Search NASA⌕ Search

SEARCH · Search NASA

Results for “data dependencies”

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 127 records · Page 7

Exploring performance and energy tradeoffs for irregular applications: A case study on the Tilera many-core architecture

High performance, parallel applications with irregular data accesses are becoming a critical workload class for modern systems. In particular, the execution of such workloads on emerging many-core systems is expected to be a significant component of applications in data mining, machine learning, scientific computing and graph analytics. However, power and energy constraints limit the capabilities of individual cores, memory hierarchy and on-chip interconnect of such systems, thus leading to architectural and software trade-os that must be understood in the context of the intended application’s behavior. Irregular applications are notoriously hard to optimize given their data-dependent access patterns, lack of structured locality and complex data structures and code patterns. We have ported two irregular applications, graph community detection using the Louvain method (Grappolo) and high-performance conjugate gradient (HPCCG), to the Tilera many-core system and have conducted a detailed study of platform-independent and platform-specific optimizations that improve their performance as well as reduce their overall energy consumption. To conduct this study, we employ an auto-tuning based approach that explores the optimization design space along three dimensions - memory layout schemes, GCC compiler flag choices and OpenMP loop scheduling options. We leverage MIT’s OpenTuner auto-tuning framework to explore and recommend energy optimal choices for different combinations of parameters. We then conduct an in-depth architectural characterization to understand the memory behavior of the selected workloads. Finally, we perform a correlation study to demonstrate the interplay between the hardware behavior and application characteristics. Using auto-tuning, we demonstrate whole-node energy savings and performance improvements of up to 49:6% and 60% relative to a baseline instantiation, and up to 31% and 45:4% relative to manually optimized variants.

auto tuning↗

Voltage-dependent excitation dynamics in UV-absorbing organic photovoltaics with efficient charge transfer exciton emission

Intermolecular charge-transfer excitons play a central role in determining the performance of organic solar cells as their voltage-dependent formation, dissociation, and recombination dynamics contribute to photocurrent generation, radiative/nonradiative voltage losses, and photovoltaic fill factor. Here, we explore the properties of brightly-emitting wide energy gap (>2 eV) charge transfer excitons by measuring the voltage-dependent photoluminescence, photocurrent, and ultrafast pump–probe transient absorption spectra of organic solar cells employing five UV-absorbing donor molecules that differ only by the length of the oligophenylene or acene group at their core. We find that organic solar cells with a strong correlation between their voltage-dependent photocurrent and charge-transfer exciton photoluminescence have low photovoltaic fill factors as they require voltage to facilitate efficient charge-transfer exciton dissociation. In contrast, solar cells that are efficient can readily generate charges without an applied field and have a separate population of tightly-bound charge-transfer excitons that are responsible for emission. Furthermore, considering that the sum of all excitation loss rates (i.e., recombination and charge extraction) must be equal to the excitation generation rate in the steady state, these voltage-dependent data allow us to solve for the voltage-dependent fate of all excitations in the solar cells and estimate upper and lower bounds for geminate and non-geminate recombination, respectively.

14 SOLAR ENERGY↗

The HEP.TrkX Project: deep neural networks for HL-LHC online and offline tracking

Particle track reconstruction in dense environments such as the detectors of the High Luminosity Large Hadron Collider (HL-LHC) is a challenging pattern recognition problem. Traditional tracking algorithms such as the combinatorial Kalman Filter have been used with great success in LHC experiments for years. However, these state-of-the-art techniques are inherently sequential and scale poorly with the expected increases in detector occupancy in the HL-LHC conditions. The HEP.TrkX project is a pilot project with the aim to identify and develop cross-experiment solutions based on machine learning algorithms for track reconstruction. Machine learning algorithms bring a lot of potential to this problem thanks to their capability to model complex non-linear data dependencies, to learn effective representations of high-dimensional data through training, and to parallelize easily on high-throughput architectures such as GPUs. This contribution will describe our initial explorations into this relatively unexplored idea space. Furthermore, we will discuss the use of recurrent (LSTM) and convolutional neural networks to find and fit tracks in toy detector data.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Practices and Standards for Data and Processing at the APS

The handling of datasets at scientific user facilities is becoming progressively more challenging as advances in sources and detectors drive increasingly aggressive data acquisition rates. The ability to share and process these data depends on the ability of the originator of the data and the persons making use of that data to unequivocally agree on what these data mean. This understanding is driven by use of standardized data formats.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

The HEP.TrkX Project: Deep Learning for Particle Tracking

Charged particle reconstruction in dense environments, such as the detectors of the High Luminosity Large Hadron Collider (HL-LHC) is a challenging pattern recognition problem. Traditional tracking algorithms, such as the combinatorial Kalman Filter, have been used with great success in HEP experiments for years. However, these state-of-the-art techniques are inherently sequential and scale quadratically or worse with increased detector occupancy. The HEP.TrkX project is a pilot project with the aim to identify and develop cross-experiment solutions based on machine learning algorithms for track reconstruction. Machine learning algorithms bring a lot of potential to this problem thanks to their capability to model complex non-linear data dependencies, to learn effective representations of high-dimensional data through training, and to parallelize easily on high-throughput architectures such as FPGAs or GPUs. In this paper we present the evolution and performance of our recurrent (LSTM) and convolutional neural networks moving from basic 2D models to more complex models and the challenges of scaling up to realistic dimensionality/sparsity.

Tsaris, Aristeidis↗

Optimizing Irregular Applications for Energy and Performance on the Tilera Many-core Architecture

Optimizing applications simultaneously for energy and performance is a complex problem. High performance, parallel, irregular applications are notoriously hard to optimize due to their data-dependent memory accesses, lack of structured locality and complex data structures and code patterns. Irregular kernels are growing in importance in applications such as machine learning, graph analytics and combinatorial scientific computing. Performance- and energy-efficient implementation of these kernels on modern, energy efficient, multicore and many-core platforms is therefore an important and challenging problem. We present results from optimizing two irregular applications { the Louvain method for community detection (Grappolo), and high-performance conjugate gradient (HPCCG) { on the Tilera many-core system. We have significantly extended MIT's OpenTuner auto-tuning framework to conduct a detailed study of platform-independent and platform-specific optimizations to improve performance as well as reduce total energy consumption. We explore the optimization design space along three dimensions: memory layout schemes, compiler-based code transformations, and optimization of parallel loop schedules. Using auto-tuning, we demonstrate whole node energy savings of up to 41% relative to a baseline instantiation, and up to 31% relative to manually optimized variants.

irregular applications, energy opttimization, many↗

Effects of irradiation and annealing on deep levels in rhodium-doped p-GaAs grown by metal-organic chemical-vapor deposition

This paper reports a detailed study of the effects of irradiation and thermal annealing on deep levels in Rh-doped p-type GaAs grown by low-pressure metal-organic chemical-vapor deposition, using deep level transient spectroscopy (DLTS) technique. It is found upon irradiation with alpha particles that, in addition to the radiation-induced defect peaks, all the Rh-related peaks observed in majority, as well as minority-carrier emission DLTS scans show an increase in their respective concentrations. The usually observed {alpha}-induced defects H{alpha}1, H{alpha}2, and H{alpha}3 are found to have lower introduction rates in Rh-doped samples, as compared to reference samples (not doped with Rh). Alpha-irradiation has been found to decompose the two minority carrier emitting bands (one at low temperature {approx}150 K and the other at {approx}380 K) observed prior to irradiation into distinct peaks corresponding to deep levels Rh1 and Rh2 and EL2 and Rh3, respectively. A similar effect is also observed for the majority-carrier emitting band composed of hole emission from deep levels RhA and RhB, which separate out well upon irradiation. Further, from the double-correlation DLTS measurements, the emission rates of carriers from the radiation-enhanced peaks corresponding to deep levels Rh1, Rh2, Rh3, and RhC were found to be dependent on junction electric field. For RhC, the field dependence data have been analyzed in terms of the Poole-Frenkel model employing a 3-dimensional Coulomb potential with q = 2e (electronic charge). Temperature dependence of the hole capture cross-sections of the levels RhA and RhC was also studied quantitatively. The observed dependence of the hole capture cross-section of RhC on temperature can be interpreted in terms of multiphonon capture model, yielding a capture barrier of 0.2 eV and {sigma}({infinity}) = 2.3 x 10{sup -14} cm{sup 2}. The results of irradiation and isochronal thermal annealing study, in combination with the theoretical analysis of the field dependence of hole emission data, lead us to interpret the levels RhA and RhB as charged centers and probably complexes of arsenic anti-site defect (As{sub Ga}) with Rh-impurity, (As{sub Ga}-Rh). The levels Rh1 and Rh2 are also proposed to be complexes, albeit of a different type, of arsenic anti-site defects (As{sub Ga}) and Rh-impurity (As{sub Ga}-Rh), while the level RhC is likely to be a doubly-charged complex center composed of arsenic interstitial (I{sub As}) and Rh-impurity (I{sub As}-Rh).

36 MATERIALS SCIENCE↗

FITTING THE UNION2.1 SUPERNOVA SAMPLE WITH THE R{sub h} = ct UNIVERSE

The analysis of Type Ia supernova data over the past decade has been a notable success story in cosmology. These standard candles offer us an unparalleled opportunity to study the cosmological expansion out to a redshift of {approx}1.5. The consensus today appears to be that {Lambda}CDM offers the best explanation for the luminosity-distance relationship seen in these events. However, a significant incompatibility is now emerging between the standard model and other equally important observations, such as those of the cosmic microwave background. {Lambda}CDM does not provide an accurate representation of the cosmological expansion at high redshifts (z >> 2). It is therefore essential to re-analyze the Type Ia supernova data in light of the cosmology (the R{sub h} = ct universe) that best represents the universe's dynamical evolution at early times. In this paper, we directly compare the distance relationship in {Lambda}CDM with that predicted by R{sub h} = ct, and each with the Union2.1 sample, and show that the two theories produce virtually indistinguishable profiles, though the fit with R{sub h} = ct has not yet been optimized. This is because the data cannot be determined independently of the assumed cosmology-the supernova luminosities must be evaluated by optimizing four parameters simultaneously with those in the adopted model. This renders the data compliant to the underlying theory, so the model-dependent data reduction should not be ignored in any comparative analysis between competing cosmologies. In this paper, we use R{sub h} = ct to fit the data reduced with {Lambda}CDM, and though quite promising, the match is not perfect. An even better fit would result with an optimization of the data using R{sub h} = ct from the beginning.

79 ASTRONOMY AND ASTROPHYSICS↗

The use of expert systems in combination with active and passive microwave data to classify sea ice

The authors report on a system for sea ice classification which combines passive and active microwave data with algorithmic and heuristic classification techniques in order to classify sea ice unambiguously. They describe initial algorithmic classification, extraction of features and symbolic information, the application of data-dependent expert systems, and the establishment of feature correspondence between data sets.

Haverkamp, Donna↗

Accelerating quantum optics experiments with statistical learning

Quantum optics experiments, involving the measurement of low-probability photon events, are known to be extremely time-consuming. We present a methodology for accelerating such experiments using physically motivated ansatzes together with simple statistical learning techniques such as Bayesian maximum a posteriori estimation based on few-shot data. We show that it is possible to reconstruct time-dependent data using a small number of detected photons, allowing for fast estimates in under a minute and providing a one-to-two order of magnitude speed-up in data acquisition time. We test our approach using real experimental data to retrieve the second order intensity correlation function, G (2) ($τ$), as a function of time delay τ between detector counts, for thermal light as well as anti-bunched light emitted by a quantum dot driven by periodic laser pulses. The proposed methodology has a wide range of applicability and has the potential to impact the scientific discovery process across a multitude of domains.

36 MATERIALS SCIENCE↗

Particle swarm optimization and gravitational wave data analysis: Performance on a binary inspiral testbed

The detection and estimation of gravitational wave signals belonging to a parameterized family of waveforms requires, in general, the numerical maximization of a data-dependent function of the signal parameters. Because of noise in the data, the function to be maximized is often highly multimodal with numerous local maxima. Searching for the global maximum then becomes computationally expensive, which in turn can limit the scientific scope of the search. Stochastic optimization is one possible approach to reducing computational costs in such applications. We report results from a first investigation of the particle swarm optimization method in this context. The method is applied to a test bed motivated by the problem of detection and estimation of a binary inspiral signal. Our results show that particle swarm optimization works well in the presence of high multimodality, making it a viable candidate method for further applications in gravitational wave data analysis.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Fractals and Spatial Methods for Mining Remote Sensing Imagery

The rapid increase in digital remote sensing and GIS data raises a critical problem -- how can such an enormous amount of data be handled and analyzed so that useful information can be derived quickly? Efficient handling and analysis of large spatial data sets is central to environmental research, particularly in global change studies that employ time series. Advances in large-scale environmental monitoring and modeling require not only high-quality data, but also reliable tools to analyze the various types of data. A major difficulty facing geographers and environmental scientists in environmental assessment and monitoring is that spatial analytical tools are not easily accessible. Although many spatial techniques have been described recently in the literature, they are typically presented in an analytical form and are difficult to transform to a numerical algorithm. Moreover, these spatial techniques are not necessarily designed for remote sensing and GIS applications, and research must be conducted to examine their applicability and effectiveness in different types of environmental applications. This poses a chicken-and-egg problem: on one hand we need more research to examine the usability of the newer techniques and tools, yet on the other hand, this type of research is difficult to conduct if the tools to be explored are not accessible. Another problem that is fundamental to environmental research are issues related to spatial scale. The scale issue is especially acute in the context of global change studies because of the need to integrate remote-sensing and other spatial data that are collected at different scales and resolutions. Extrapolation of results across broad spatial scales remains the most difficult problem in global environmental research. There is a need for basic characterization of the effects of scale on image data, and the techniques used to measure these effects must be developed and implemented to allow for a multiple scale assessment of the data before any useful process-oriented modeling involving scale-dependent data can be conducted. Through the support of research grants from NASA, we have developed a software module called ICAMS (Image Characterization And Modeling System) to address the need to develop innovative spatial techniques and make them available to the broader scientific communities. ICAMS provides new spatial techniques, such as fractal analysis, geostatistical functions, and multiscale analysis that are not easily available in commercial GIS/image processing software. By bundling newer spatial methods in a user-friendly software module, researchers can begin to test and experiment with the new spatial analysis methods and they can gauge scale effects using a variety of remote sensing imagery. In the following, we describe briefly the development of ICAMS and present application examples.

Lam, Nina↗

Extraction of $P_{11}$ Resonance from piN Data and Its Stability

We address a question about how much resonance poles and residues extracted from data depend on a model used for the extraction, and on the precision of data. We focus on the P{sub 11} {pi}-N scattering and use the dynamical coupled-channel (DCC) model developed in Excited Baryon Analysis Center (EBAC) at JLab. We examine the model-dependence of the poles by varying parameters largely within the EBAC-DCC model. We find that two poles associated with the Roper resonance are fairly stable against the variation. We also study the stability of the Roper poles against different analytic structure of the P{sub 11} amplitude below {pi}-N threshold by using a bare nucleon model. We again find a good stability of the Roper poles.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Extraction of P11 Resonance from pi-N Data and Its Stability

An important question about resonance extraction is how much resonance poles and residues extracted from data depend on a model used for the extraction, and on the precision of data. We address this question with the dynamical coupled-channel (DCC) model developed in Excited Baryon Analysis Center (EBAC) at JLab. We focus on the P11 pi-N scattering. We examine the model-dependence of the poles by varying parameters to a large extent within the EBAC-DCC model. We find that two poles associated with the Roper resonance are fairly stable against the variation. We also develop a model with a bare nucleon, thereby examining the stability of the Roper poles against different analytic structure of the P11 amplitude below pi-N threshold. We again find a good stability of the Roper poles.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Anomalous Hall Effect in Kagome Ferrimagnet GdMn 6 Sn 6

We present magnetotransport data on the ferrimagnet GdMn 6 Sn 6 . From the temperature-dependent data we are able to extract a large intrinsic contribution to the anomalous Hall effect σ int xz ~ 32 Ω –1 cm –1 and σ int xz ~ 223 Ω –1 cm –1 , which is comparable to values found in other systems also containing kagome nets of transition metals. From our transport anisotropy, as well as our density functional theory calculations, we argue that the system is electronically best described as a three-dimensional system. Furthermore we show that reduced dimensionality is not a strong requirement for obtaining large Berry phase contributions to transport properties. In addition, the coexistence of rare-earth and transition-metal magnetism makes the hexagonal MgFe 6 Ge 6 structure type a promising system to tune the electronic and magnetic properties in future studies.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Quantifying uncertainty in regional-scale seismic moment tensors

We examine the ability of three different inversion methods: 1) first motion (FM) inversions, 2) amplitude inversions, and 3) full waveform (FW) time variable moment tensor (TVMT) inversions, to recover an accurate source mechanism for simulated data for an earthquake as well as an explosion. The ability of inversion methods, such as the ones described above, to recover accurate models representative of the data depends on both a priori information as well as the quality of data being inverted. Therefore, we examine the effect that station geometry, geologic model, and noise level has on the inversion and estimated source mechanism. We find that FM data can provide more robust solutions than amplitude data and are not as sensitive to inaccurate earth models, especially in low-noise cases, and overall FW inversions are the most accurate out of the methods examined, but can still be biased by inaccurate earth models.

58 GEOSCIENCES↗

Deriving inertial wave characteristics from surface drifter velocities - Frequency variability in the tropical Pacific

Two techniques were developed for estimating statistics of inertial oscillations from satellite-tracked drifters that overcome the difficulties inherent in estimating such statistics from data dependent upon space coordinates that are a function of time. Application of these techniques to tropical surface drifter data collected during the NORPAX, EPOCS, and TOGA programs reveals a latitude-dependent, statistically significant 'blue shift' of inertial wave frequency. The latitudinal dependence of the blue shift is similar to predictions based on 'global' internal-wave spectral models, with a superposition of frequency shifting due to modification of the effective local inertial frequency by the presence of strongly sheared zonal mean currents within 12 deg of the equator.

Poulain, Pierre-Marie↗

Sensitivity of Temperature Profiles Retrieved from Mars Global Surveyor Thermal Emission Spectrometer (MGS/TES) Observations to the GSFC Synthetic Mars Model Atmosphere

Part of the task of interpreting IR spectral features observed by MGS/TES due to surface minerals requires distinguishing those IR signatures from atmospheric signatures of gas and dust. Surface-atmosphere separation for MGS/TES depends on knowledge of the retrieved temperature profile. In turn, the temperature retrieval Erom the observed data depends on molecular parameters including 15 micron CO2 line shape or line intensities which contribute to defining the Mars synthetic radiative transfer model. Using a simple isothermal, homogeneous single layer model of Pinnock and Shine, we find the ratio of (the error in degrees Kelvin of the retrieved temperature profile) to (the percentage error in the absorption coefficient) (deg K/percent) to be 0.4 at 200K. This ratio at 150K and 250K is 0.2 and 0.6, respectively. A more refined model, incorporating observed MGS/TES retrieved temperature profiles, the TES instrumental resolution and the most recent molecular modelling, will yield an improved knowledge of this error sensitivity. We present results of such a sensitivity study to determine the dependence of temperature profiles inverted from MGS/TES on these and other molecular parameters. This work was supported in part by NASA's Mars Data Analysis Program.

Maguire, William C.↗