Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed clustering methods”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 901 records · Page 50

Reducing Communication in Graph Neural Network Training

Graph Neural Networks (GNNs) are powerful and flexible neural networks that use the naturally sparse connectivity information of the data. GNNs represent this connectivity as sparse matrices, which have lower arithmetic intensity and thus higher communication costs compared to dense matrices, making GNNs harder to scale to high concurrencies than convolutional or fully-connected neural networks. Here, we introduce a family of parallel algorithms for training GNNs and show that they can asymptotically reduce communication compared to previous parallel GNN training methods. We implement these algorithms, which are based on 1D, 1. 5D, 2D, and 3D sparse-dense matrix multiplication, using torch.distributed on GPU-equipped clusters. Our algorithms optimize communication across the full GNN training pipeline. We train GNNs on over a hundred GPUs on multiple datasets, including a protein network with over a billion edges.

97 MATHEMATICS AND COMPUTING↗

Robust Technique for Measuring and Simulating Silicon Wafer Quality Characteristics that Enable the Prediction of Solar Cell Electrical Performance of MEMC Silicon Wafer. Cooperative Research and Development Final Report, CRADA Number CRD-11-438

NREL and MEMC Electronic Materials are interested in developing a robust technique for monitoring material quality of mc-Si and mono-Si wafers -- a technique that can provide relevant data to accurately predict the performance of solar cells fabricated on them. Previous work, performed under two TSAs between NREL and MEMC, has established that dislocation clusters are the dominant performance-limiting factor in MEMC mc-Si solar cells. The work under this CRADA will go further in verifying these results on a larger data set, evaluate possibilities of faster method(s) for mapping dislocations in wafers/ingots, understanding dislocation generation during ingot casting, and helping MEMC to have an internal capability for basic characterization that will provide feedback needed for more accurate crystallization simulations. NREL has already developed dislocation mapping technique and developed a basic electronic model (called Network Model) that uses spatial distribution of dislocations to predict the cell performance. In this CRADA work, we will use these techniques to: (i) establish dislocation, grain size, and grain orientation distributions of the entire ingots (through appropriate DOE) and compare these with theoretical models developed by MEMC, (ii) determine concentrations of some relevant impurities in selected wafers, (iii) evaluate potential of using photoluminescence for dislocation mapping and identification of recombination centers, (iv) evaluate use of diode array analysis as a detailed characterization tool, and (v) establish dislocation mapping as a wafer-quality monitoring tool for commercial mc-Si production.

14 SOLAR ENERGY↗

SU-E-T-581: On the Value of LET as a Radiation Quality Descriptor for RBE

Purpose: To investigate the relationship between linear energy transfer (LET) and relative biological effectiveness (RBE) for protons and light ions, and the corresponding role of LET as a descriptor of radiation quality of hadron therapy. Methods: Monte Carlo (MC) proton and light ion (He, Li, C) tracks with LET < 30 eV nm{sup -1} were generated in an event-by-event mode. They were overlaid on a cell nucleus model containing 6×10{sup 9} nucleotide base pairs using an isotropic irradiation procedure that provides electronic equilibrium. Strand breaks (sbs) were scored in the DNA sugar-phosphate groups and further sub-classified into single or double sbs (ssbs or dsbs). Distributions of ssbs and dsbs for 2 Gy fractions were calculated to estimate RBE for the induction of initial dsbs with reference to {sup 60}Co. Additionally, sbs were classified based on their complexity (i.e. the number of sbs in each cluster). Results: An increase in LET for light ions of the same atomic number or a decrease in atomic number for ions of the same LET resulted in a lower kinetic energy of emitted secondary electrons. The clustering of DNA damage was more pronounced as reflected by the increase in proton RBE from ∼ 1.75 to 4 for LET values of 7 to 28 eV nm{sup -1}. A significant RBE decrease between protons, He, Li and C ions of the same LET was also noticed as function of the atomic number. Significant differences in ssbs and dsbs complexities were also seen for particles with the same LET, potentially supporting a clustering-based radiation quality descriptor. Conclusion: The LET-RBE relationships were simulated for proton and light ions and exhibited expected trends, including different RBEs for particles with the same LET but different atomic numbers. A complexity based radiation quality descriptor may allow better differentiation of RBE between radiation fields of similar LET. We would like to acknowledge support from the Fonds de recherche du Quebec Sante (FRQS), from the CREATE Medical Physics Research Training Network grant (number 432290) of NSERC, support from NSERC under grants RGPIN 397711-11 and RGPIN-2014-06475 and support from CIHR under grants MOP-114910, MOP-136774 and MOP-102550.

61 RADIATION PROTECTION AND DOSIMETRY↗

Di-neutron correlations in {sup 7}H

We investigate the dineutron correlations in {sup 7}H. The ground state of {sup 7}H is solved by superposing many AMD (antisymmetrized molecular dynamics) Slater determinants with various t+n+n+n+n configurations based on AMD triple-S (AMD superposition of selected snapshots). The mixing of dineutron cluster components is estimated by calculating the squared overlap with the wave function of the dineutron condensate, which is an extension of the studies on the {alpha} condensate in 4N nuclei. The calculated results show significant mixing of components of dineutron clusters (t+{sup 2}n+{sup 2}n) with a spatially extended distribution ({approx}70%) much larger than that of {sup 8}He, where the core nucleus is changed from t to {sup 4}He.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

The Quantum Liquid of Alpha Clusters - A Variational Approach

Within the variational approach of Bose liquids we analyze the g.s. energy of charge neutral alpha matter at T = 0. As a prerequisite for such calculation we take from the literature or propose new {alpha}-{alpha} potentials that are particularly suitable for this task, i.e. posses a repulsive core and/or reproduce the low energy scattering data and the resonance properties of the {alpha}-{alpha} system. The alpha matter EOS is then obtained with the HNC method using Pandharipande-Bethe correlation derived variationally in the lowest order expansion of the energy functional or a simple gaussian function with a healing range determined by the normalization of the radial distribution function in the lowest order. We show that saturation is achieved only via repulsive and shallow potentials that are not consistent with the scattering and resonance constraints.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Accurate ab initio potential energy surface, thermochemistry, and dynamics of the F{sup −} + CH{sub 3}F S{sub N}2 and proton-abstraction reactions

We develop a full-dimensional global analytical potential energy surface (PES) for the F{sup −} + CH{sub 3}F reaction by fitting about 50 000 energy points obtained by an explicitly correlated composite method based on the second-order Møller–Plesset perturbation-F12 and coupled-cluster singles, doubles, and perturbative triples-F12a methods and the cc-pVnZ-F12 [n = D, T] basis sets. The PES accurately describes the (a) back-side attack Walden inversion mechanism involving the pre- and post-reaction (b) ion-dipole and (c) hydrogen-bonded complexes, the configuration-retaining (d) front-side attack and (e) double-inversion substitution pathways, as well as (f) the proton-abstraction channel. The benchmark quality relative energies of all the important stationary points are computed using the focal-point analysis (FPA) approach considering electron correlation up to coupled-cluster singles, doubles, triples, and perturbative quadruples method, extrapolation to the complete basis set limit, core-valence correlation, and scalar relativistic effects. The FPA classical(adiabatic) barrier heights of (a), (d), and (e) are −0.45(−0.61), 46.07(45.16), and 29.18(26.07) kcal mol{sup −1}, respectively, the dissociation energies of (b) and (c) are 13.81(13.56) and 13.73(13.52) kcal mol{sup −1}, respectively, and the endothermicity of (f) is 42.54(38.11) kcal mol{sup −1}. Quasiclassical trajectory computations of cross sections, scattering (θ) and initial attack (α) angle distributions, as well as translational and internal energy distributions are performed for the F{sup −} + CH{sub 3}F(v = 0) reaction using the new PES. Apart from low collision energies (E{sub coll}), the S{sub N}2 excitation function is nearly constant, the abstraction cross sections rapidly increase with E{sub coll} from a threshold of ∼40 kcal mol{sup −1}, and retention trajectories via double inversion are found above E{sub coll} = ∼ 30 kcal mol{sup −1}, and at E{sub coll} = ∼ 50 kcal mol{sup −1}, the front-side attack cross sections start to increase very rapidly. At low E{sub coll}, the indirect mechanism dominates (mainly isotropic backward-forward symmetric θ distribution and translationally cold products) and significant long-range orientation effects (isotropic α distribution) and barrier recrossings are found. At higher E{sub coll}, the S{sub N}2 reaction mainly proceeds with direct rebound mechanism (backward scattering and hot product translation)

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A New Spin for Understanding the Peculiar Horizontal Branch Morphology of the Galactic Globular Clusters NGC 6388 and NGC 6441

In this paper we present multiband optical and UV Hubble Space Telescope photometry of the two Galactic globular clusters NGC 6388 and NGC 6441 Aims. We investigate the properties of their anomalous horizontal branches (HB) in different photometric planes in order to shed light on the nature of the physical mechanism(s) responsible for the existence of an extended HB blue tail, and of a slope in the HB, visible in all the color-magnitude diagrams. Methods. New photometric data have been collected and carefully reduced. Empirical data have been compared with updated stellar models of low-mass, metal-rich, He-burning structures, transformed to the observational plane with appropriate atmosphere models. Results. We have obtained the first UV color-magnitude diagrams for NGC 6388 and NGC 6441. These diagrams confirm previous results, obtained in optical bands, about the presence of a sizeable stellar population of extremely hot Horizontal Branch stars. At least in NGC 6388, we find a clear indication that at the hot end of the horizontal branch the distribution of stars forms a hook-like feature, closely resembling those observed in NGC 2808 and w Centauri. We briefly review the theoretical scenarios which have been suggested for interpreting this observational feature. We investigate also on the tilt in the horizontal branch morphology, and provide further evidence that supports early suggestions according to which this feature cannot be interpreted as an effect of differential reddening or radiative levitation, though these effects contribute to create the anomaly. We demonstrate that a possible solution of the puzzle is to assume that a small fraction (approx. 13% in NGC 6388 and approx. 8% NGC 6441) of the stellar population in the two clusters is strongly helium enriched (Y approx. 0.40 in NGC6388 and Y approx. 0.35 in NGC 6441). This solution necessarily implies the presence of a double generation of stars in the two clusters.

Busso, G.↗

Application of unsupervised deep learning to image segmentation and in-situ contact angle measurements in a CO 2 -water-rock system

Rock surface wettability is a critical property that regulates multiphase flows in porous media, which can be quantified using the surface contact angle (CA). X-ray micro-computed tomography (μCT) provides an effective approach to in-situ measurements of surface CAs. However, the CA measurement accuracy depends significantly on the quality of CT image segmentation, which is the clustering of CT pixels into separate phases. Inspired by this, we developed a deep learning (DL)-based CA measurement workflow. Motivated by the recent tremendous progress in unsupervised learning techniques and aiming to avoid expensive manual data annotations, an unsupervised DL pipeline for CT image segmentation was proposed and implemented, which includes unsupervised model training and post-processing. The unsupervised model training was driven by a novel loss function constrained with feature similarity and spatial continuity and implemented by iterative forward and backward paths; the former clustered the pixel-wise feature vectors extracted by convolution neural networks, whereas the latter updated the parameters using gradient descent. An over-segmentation strategy was adopted for model training. The post-processing steps based on agglomerative hierarchical clustering (AHC) were implemented to further merge the over-segmented model output to the desired cluster number, which is intended to improve the efficiency of image segmentation. The developed unsupervised DL pipeline was compared with other commonly-used image segmentation methods using pixel-wise and physics-based evaluation metrics on a synthetic raw-image dataset, which had a known ground truth. The unsupervised DL pipeline showed the best performance. Next, the segmented images were input to an automatic CA measurement tool, and the results were validated by comparisons with manual measurements. The CA values from the manual and automatic measurements showed similar distributions and statistical properties. The automatic measurement demonstrated a wider spectrum because of the much larger number of measurement data points. The primary novelty of the unsupervised DL pipeline developed in this study lies in the novel loss function and the over-segmentation strategy associated with AHC post-processing. Finally, the workflow has been proven an efficient tool for pore-scale wettability characterization, which has a wide range of applications in fundamental studies of multiphase flows in natural porous media, which have critical implications to geological carbon sequestration, hydrocarbon energy recovery, and contaminant transport in groundwater.

42 ENGINEERING↗

DISENTANGLING OVERLAPPING ASTRONOMICAL SOURCES USING SPATIAL AND SPECTRAL INFORMATION

We present a powerful new algorithm that combines both spatial information (event locations and the point-spread function) and spectral information (photon energies) to separate photons from overlapping sources. We use Bayesian statistical methods to simultaneously infer the number of overlapping sources, to probabilistically separate the photons among the sources, and to fit the parameters describing the individual sources. Using the Bayesian joint posterior distribution, we are able to coherently quantify the uncertainties associated with all these parameters. The advantages of combining spatial and spectral information are demonstrated through a simulation study. The utility of the approach is then illustrated by analysis of observations of FK Aqr and FL Aqr with the XMM-Newton Observatory and the central region of the Orion Nebula Cluster with the Chandra X-ray Observatory.

79 ASTRONOMY AND ASTROPHYSICS↗

Are dark energy models with variable EoS parameter w compatible with the late inhomogeneous Universe?

We study the late-time evolution of the Universe where dark energy (DE) is presented by a barotropic fluid on top of cold dark matter (CDM) . We also take into account the radiation content of the Universe. Here by the late stage of the evolution we refer to the epoch where CDM is already clustered into inhomogeneously distributed discrete structures (galaxies, groups and clusters of galaxies). Under this condition the mechanical approach is an adequate tool to study the Universe deep inside the cell of uniformity. More precisely, we study scalar perturbations of the FLRW metric due to inhomogeneities of CDM as well as fluctuations of radiation and DE. For an arbitrary equation of state for DE we obtain a system of equations for the scalar perturbations within the mechanical approach. First, in the case of a constant DE equation of state parameter w, we demonstrate that our method singles out the cosmological constant as the only viable dark energy candidate. Then, we apply our approach to variable equation of state parameters in the form of three different linear parametrizations of w, e.g., the Chevallier-Polarski-Linder perfect fluid model. We conclude that all these models are incompatible with the theory of scalar perturbations in the late Universe.

79 ASTRONOMY AND ASTROPHYSICS↗

Towards optimal extraction of cosmological information from nonlinear data

One of the main unsolved problems of cosmology is how to maximize the extraction of information from nonlinear data. If the data are nonlinear the usual approach is to employ a sequence of statistics (N-point statistics, counting statistics of clusters, density peaks or voids etc.), along with the corresponding covariance matrices. However, this approach is computationally prohibitive and has not been shown to be exhaustive in terms of information content. Here we instead develop a hierarchical Bayesian approach, expanding the likelihood around the maximum posterior of linear modes, which we solve for using optimization methods. By integrating out the modes using perturbative expansion of the likelihood we construct an initial power spectrum estimator, which for a fixed forward model contains all the cosmological information if the initial modes are gaussian distributed. We develop a method to construct the window and covariance matrix such that the estimator is explicitly unbiased and nearly optimal. We then generalize the method to include the forward model parameters, including cosmological and nuisance parameters, and primordial non-gaussianity. We apply the method in the simplified context of nonlinear structure formation, using either simplified 2-LPT dynamics or N-body simulations as the nonlinear mapping between linear and nonlinear density, and 2-LPT dynamics in the optimization steps used to reconstruct the initial density modes. We demonstrate that the method gives an unbiased estimator of the initial power spectrum, providing among other a near optimal reconstruction of linear baryonic acoustic oscillations.

79 ASTRONOMY AND ASTROPHYSICS↗

Identification of a typical flight patterns

Method and system for analyzing aircraft data, including multiple selected flight parameters for a selected phase of a selected flight, and for determining when the selected phase of the selected flight is atypical, when compared with corresponding data for the same phase for other similar flights. A flight signature is computed using continuous- valued and discrete-valued flight parameters for the selected flight parameters and is optionally compared with a statistical distribution of other observed flight signatures, yielding a typicality scores for the same phase for other similar flights. A cluster analysis is optionally applied to the flight signatures to define an optimal collection of clusters. A level of atypicality for a selected flight is estimated, based upon an index associated with the cluster analysis.

Irving C Statler↗

Identification of atypical flight patterns

Method and system for analyzing aircraft data, including multiple selected flight parameters for a selected phase of a selected flight, and for determining when the selected phase of the selected flight is atypical, when compared with corresponding data for the same phase for other similar flights. A flight signature is computed using continuous-valued and discrete-valued flight parameters for the selected flight parameters and is optionally compared with a statistical distribution of other observed flight signatures, yielding atypicality scores for the same phase for other similar flights. A cluster analysis is optionally applied to the flight signatures to define an optimal collection of clusters. A level of atypicality for a selected flight is estimated, based upon an index associated with the cluster analysis.

Statler, Irving C.↗

A Catalog of Galaxy Clusters Observed by XMM-Newton

Images and the radial profiles of the temperature, abundance, and brightness for 70 clusters of galaxies observed by XMM-Newton are presented along with a detailed discussion of the data reduction and analysis methods, including background modeling, which were used in the processing. Proper consideration of the various background components is vital to extend the reliable determination of cluster parameters to the largest possible cluster radii. The various components of the background including the quiescent particle background, cosmic diffuse emission, soft proton contamination, and solar wind charge exchange emission are discussed along with suggested means of their identification, filtering, and/or their modeling and subtraction. Every component is spectrally variable, sometimes significantly so, and all components except the cosmic background are temporally variable as well. The distributions of the events over the FOV vary between the components, and some distributions vary with energy. The scientific results from observations of low surface brightness objects and the diffuse background itself can be strongly affected by these background components and therefore great care should be taken in their consideration.

Snowden, S. L.↗

An Orthogonal Recursive Bisection (ORB) Based Time Advancement Algorithm for CFD-DEM Solvers

The time integration of the granular phase in coupled computational fluid dynamics (CFD) – discrete element method (DEM) simulations presents a unique computational challenge brought about by the large variations in particle collisional time scales. Particles in the dilute regions of the computational domain can be advanced with large time steps while dense regions require much smaller time increments. However, the time step size in most solvers is globally set as the limit for accuracy and stability imposed by the collisions and is typically orders of magnitude less than that required away from collisions. This work addresses this precise issue and provides a strategy to avoid the use of a global conservative small time step size for the entire set of particles.A novel time stepping algorithm for CFD-DEM solvers using a partitioning approach using orthogonal recursive bisection (ORB) that allows for variable time steps among particles is described and its computational performance is compared against baseline explicit methods, typically used in several CFD-DEM solvers. ORB has advantages of being relatively quick and easy to update incrementally and has the required heuristic behavior (i.e., it will split the region in half with a cluster on each side) when groups of particles are well separated (clustered). The algorithm presented in this work uses a local time stepping approach to resolve collisional time scales for subsets of particles that are present at the leaves of the ORB, thereby resulting in substantial reduction of computational cost. The parallel implementation of this method where a ``knapsack” algorithm is used in tandem with ORB for effective load-balancing is also presented, where a best possible partitioning is obtained based on number of particles and local time-stepping costs. The algorithm is tested against benchmark problems with varying particle distributions that include fluidized bed and riser flow scenarios. Preliminary results indicate that the approach is 2-3X faster than traditional explicit methods for problems that involve both dense and dilute regions, while maintaining the same level of accuracy.

adaptive timestepping↗

Parallel Optical Flow Processing of 4D Cardiac CT Data on Multicore Clusters, In: 2014 IEEE 17th International Conference on Computational Science and Engineering

Optical flow is the distribution of apparent velocities of movement of brightness patterns in a sequence of images. For large 3D image sequences, optical flow applications are time consuming and memory-bound. To cope with these problems, in this paper, we present parallel optical flow processing of 4D cardiac CT data on multicore cluster systems to significantly shorten the time for computing velocity fields of the heart in order to aid cardiologists in diagnosing heart disease such as myocardial infarction and cardiac dysrhythmia in time. First, we modify and extend two traditional 2D optical flow methods Horn-Schunck and Lucas-Kanade to three-dimensional cases to process the 4D cardiac CT data. Second, we extend Mat lab MPI to support parallel computing with Mat lab and Octave on these cluster systems. Then we develop the parallel Mat lab/Octave optical flow applications for the 4D cardiac CT data in detail. Our experimental results show that these parallel optical flow applications have good scalability with close to linear speedup, and are able to shorten the image processing time significantly from more than 5 hours on 4 cores to 1.5 minutes on 1024 cores.

Engineering↗

New particle formation in the sulfuric acid–dimethylamine–water system: reevaluation of CLOUD chamber measurements and comparison to an aerosol nucleation and growth model

Here, a recent CLOUD (Cosmics Leaving OUtdoor Droplets) chamber study showed that sulfuric acid and dimethylamine produce new aerosols very efficiently and yield particle formation rates that are compatible with boundary layer observations. These previously published new particle formation (NPF) rates are reanalyzed in the present study with an advanced method. The results show that the NPF rates at 1.7 nm are more than a factor of 10 faster than previously published due to earlier approximations in correcting particle measurements made at a larger detection threshold. The revised NPF rates agree almost perfectly with calculated rates from a kinetic aerosol model at different sizes (1.7 and 4.3 nm mobility diameter). In addition, modeled and measured size distributions show good agreement over a wide range of sizes (up toca. 30 nm). Furthermore, the aerosol model is modified such that evaporation rates for some clusters can be taken into account; these evaporation rates were previously published from a flow tube study. Using this model, the findings from the present study and the flow tube experiment can be brought into good agreement for the high base-to-acid ratios ( ~ 100)relevant for this study. This confirms that nucleation proceeds at rates that are compatible with collision-controlled (a.k.a.kinetically controlled) NPF for the conditions during the CLOUD7 experiment (278 K, 38 % relative humidity, sulfuric acid concentration between 1 × 10 6 and 3 × 10 7 cm -3 , and dimethylamine mixing ratio of ~ 40 pptv, i.e., 1 × 10 9 cm -3 ) .

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Mercury Binding Sites in Thiol-Functionalized Mesostructured Silica

Thiol-functionalized mesostructured silica with anhydrous compositions of (SiO{sub 2}){sub 1-x}(LSiO{sub 1.5}){sub x}, where L is a mercaptopropyl group and x is the fraction of functionalized framework silicon centers, are effective trapping agents for the removal of mercuric(II) ions from water. In the present work, we investigate the mercury-binding mechanism for representative thiol-functionalized mesostructures by atomic pair distribution function (PDF) analysis of synchrotron X-ray powder diffraction data and by Raman spectroscopy. The mesostructures with wormhole framework structures and compositions corresponding to x = 0.30 and 0.50 were prepared by direct assembly methods in the presence of a structure-directing amine porogen. PDF analyses of five mercury-loaded compositions with Hg/S ratios of 0.50-1.30 provided evidence for the bridging of thiolate sulfur atoms to two metal ion centers and the formation of chain structures on the pore surfaces. We find no evidence for Hg-O bonds and can rule out oxygen coordination of the mercury at greater than the 10% level. The relative intensities of the PDF peaks corresponding to Hg-S and Hg-Hg atomic pairs indicate that the mercury centers cluster on the functionalized surfaces by virtue of thiolate bridging, regardless of the overall mercury loading. However, the Raman results indicate that the complexation of mercury centers by thiolate depends on the mercury loading. At low mercury loadings (Hg/S {le} 0.5), the dominant species is an electrically neutral complex in which mercury most likely is tetrahedrally coordinated to bridging thiolate ligands, as in Hg(SBu{sup t}){sub 2}. At higher loadings (Hg/S 1.0-1.3), mercury complex cations predominate, as evidenced by the presence of charge-balancing anions (nitrate) on the surface. This cationic form of bound mercury is assigned a linear coordination to two bridging thiolate ligands.

43 PARTICLE ACCELERATORS↗