关键词： data center   accelerator   graph processing   distributed architecture   communication optimization  

摘要： With the rapid growth of real-world graphs,the size of which can easily exceed the on-chip(board)storage capacity of an accelerator,processing large-scale graphs on a single Field Programmable Gate Array(FPGA)becomes *** multi-FPGA acceleration is of great necessity and *** cloud providers(e.g.,Amazon,Microsoft,and Baidu)now expose FPGAs to users in their data centers,providing opportunities to accelerate large-scale graph *** this paper,we present a communication library,called FDGLib,which can easily scale out any existing single FPGA-based graph accelerator to a distributed version in a data center,with minimal hardware engineering *** provides six APIs that can be easily used and integrated into any FPGA-based graph accelerator with only a few lines of code *** the torus-based FPGA interconnection in data centers,FDGLib also improves communication efficiency using simple yet effective torus-friendly graph partition and placement *** interface FDGLib into AccuGraph,a state-of-the-art graph *** results on a 32-node Microsoft Catapult-like data center show that the distributed AccuGraph can be 2.32x and 4.77x faster than a state-of-the-art distributed FPGA-based graph accelerator ForeGraph and a distributed CPU-based graph system Gemini,with better scalability.

关键词： GNSS钟差   数据处理   钟差量级   稳定性  

摘要： 钟差在GNSS应用中是十分重要的参数信息,包括卫星钟差和接收机钟差,其精度直接影响GNSS的服务水平。卫星钟差通常是利用IGS等GNSS组织的全球跟踪站的观测数据解算的。解算过程中,选择稳定的钟差基准可获得更准确的卫星钟差,进而正确反映卫星钟的性能,以及更好地对卫星钟差进行建模和预报。钟差基准通常选用具有高性能接收机钟的测站,因此,针对IGS测站的接收机钟进行性能分析,本项目评估了所有IGS测站的接收机钟性能,主要包括钟差的量级和时钟的短、长期稳定性。通过实验得出:IGS作为国际上知名的GNSS服务组织,其测站配备的接收机钟的钟差量级大部分优于1×10^(-7) s,占比约58.0%,且有74个测站的钟差量级在ns级;所有IGS测站配备的接收机钟的短期和长期稳定性均较高,大部分的短期(300 s)稳定性集中在1×10^(-12)~1×10^(-11)区间,优于1×10^(-12)的数量占比约24.1%,而长期稳定性(天稳)优于1×10^(-13)的测站数众多,占比约65.0%。基于评估结果,从众多高性能的接收机钟中,筛选出100多个能够作为钟差基准的测站,为GNSS服务端解算卫星钟差提供基准选择的参考。

关键词： 流数据处理   数据清洗   Storm结构   支持向量机   电力物联网  

摘要： 数据的采集、分析和处理是目前电力系统重点关注的问题之一。电力物联网系统中数据业务处理需求多样,尤其是流数据的处理对时延要求严格。然而,现有的数据处理方法形式单一,不能很好地满足低时延处理要求。因此,提出了基于Storm架构的电力物联网流数据处理方法。首先,基于Storm拓扑结构提出分布式流数据处理框架。进而采用流水线式的处理方式,从而达到缩短处理时间的效果。在数据接入后,采用循环队列和流转算子的方法。然后再通过边缘计算处理海量的数据,实现高效的协同工作。此外,通过支持向量机预测算法预测数据的发展趋势,采用清洗技术对脏数据进行处理,解决数据传输过程中的污染情况。仿真结果表明,与传统数据处理方法相比,所提流数据处理方法大大缩短了数据处理时间,满足了大量流数据处理的需求。

关键词： 条件生成对抗网络   小样本问题   机器学习   数据处理   泛化能力   高维数据  

摘要： 在处理小样本机器学习数据时,常规的数据处理算法在处理过程中容易产生不平衡数据,导致数据处理算法的泛化能力比较差.为解决上述问题,基于条件生成对抗网络,提出小样本机器学习数据处理算法.首先将小样本机器学习数据实现在Hibert空间内的映射,然后将数据特征提取问题转化为线性求解问题,得到样本数据特征矩阵,同时计算各个样本数据特征矩阵的信息熵和互信息,根据互信息计算结果调整目标的相对熵,使数据达到平衡,最后将处理后的数据输入到条件生成对抗网络中,以数据处理目标作为目标函数,得到输出结果,完成数据处理.实验结果表明,设计的基于条件生成对抗网络的数据处理算法数据分类准确,数据平衡水平高,整体泛化能力得到了提高.

关键词： 泥石流   点云处理   地貌形态   摄影测量  

摘要： 针对重庆綦江~万盛高速公路边坡在强降雨条件下形成的坡面泥石流,采用消费级无人机航拍生成坡面泥石流点云数据。采用C++编程语言对点云数据进行切片、投影、排序、深度计算、体积计算等操作,提出了基于点云切片提取泥石流各断面侵蚀-堆积极值点平面/剖面分布、侵蚀-堆积体积等地貌特征的方法。结合该研究区进行分析可知,坡面泥石流内部存在侵蚀-堆积优势区域;该坡面泥石流侵蚀-堆积体积与平均侵蚀-堆积深度重合度高,与坡面泥石流宽度大、侵蚀深度浅的特点有关。基于点云数据处理的坡面泥石流形态特征提取方法可从不同角度对地貌特征进行分析,对坡面泥石流侵蚀-堆积关系与其产流-汇流-堆积演化过程研究具有积极意义。

ISBN: (纸本)9787030697097

ISBN: (纸本)9787115571861

关键词： Data processing   distributed systems  

摘要： Optimizing scheduling and communication of distributed data processing for resource and data characteristics is crucial for achieving high performance. Existing approaches to such optimizations largely fall into two categories. First, distributed runtimes provide low-level policy interfaces to apply the optimizations, but do not ensure the maintenance of correct application semantics and thus often require significant effort to use. Second, policy interfaces that extend a high-level application programming model ensure correctness, but do not provide sufficient fine control. We describe Apache Nemo, an optimization framework for distributed dataflow processing that provides fine control for high performance and also ensures correctness for ease of use. We combine several techniques to achieve this, including an intermediate representation of dataflow, compiler optimization passes, and runtime extensions. Our evaluation results show that Nemo enables composable and reusable optimizations that bring performance improvements on par with existing specialized runtimes tailored for a specific deployment scenario. Apache Nemo is open-sourced at https://*** as an Apache incubator project.

关键词： LIBS   Quantification   Uncertainty reduction   Plasma modulation  

摘要： Laser-induced breakdown spectroscopy (LIBS) is regarded as the future superstar for chemical analysis, but the relatively high measurement uncertainty and error remain the persistent challenges for its technological development as well as wide applications. In the present work, mechanisms of measurement uncertainty generation and basic principle of signal uncertainty and matrix effects impacting quantification performance were explained. Furthermore, methods for raw signal improvement including sample preparation, system optimization, and especially plasma modulation, which modulates the laser-induced plasma evolution process for higher signal repeatability and signal-to-noise ratio, were reviewed and discussed. Different LIBS mathematical quantification methods including calibration-free methods and calibration methods, which were classified into physical-principle based calibration model, data-driven based calibration model, and hybrid model, were discussed and compared. Overall, a framework of quantification improvement strategy including key steps and main way-out was summarized and recommended for LIBS future development. (C) 2021 Elsevier B.V. All rights reserved.

关键词： X-ray emission spectroscopy   Chemical speciation   Principal component analysis   Perovskites  

摘要： Wet chemical method was applied with the intention of obtaining Lu1-xSrxFe0.5Cr0.5O3 (0 < x < 1) perovskites in polycrystalline form. A systematic study of the crystal structure of the main perovskite phase obtained as well as a characterization of impurities was performed by X-ray techniques: powder diffraction (PXRD), high resolution emission spectroscopy (HR-XES), and photoelectron spectrometry (XPS). The PXRD data of Lu1-xSrxFe0.5Cr0.5O3 (0 < x < 1) were refined using Pbnm orthorhombic unit cell for the perovskite structure. Only samples with x = 0 and x = 0.1 resulted pure;as the amount of Sr increases, it is segregated as impurities, first as SrCrO4 and then as SrFe12O19. The mean oxidation state of Cr in the analyzed samples was obtained from HR-XES by quantifying spectral changes in the Cr-K beta" and Cr-K beta 2,5 regions by different methodologies: statistical parameters and multivariate methods. The results were compared with the ones obtained by PXRD and XPS. With the last technique Cr6+ ionic concentrations were obtained through the spectral analysis in the region Cr2p3/2-580eV. Similar Cr6+ ionic compositions were obtained using the three X-ray techniques.