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.     

Magnetotelluric investigations for imaging electrical structure of Garhwal Himalayan corridor,Uttarakhand, India

Magnetotelluric investigations have been carried out in the Garhwal Himalayan corridor to delineate the electrical structure of the crust along a profile extending from Indo-Gangetic Plain to Higher Himalayan region in Uttarakhand, India. The profile passing through major Himalayan thrusts: Himalayan Frontal Thrust (HFF), Main Boundary Thrust (MBT) and Main Central Thrust (MCT), is nearly perpendicular to the regional geological strike. Data processing and impedance analysis indicate that out of 44 stations MT data recorded, only 27 stations data show in general, the validity of 2D assumption. The average geoelectric strike, N70°W, was estimated for the profile using tensor decomposition. 2D smooth geoelectrical model has been presented, which provides the electrical image of the shallow and deeper crustal structure. The major features of the model are (i) a low resistivity (<50Ωm), shallow feature interpreted as sediments of Siwalik and Indo-Gangetic Plain, (ii) highly resistive (> 1000Ωm) zone below the sediments at a depth of 6 km, interpreted as the top surface of the Indian plate, (iii) a low resistivity (< 10Ωm) below the depth of 6 km near MCT zone coincides with the intense micro-seismic activity in the region. The zone is interpreted as the partial melting or fluid phase at mid crustal depth. Sensitivity test indicates that the major features of the geoelectrical model are relevant and desired by the MT data.     

Nitrogen forms and decomposition of organic carbon in the southern Bohai Sea core sediments

Study on form characteristics of nitrogen in marine sediments is the primary method to research its biogeochemical cycling and nitrogen form characteristics in core sediments can reflect the process and results of early diagenesis in a certain degree. In this paper, Sequential extraction process in natural grain size was used for studying the existent forms of nitrogen in five core sediments of the southern Bohai Sea for the first time. Nitrogen was divided into two parts -transferable and fixed based on whether it could be extracted by the reagent. Distributions and early diagenesis of transferable nitrogen forms in the southern Bohai Sea were researched integratedly. Results indicate that IEF - N and OSF-N are predominant forms in transferable part in the studied core sediments. Contents of different nitrogen forms vary differently with depth, and have different diagenesis process. Decomposition constant of organic nitrogen (ON) and OC are about 15.51 × 10~(-3)a~(-1)and 4.79× 10~(-3)a~(-1) respective     

关于多项式特征值的条件数

Tisseur(2000)研究了非齐次多项式特征值问题的条件数，所得结果的不足是对无穷特征值必须另外处理；Declieu和Tisseur(2003)运用微分几何方法研究了齐次多项式特征值问题的条件数，所得结果的不足之处是与系数矩阵的尺度化(scaling)有关。这两种条件数都有各自的应用范围。本文基于投影尺度引入齐次多项式特征值问题的条件数，其优点是与系数矩阵的尺度化无关，因而也许会有较广泛的应用。     

低硼氢氧化镁的制备及热分解动力学参数

本文以水合氯化镁、工业氨水为原料制备出了低硼含量的纯度 >99%的氢氧化镁和氧化镁。利用差热 -热重 (DTA- TG)、X-射线衍射 (XRD)、透射电子显微镜 (TEM)等手段 ,对氢氧化镁的热稳定性、相结构、表观状态以及脱水动力学参数进行了分析。结果为 :氢氧化镁以束状纤维和片状粉末共存 ,在 6 2 3K基本完成脱水反应生成氧化镁 ,其脱水反应的表观活化能 E为135 .4 k J· mol-1,指前因子 A为 5 .15× 10 10 ,反应级数 n为 1.10     

基于EMD与神经网络的机械故障诊断技术

经验模式分解 (EMD)是分析非线性、非平稳信号的有力工具 ,它将信号分解为突出了原信号的不同时间尺度的局部特征信息的内在模函数 (IMF)分量。本文通过将各 IMF分量输入到 BP网络中进行训练学习和故障诊断 ,比直接输入原信号可以提高 BP网络对故障诊断的准确率 ,而且减少了训练时间。     

二次特征值问题的最佳向后扰动分析

二次特征值问题 (QEP)的主要的求解方法之一是转化为广义特征值问题 (GEP) ,然后用求解广义特征值的方法 (比如 QZ方法 )求解。本文研究由此获得的计算解的范数意义下的最佳向后扰动分析 ,所得结果是 Tisseur最近所得结果的加强。     

Symbiotic Relations of Sediment-Agglutinating Nematodes and Bacteria in Detrital Habitats: The Enzyme-Sharing Concept

Abstract. A new concept (‘enzyme sharing’) concerning the interaction of marine nematodes and microbes in the degradation of sedimentary detritus is presented. Elements of this concept are (1) the notorious tendency of many aquatic nematodes to agglutinate detrital particles by mucus secretions, (2) new observations on the stimulation of microbial growth by nematodes in agar plates, and (3) literature data on limited endogenous proteolytic capacities of aquatic nematodes.
Observations on nematode‐microbe associations in agar plates prompted the conceptual synthesis. In agar medium without the addition of any nutrients a spectacular growth of bacteria was visible on the sinusoidal crawling trails of nematodes only 2 – 3 days after introduction of the worms (species of Adoncholaimus, Anoplostoma and Sabatieria). Juveniles of Anoplostoma that hatched in the agar cultures left their minute trails in the medium and these were rapidly occupied by bacteria. The nematodes repeatedly visited their bacterial trails, which persisted as a peculiar biotic structure for more than one year and survived the nematodes.
In sterile agar preparations containing the fluorogenic methylumbelliferyl‐β‐glucoside in the presence of the nematode Adoncholaimus, an enhanced fluorescence of the medium was visible, indicating β‐glucosidase activity. We therefore assume that oncholaimid nematodes discharge enzymes that alone, or in concert with microbial activities, contribute to the hydrolytic cleavage of refractory polysaccharides containing β‐glucosidic bonds such as agar components and cellulose. The sugars thus produced may then be taken up by the nematodes and concomitantly support the conspicuous growth of microbes. Since we did not observe any feeding of the nematodes on the associated microbes in agar plates, we question the nutritive potential of intact microbial cells for a number of nematodes abounding in detrital habitats, and call attention to the significance of ambient dissolved or adsorbed organic monomeric nutrients.
Consequently, we perceive the puzzling perpetual accretion of detrital organic particles to sediment agglutinations by nematodes as an adaptation for operating an ‘enzymatic reactor’ for the production of dissolved nutrients. We hypothesise a relationship of mutual commensalism of nematodes and heterotrophic microbes in detrital habitats and propose the term ‘enzyme sharing’ for this relationship. Both parties invest in a common enzyme pool that decomposes organic detritus for their nutrition. We present here evidence that nematodes contribute β‐glucosidase, which is involved in the cellulase system. Data from the literature suggest that microbial enzymatic processing of detrital proteins yields amino acids available to nematodes, which apparently have no efficient proteolytic enzyme system in their intestines.     

基于坐标归一化和奇异值分解的直接线性变换解法

直接线性变换方法是数字摄影测量和计算机视觉领域最常用的解析处理方法之一,特别适用于未检校数码相机的三维测量,但直接线性参数间的相关性、物方控制的约束和设计矩阵元素数量级的较大差异,均可导致法矩阵严重病态,从而影响解的稳定性。本文借鉴改进的八点基本矩阵估计算法,采用基于坐标归一化和奇异值分解的解法,即首先将像点和物点坐标进行相似变换得到归一化坐标后组成法矩阵,其次利用矩阵奇异值分解方法代替常规的最小二乘方法,模拟和真实数据表明,此方法可以有效提高解算精度和稳定性。     

经验模式分解的尺度阈值滤波在变形分析中的应用

本文将信号处理领域的经验模式分解算法应用于变形信息的提取中,通过引入阈值函数,建立了基于经验模式分解的尺度阈值滤波模型,采用优化模型确定了经验模式分解的次数。分别通过模拟试验和实测数据与小波阈值法和多项式拟合法进行了比对,分析表明:在低噪声情况下,三种方法都有一定的滤波效果;在高噪声情况下,经验模式分解的尺度阈值滤波法具有与小波阈值去噪法相等的精度,而且瞬时强噪声识别能力更好,优于多项式拟合法。