can be processed more efficiently     

基于相关滤波技术实现激光陀螺抖动信号的剥除

为了满足惯导系统对机械抖动激光陀螺快速跟踪方面的应用需求,基于相关滤波技术实现了激光陀螺抖动误差信号的剥除.首先从理论上对基于相关滤波去除抖动的思想进行了详细的分析,提出了基于相关滤波进行抖动剥除的开环模型,为了提高系统的稳定性,在开环模型的基础上提出了闭环模型.其次介绍了基于相关滤波闭环模型实现抖动剥除的软硬件设计,最后用所研制的系统对某一型号的激光陀螺进行了测试,通过对采样信号和误差信号的观测,系统工作稳定,在5 kHz的采样条件下,陀螺剥除后的剩余脉冲控制在±1内.