A machine learning approach for predicting computational intensity and domain decomposition in parallel geoprocessing

ABSTRACT

High performance computing is required for fast geoprocessing of geospatial big data. Using spatial domains to represent computational intensity (CIT) and domain decomposition for parallelism are prominent strategies when designing parallel geoprocessing applications. Traditional domain decomposition is limited in evaluating the computational intensity, which often results in load imbalance and poor parallel performance. From the data science perspective, machine learning from Artificial Intelligence (AI) shows promise for better CIT evaluation. This paper proposes a machine learning approach for predicting computational intensity, followed by an optimized domain decomposition, which divides the spatial domain into balanced subdivisions based on the predicted CIT to achieve better parallel performance. The approach provides a reference framework on how various machine learning methods including feature selection and model training can be used in predicting computational intensity and optimizing parallel geoprocessing against different cases. Some comparative experiments between the approach and traditional methods were performed using the two cases, DEM generation from point clouds and spatial intersection on vector data. The results not only demonstrate the advantage of the approach, but also provide hints on how traditional GIS computation can be improved by the AI machine learning.     

Calculation of the added mass of circular cylinders in shallow water

The added mass of hull section shapes must be known, in order to carry out any studies of ship motions, which may be associated with either seakeeping or manoeuvring. Determination of the added mass in deep water can involve conformal mapping or surface singularity distributions. However, in shallow water the problem of determining the added mass is made more difficult by the proximity of the seabed. In this paper, three approaches to the problem of finding the added mass of semi-circular sections are explored. Although there is no known exact solution to this problem, the relationships given by the three methods are compared and the approximate nature of the solutions is examined. The first method is a series solution, which is taken here to include more terms than are normally given in the existing literature. The other two solutions are considered to be new, and so are given in some mathematical detail.     

松辽盆地白垩系的密集段及海水进侵的新证

在总结松辽盆地白恶系层地层特征的基础上，阐述了密集段的地质学和地球物理学标志，经分析，大多数重要反射与密集段相关而并非是层序边界，根据扫描电镜分析结果，在密集段中首次发现了钙质超微化石，为区内白垩纪海水进侵的提供了新的证据，由稳定同位素分析，密集段形成了缺氧，相对可容纳空间最大及水体最深时期。     

Numerical simulations for coupled rock deformation and gas leak flow in parallel coal seams

Based on the new viewpoint of interaction mechanics for solid and gas, gas leakage in parallel deformable coal seams can be understood. That is, under the action of varied geophysical fields, the methane gas flow in a double deformable coal seam can be essentially considered to be compressible with time-dependent and mixed permeation and diffusion through a pore-cleat deformable, heterogeneous and anisotropic medium. From this new viewpoint, coupled mathematical models for coal seam deformation and gas leak flow in parallel coal seams were formulated and the numerical simulations for slow gas emission from the parallel coal seams are presented. It is found that coupled models might be close to reality. Meanwhile, a coupled model for solid deformation and gas leak flow can be applied to the problems of gas leak flow including mining engineering, gas drainage engineering and mining safety engineering in particular the prediction of the safe range using protective layer mining where coal and gas outbursts can efficiently be prevented. This revised version was published online in July 2006 with corrections to the Cover Date.     

一种基于交叉剖面的地质模型构建方法

针对基于近似平行断面数据的建模方法的缺点,利用交叉地质剖面数据,提出了一种通过划分空间单元格的思路来构建地质结构模型的方法。该方法基于“分而自治”的思想,将复杂的整体的建模区域划分为单个单元格内的建模,待每个格子内的建模完成之后,采用相反的思路,将所有格子内的模型拼接起来形成最终的建模区域内的整体模型     

Python在短临气象预报检验中的应用

基于机器学习方法的短临多气象要素预报系统(Weather Elements Nowcasting based on machine learning,简称WEN)具有高发布频次、高时间分辨率、基于候和时辰的复杂预报模型等特点。应用多维标签数组、机器学习工具、并行计算框架等Python库,以快速计算为目标,建立以预报模型覆盖时间范围为统计检验时间边界的检验子系统,客观给出预报性能,为模型调优效果评估、产品业务化运行提供依据。     

基于VisIt的全球科学数据并行可视化——以大气温度场为例

针对当前大规模全球科学数据可视化中存在的单机可视化数据量有限、从底层开发并行可视化系统难度大等问题,该文基于分布式环境和VisIt,提出了一种简便、开放而又有效的大规模全球科学数据可视化方法。介绍了VisIt的体系结构及运行机制,给出了自定义数据的并行可视化方法;并基于NCEP数据集及全球空间格网,在小规模集群环境下实现了小粒度适应性球体退化八叉树格网(SDOG)下的全球大气温度场的并行可视化。VisIt的并行可视化性能测试结果表明:通过增加计算节点,VisIt能有效摆脱传统单机可视化对数据量的限制,可实现大规模全球科学数据的并行可视化。     

利用部分投影的二维CT旋转中心偏移快速校正

现有的CT重建算法都依赖于转台旋转中心的准确定位,否则会引起重建图像的伪影。目前已有的旋转中心估计方法的共同特点是利用180°或360°的投影数据。获取这些投影,需要比较长的扫描时间,占用比较多的资源。为此,在平行束扫描和扇形束扫描两种情形下,本文提出利用部分视角的投影来对转台旋转中心进行快速定位的方法。在平行束情形,利用物体空间质心与投影空间的质心的对应关系定理,由60°的投影即可准确测量出旋转中心的位置。在扇形束情形,通过将扇形束投影重排成平行束投影,仍可由扇形角再进行60°的投影准确测量出旋转中心的位置。     

多视密集匹配并行传播GPU-PatchMatch算法

针对多视密集匹配的效率较低的问题,提出了GPU-PatchMatch多视密集匹配算法。该算法使用GPU提高PatchMatch的计算效率;同时充分利用稀疏场景信息,对深度信息进行规则初始化;为提高传播效率,使用了金字塔红黑板并行传播深度信息。最后在DTU、Strecha和Vaihigen数据集上进行了试验,并与常用的多视密集匹配算法进行对比。试验结果表明,本文算法在重建效率上有较大提高,与CPU算法(PMVS、MVE、OpenMVS)相比有7倍以上提升,与GPU算法相比也有2.5倍以上提升,表明本文算法的有效性。