GPU加速不完全Cholesky分解预条件共轭梯度法

不完全 Cholesky 分解预条件共轭梯度（incomplete Cholesky factorization preconditioned conjugate gradient ,ICCG）法是求解大规模稀疏对称正定线性方程组的有效方法。然而ICCG法要求在每次迭代中求解2个稀疏三角方程组,稀疏三角方程组求解固有的串行性成为了ICCG法在GPU上并行求解的瓶颈。针对稀疏三角方程组求解,给出了一种利用GPU 加速的有效方法。为了增加稀疏三角方程组求解在GPU上的多线程并行性,提出了对不完全Cholesky分解产生的稀疏三角矩阵进行分层调度（level scheduling ）的方法。为了进一步提高稀疏三角方程组求解的并行性能,提出了在分层调度前通过近似最小度（approximate minimum degree ,AMD）算法对系数矩阵进行重排序、在分层调度后对稀疏三角矩阵进行层排序的方法,降低了分层调度过程中产生的层数,优化了稀疏三角方程组求解的GPU内存访问模式。数值实验表明,与利用NVIDIA CUSPARSE实现的ICCG法相比,采用上述方法性能可以获得平均1倍以上的提升。     

基于雅可比矩阵逆预处理的快速潮流计算方法

随着电网规模变大，利用稳定双共轭梯度法（Bi-CGSTAB）求解潮流计算中的修正方程组时，收敛速度会变得很慢。通过寻找合适的预处理矩阵是解决问题的关键。研究了雅可比矩阵预处理方法，针对牛顿法求解潮流过程中雅可比矩阵的变化特性，提出将第一次外迭代的雅可比矩阵逆作为预处理矩阵，并与稳定双共轭梯度法相结合，提高潮流计算的收敛速度。借助InterPSS电力系统仿真软件，对IEEE118、IEEE162、IEEE300和一个欧洲大陆真实电力系统进行仿真计算，验证了在处理大规模电网时，所提方法相对稀疏近似逆预处理具备更好的有效性。     

预处理2步非线性共轭梯度法的并行算法与数值试验

§1．引言在各种非线性问题的数值求解中，常归结为求解大型非线性方程组F(x)=0其中x∈Rn，F：DRn→Rn，且n是很大的整数．[1]中给出的非线性共轭梯度法（NCG法）是一种适合F的Jacobi阵非对称的算法，[2]中给出的2步NCG法是[1]的改进且节省计算时间．本文将预处理技术应用于2步NCG法，给出预处理2步NCG法．文中讨论了适合并行计算的最小二乘多项式预处理方法，通过数值试验表明多项式预处理2步NCG法比其他预处理方法优越，比不预处理效果好得多，且方程维数越大效果越明显．本文还给出预处理2步NCG法的并行实现及数值试验，它比…     

对称逐步超松弛预处理共轭梯度法的改进迭代格式

51．引言线性方程组的求解方法可分为两大类：直接法和迭代法．对于大型问题，当系数矩阵为条件数较小的稀疏矩阵且右端项不多时，迭代法的求解效率高．尽管迭代法多种多样，但其迭代收敛速度毫不例外地取决于迭代矩阵的条件数，而预处理的唯一目的就是降低迭代矩阵的条件数，从而达到减少迭代次数和计算量的目的．共轭梯度法（CG法）具有许多内在的优点，如有限步收敛性质．在实际计算中，由于舍入误差的影响，特别是由于系数矩阵的条件数常常较大，CG法往往出现收敛慢的问题．预处理共轭梯度法（PCG法）就是在共轭梯度法中采用了预处理…     

大型稀疏线性方程组新的ICCG方法

有限元线性方程组的系数矩阵一般具有稀疏性和对称性的特点,全稀疏存贮方法就是利用这些特点,只存贮对称部分的非零元素,采用链表式管理,即节省存贮空间,又便于动态更改．在完全Cholesky分解的基础上,构造出了新的预处理方法,应用适当的对角元修正策略,得到了一种新的ICCG方法,能够确保方程组高效准确的分解和求解．数值算例证明该算法在时间和存贮上都较为占优,可靠高效,能够应用于有限元线性方程组的求解．     

求实对称矩阵部分特征值的并行算法

提出了并行求解实对称稠密矩阵部分特征值的反幂法的预处理方法.该方法基于带状矩阵特征问题反幂法的信息传递复杂度低的特点,采用Householder变换并行算法约化大型实对称稠密矩阵为一定带宽的带状矩阵,针对带状矩阵用反幂法求解矩阵的在某一点的近似特征值;其中针对反幂法迭代中遇到的线性方程组,采用文献中的并行预处理共轭梯度算法求解.最后在Lenovo深腾1800集群上进行数值实验,并与预处理前反幂法的计算结果进行了比较,实验结果表明,经过预处理后的并行性远高于直接采用反幂法的并行性.     

视频监控系统中的图象预处理技术

主要分析和讨论工业图象预处理的基本技术与方法，并介绍了近年来发展的图象预处理中的软计算方法。     

面向遥感图像BNs分类的预处理技术及算法实现

遥感图像BNs分类分为预处理、BNs模型构建和分类3个前后联系的过程。其中对预处理技术是后面两个步骤的基础，其算法实现过程对分类结果的影响很大，预处理的目标是高效、准确地提取图像分类所需要的重要特征，剔除干扰因素。该文介绍了一种简单数据预处理技术和其算法实现过程。针对遥感数据的特点和BNs方法的需要，将该预处理过程分成波谱空间分割和关系信息计算两个部分，分别介绍了两部分的原理并给出了实现的算法。对于遥感数据分类预处理技术的研究和实现具有很强的借鉴作用。     

近断层危害场的三维并行数值模拟

在采用Metis软件包进行区域分解的基础上,结合大型稀疏矩阵的行压缩存储方式,求解基于预处理共轭梯度法的有限元方程组,开发并行有限元计算程序。对三维地质模型进行计算模拟,得到的张店-仁河断裂近断层地震危害场的位移场数值计算结果表明,该并行程序取得了较好的计算加速比。     

设备监控图像预处理算法改进及其FPGA实现

为实现对现场监控图像的实时预处理,针对灰度化和二值化算法复杂度高、计算量大和不易于硬件实现等问题,对图像预处理方法进行了研究。通过重新设计加权平均灰度算法的权重,使灰度算法由浮点运算简化为定点运算,易于现场可编程门阵列（FPGA）实现;提出一个基于Otsu算法的改进评价函数,证明了其与最大类间方差法、最小类内方差法的等价性。进而提出一种适于硬件实现的图像预处理算法,并通过Verilog编程以及在Matlab和ModelSim上联合仿真,证明了整个预处理功能的可实现性,最终设计出一种可用于实时图像处理系统前端的基于硬件实现的高速图像预处理模块。     

社会Agent的BDO模型

目前MAS中思维状态的研究趋势是在个体模型中加入社会思维属性,研究社会承诺、依赖、联合意图的推理关系。在BDI模型中,以意图为中心的观点不描述社会Agent。该文分析了以竭力为中心的Agent思维状态研究存在的问题,提出MAS的分层模型,并提出以信念、愿望和义务作为基本思维属性（简称BDO）来描述Agent的思维状态和社会属性,给出了BDO逻辑和语义模型,考虑了奖励、惩罚、承诺和承诺解除问题,并对队、组织、组织意图等现象给出了描述。该文力图更自然地描述社会性Agent的思维状态和群体概念,是对Rao和Georgeff提出的BDI模型的改进。最后通过一个例子说明了BDO的表达能力。进一步的工作包括建立更为完善的语义模型、结合各个思维属性的动态修正语义给出BDO Agent的动态模型以及给出基于BDO逻辑框架的Agent／MAS实现结构。     

Biomechanical analysis of materials handling manipulators in short distance transfers of moderate mass objects: joint strength,spine forces and muscular antagonism

Although often suggested as a control measure to alleviate musculoskeletal stresses, the use of mechanical assistance devices (i.e. manipulators) in load transfers has not been extensively studied. Without data describing the biomechanical effects of such devices, justification for decisions regarding implementation of such tools is difficult. An experimental study of two types of mechanical manipulators (articulated arm and overhead hoist) was conducted to determine whether biomechanical stresses, and hence injury risk, would be alleviated. Short distance transfers of loads with moderate mass were performed both manually and with manipulator assistance under a variety of task conditions. Using analysis and output from new dynamic torso models, strength demands at the shoulders and low back, lumbar spine forces, and lumbar muscle antagonism were determined. Strength requirements decreased significantly at both the shoulders and low back when using either manipulator in comparison with similar transfers performed manually. Peak spine compression and anteriorposterior (a-p) shear forces were reduced by about 40% on average, and these reductions were shown to be primarily caused by decreases in hand forces and resultant spinal moments. Two metrics of muscular antagonism were defined, and analysis showed that torso muscle antagonism was largest overall when using the hoist. The results overall suggest that hoist-assisted transfers, although better in reducing spine compression forces, may impose relatively higher demands on coordination and/or stability at extreme heights or with torso twisting motions. The relatively higher strength requirements and spine compression associated with the articulated arm may be a result of the high inertia of the system. Potential benefits of practice and training are discussed, and conclusions regarding implementation of mechanical manipulators are given.     

Algebraic pruning: a fast technique for curve and surface intersection

Computing the intersection of parametric and algebraic curves and surfaces is a fundamental problem in computer graphics and geometric modeling. This problem has been extensively studied in the literature and different techniques based on subdivision, interval analysis and algebraic formulation are known. For low degree curves and surfaces algebraic methods are considered to be the fastest, whereas techniques based on subdivision and Bézier clipping perform better for higher degree intersections. In this paper, we introduce a new technique of algebraic pruning based on the algebraic approaches and eigenvalue formulation of the problem. The resulting algorithm corresponds to computing only selected eigenvalues in the domain of intersection. This is based on matrix formulation of the intersection problem, power iterations and geometric properties of Bézier curves and surfaces. The algorithm prunes the domain and converges to the solutions rapidly. It has been applied to intersection of parametric and algebraic curves, ray tracing and curve-surface intersections. The resulting algorithm compares favorably with earlier methods in terms of performance and accuracy.     

Process Control in metallurgical plants—From an Xstrata perspective

Some history and modern practice of Process Control in metallurgical operations is reviewed. Clearly the early deliverables from the pioneer days in the 1950s through to the 1970s and early 1980s were under-appreciated. The discipline has since grown into a more visible, sophisticated and accepted practice as a result of the assembly of appropriately recruited and trained individuals and teams, who have successfully negotiated deliverable projects that impact all metallurgical performances beyond early milling processes. The skill set in these individuals and teams essentially includes organisational behaviour in addition to their specialist technical attributes. A strong network to internal and external specialists and experts is essential. Furthermore, instrumentation and control technology has improved immensely. The challenge in the current modern practice is to win support of senior management in operations for the project cost, schedule and deliverables of Process Control. Once gained, this acceptance then amounts to the logistics of project scope and delivery—a track record well-demonstrated by the Xstrata Process Control Group.     

基于.NET的水站实时监测系统设计与实现

介绍了采用微软公司的.NET技术实现水站实时监测查询系统的设计思想、采用的技术方法和实现过程.系统基于B/S模式,采用四层逻辑设计结构,利用.NET组件设计技术和ADO.NET数据访问技术,实现了对水站水情数据查询的实时监测,为调控提供了科学、及时、准确的数据依据.系统的设计与实现是.NET技术在实时监测领域更具实用价值的应用.     

Ocean transparency from space: Validation of algorithms estimating Secchi depth using MERIS, MODIS and SeaWiFS data

Ocean transparency, often measured using Secchi disk, is a useful index of water quality or productivity and is used in many environmental studies. The spaceborne ocean color sensors provide synoptic and regular radiometric data and can be used for applying environmental policies if the data is converted into relevant biogeochemical properties. We adapted and developed semi-analytical and empirical algorithms to estimate the Secchi depth from satellite ocean color data in both coastal and oceanic waters. The development of the algorithms is based on the use of a comprehensive in situ bio-optical dataset. The algorithms are validated using an extensive set of coincident satellite estimates and in situ measurements of the Secchi depth (so-called matchups). More than 400 matchups are compiled for the MERIS, MODIS and SeaWiFS sensors. The comparison between Secchi depth retrievals from remote sensing data and in situ measurements yields determination coefficients (R²) between 0.50 and 0.73, depending on the sensor and algorithm. The type II linear regression slopes and intercepts vary between 0.95 and 1.46, and between − 0.8 and 6.2 m, respectively. While semi-analytical algorithms provide the most promising results on in situ data, the empirical one proves to be more robust on remote sensing data because it is less sensitive to error due to erroneous atmospheric corrections. Using ocean color archives, one can derive maps of ocean transparency for different areas. Our climatology of the Secchi depth based on ocean color for the transition zone between the North Sea and Baltic Sea is compared to an historical dataset.     

不确定性人工智能

在主、客观世界普遍存在的不确定性中,随机性和模糊性是最重要的两种形式.研究了随机性和模糊性之间的关联性,统一用熵作为客观事物和主观认知中不确定状态的度量,用超熵来度量不确定状态的变化,并利用熵和超熵进一步研究了混沌、分形和复杂网络中的不确定性,以及由此带来的种种进化和变异,为实现不确定性人工智能找到了一种简单、有效的形式化方法,也为包括形象思维在内的不确定性思维的自动化打下了基础.不确定性人工智能是人工智能进入21世纪的新发展.这个由多学科交叉渗透构成的新学科,必将使得机器能够具备人脑一样的不确定性信息和知识的表示能力、处理能力和思维能力.     

人和人工智能系统的概念形成过程研究

概念的形成是实现人工智能的基础,为研究人工智能系统中概念的形成过程,从人对事物形成概念的过程出发进行了研究。比较人和人工智能系统的概念形成过程得到了如下特点:人的优势在于能自主地确定对象表象和对象功能中的各种特征和划分等,能在对象、描述性定义和功能性定义对应关系不完备情况下通过思维和联想建立概念;人工智能系统的优势在于丰富的对象表象感知能力,对象的各种特征和划分的长期存储、运算和分析能力;而人工智能的概念形成过程存在的缺点基本与人的概念形成过程的优点对应。因此本文认为人工智能的概念形成过程必须关注因素的智能识别、功能的系统实践和人经验知识的有师学习。现有技术在缺乏人经验知识的情况下,人工智能系统不能自主建立概念和知识库,不能实现智能过程。     

面向主动防御的多样性研究进展

不同软件或执行过程通常存在不同的脆弱性,多样性技术基于该前提应用于系统的可靠性、安全性设计中,显著增强了系统的防御能力和入侵容忍能力,然而也存在系统代价高、复杂性高等不足。已有研究中出现了大量的多样性技术实现、系统设计以及相关的评估工作,覆盖范围广泛。针对主动防御领域内的多样性应用,围绕多样性应用性价比的问题,本文梳理了多样性研究中的典型工作和最新进展。首先对多样性综述研究工作进行了对比分析,讨论了多样性研究的主要内容和研究侧重点。其次对多样性概念进行了梳理,给出了时、空多样性的定义。再次,按照时空多样性的分类方法,对基于多样性的主动防御系统的架构和实现技术进行介绍,分析了时、空多样性系统的特点和实现方式。然后,对多样性度量和有效性评估方法进行了分类总结,分析了不同度量、评估方法的优势和不足。最后,提出了多样性技术的下一步研究方向。     

A hierarchical organization scheme for video data

Video stream is based on bits of imagery and is thus difficult to be perceived (by machine) in the content level. To access video content, a suitable organization of video data is critical. This paper proposes a hierarchical structure and a process scheme for organizing video data to facilitate indexing, browsing and querying. Four layers can be distinguished, that is: video program, episode, shot and frame. This hierarchy provides an efficient and flexible structure as well as compact and meaningful abstraction of video program. To achieve such an organization, not only the boundary detection of shots and episodes, but also the extraction of key-frames for shots and the selection of representative shots and frames for episodes are important. Suitable criteria and methods for above tasks are proposed and these techniques have been integrated into a workable system. A number of organization experiments using real video data are performed and some results are presented, which show the effectiveness of the proposed organization scheme and techniques.