Search NASA⌕ Search

SEARCH · Search NASA

Results for “reduced set 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 811 records · Page 45

Precision Measurement of the Proton Magnetic Form Factor at High Q2

Elastic electromagnetic form factors characterize the distribution of electric charge and magnetization current inside the nucleon and reflect the internal structure determined by Quantum Chromodynamics. Existing data on the proton magnetic form factor at high Q2 (the 4-momentum transfer squared) have large statistical and systematic uncertainties. The GMp experiment E12-07-108 was one of the first set of experiments to run in Hall A at Jefferson Lab after the 12 GeV upgrade with the goal of this experiment to precisely measure the electron-proton elastic cross section for Q2 up to 17 GeV2 with an accuracy of better than 2% - several times better than existing data in this Q2 range. This will allow further tests of form factor scaling predicted by pQCD and, additionally, will be an important benchmark for many other experiments where elastic electron-proton scattering is used for normalization. Since, GMp data were taken at lower ?(virtual photon polarization) than SLAC data, the measured cross sections recieve a smaller contribution from the electric form factor. This dissertation analyzed the GMp experiment data for five different kinematic settings determining the form factor with total uncertainties nearly two times smaller than the existing data at these Q2. This dissertation also presents the results of all Fall 2016 and Spring 2016 GMp experiment data. The proton magnetic form factors were extracted using both the Rosenbluth separation method and a parameterization of the form factor ratio (?pGpE=GpM) obtained from a fit to the existing cross section data. At the end of this thesis, the extracted GMp form factor data is compared to the existing data. The results for ?pGpE=GpM from Rosenbluth separations are found to be consistent with previous extractions, but with significantly reduced uncertainties, further highlighting the discrepancy with the polarization transfer technique. This data set can provide significant additional constraints on contributions from two-photon exchange.

Narayan Gautam, Thir↗

SALSA3D : a global 3D p-velocity model of the Earth's crust and mantle for improved event location.

To test the hypothesis that high quality 3D Earth models will produce seismic event locations which are more accurate and more precise, we are developing a global 3D P wave velocity model of the Earth's crust and mantle using seismic tomography. In this paper, we present the most recent version of our model, SALSA3D version 1.5, and demonstrate its ability to reduce mislocations for a large set of realizations derived from a carefully chosen set of globally-distributed ground truth events. Our model is derived from the latest version of the Ground Truth (GT) catalog of P and Pn travel time picks assembled by Los Alamos National Laboratory. To prevent over-weighting due to ray path redundancy and to reduce the computational burden, we cluster rays to produce representative rays. Reduction in the total number of ray paths is {approx}50%. The model is represented using the triangular tessellation system described by Ballard et al. (2009), which incorporates variable resolution in both the geographic and radial dimensions. For our starting model, we use a simplified two layer crustal model derived from the Crust 2.0 model over a uniform AK135 mantle. Sufficient damping is used to reduce velocity adjustments so that ray path changes between iterations are small. We obtain proper model smoothness by using progressive grid refinement, refining the grid only around areas with significant velocity changes from the starting model. At each grid refinement level except the last one we limit the number of iterations to prevent convergence thereby preserving aspects of broad features resolved at coarser resolutions. Our approach produces a smooth, multi-resolution model with node density appropriate to both ray coverage and the velocity gradients required by the data. This scheme is computationally expensive, so we use a distributed computing framework based on the Java Parallel Processing Framework, providing us with {approx}400 processors. Resolution of our model is assessed using a variation of the standard checkerboard method. We compare the travel-time prediction and location capabilities of SALSA3D to standard 1D models via location tests on a global event set with GT of 5 km or better. These events generally possess hundreds of Pn and P picks from which we generate different realizations of station distributions, yielding a range of azimuthal coverage and ratios of teleseismic to regional arrivals, with which we test the robustness and quality of relocation. The SALSA3D model reduces mislocation over standard 1D ak135 regardless of Pn to P ratio, with the improvement being most pronounced at higher azimuthal gaps.

99 GENERAL AND MISCELLANEOUS↗

A global 3D P-velocity model of the Earth's crust and mantle for improved event location : SALSA3D.

To test the hypothesis that high quality 3D Earth models will produce seismic event locations which are more accurate and more precise, we are developing a global 3D P wave velocity model of the Earth's crust and mantle using seismic tomography. In this paper, we present the most recent version of our model, SALSA3D version 1.5, and demonstrate its ability to reduce mislocations for a large set of realizations derived from a carefully chosen set of globally-distributed ground truth events. Our model is derived from the latest version of the Ground Truth (GT) catalog of P and Pn travel time picks assembled by Los Alamos National Laboratory. To prevent over-weighting due to ray path redundancy and to reduce the computational burden, we cluster rays to produce representative rays. Reduction in the total number of ray paths is {approx}50%. The model is represented using the triangular tessellation system described by Ballard et al. (2009), which incorporates variable resolution in both the geographic and radial dimensions. For our starting model, we use a simplified two layer crustal model derived from the Crust 2.0 model over a uniform AK135 mantle. Sufficient damping is used to reduce velocity adjustments so that ray path changes between iterations are small. We obtain proper model smoothness by using progressive grid refinement, refining the grid only around areas with significant velocity changes from the starting model. At each grid refinement level except the last one we limit the number of iterations to prevent convergence thereby preserving aspects of broad features resolved at coarser resolutions. Our approach produces a smooth, multi-resolution model with node density appropriate to both ray coverage and the velocity gradients required by the data. This scheme is computationally expensive, so we use a distributed computing framework based on the Java Parallel Processing Framework, providing us with {approx}400 processors. Resolution of our model is assessed using a variation of the standard checkerboard method. We compare the travel-time prediction and location capabilities of SALSA3D to standard 1D models via location tests on a global event set with GT of 5 km or better. These events generally possess hundreds of Pn and P picks from which we generate different realizations of station distributions, yielding a range of azimuthal coverage and ratios of teleseismic to regional arrivals, with which we test the robustness and quality of relocation. The SALSA3D model reduces mislocation over standard 1D ak135 regardless of Pn to P ratio, with the improvement being most pronounced at higher azimuthal gaps.

99 GENERAL AND MISCELLANEOUS↗

Improving efficiency of semi‐direct møller–plesset second‐order perturbation methods through oversubscription on multiple nodes

The purpose of this work is to evaluate the efficacy of oversubscription, at the 1 n , 2 n , and 3 n levels for n physical cores, on semi‐direct MP2 methods within NWChem when using two and three Intel nodes. Semi‐direct MP2 energy and gradient calculations were performed on chemical systems ranging from 824 to 1626 basis functions using the cc‐pVDZ basis set. Wall times for semi‐direct MP2 energies were reduced by as much as 36% using two nodes and 44% using three nodes compared to no oversubscription. Total energy consumed by the CPU and DRAM was also reduced by as much as 12% using two nodes and as much as 20% using three nodes when oversubscribing. MP2 gradient wall times improved by as much as 16% using two nodes and 18% using three nodes compared to execution at the 1 n level; however, energy savings were insignificant. Intel performance‐counter data show a strong correlation between total wall time saved and less time spent in the idle state, indicating a more efficient use of the processors when oversubscribing. © 2019 Wiley Periodicals, Inc.

Fought, Ellie L.↗

First principles study of the stability and thermal conductivity of novel Li-Be hybrid ceramics

Hybrid Li-Be ceramics, combining both tritium (T) breeder (Li) and neutron multiplier (Be) for use as part of the fuel cycle of future nuclear fusion reactors are proposed. The development of such hybrid materials may reduce thermal gradients through better matching of thermal properties, mitigating the detrimental effects that may accompany traditional breeder systems while maintaining acceptable neutron multiplication and T breeding. Here, first-principles methods are used to investigate stability and thermal transport properties of a set of compounds containing both Li and Be. It is demonstrated that BeLi 2 O 4 Ge and BeLi 2 O 4 Si are mechanically stable and have formation energies comparable with leading candidates for solid state breeder materials, Li 2 TiO 3 , Li 2 ZrO 3 . It is also demonstrated that similar to the leading candidates, these compounds are insulators with thermal transport defined by phonons. The calculated thermal conductivity of BeLi 2 O 4 Ge is slightly higher compared to Li 2 TiO 3 or Li 2 ZrO 3 while in the BeLi 2 O 4 Si it is almost three times higher compared to the rest of compounds due to a higher phonon group velocities and increased phonon lifetimes. These results indicate that hybrid Li-Be ceramics offer a potential route towards better matching of thermal properties with minimal functional property degradation, thereby offering better overall fuel cycle performance.

36 MATERIALS SCIENCE↗

Prevalence of Modifiable Cancer Risk Factors Among U.S. Adults Aged 18–44 Years

Introduction: Carcinogen exposure and unhealthy habits acquired in young adulthood can set the stage for the development of cancer at older ages. This study measured the current prevalence of several cancer risk factors among young adults to assess opportunities to intervene to change the prevalence of these risk factors and potentially reduce cancer incidence. Methods: Using 2015 National Health Interview Survey data (analyzed in 2016), the prevalence of potential cancer risk factors was estimated among U.S. adults aged 18–44 years, based on responses to questions about diet, physical activity, tobacco product use, alcohol, indoor tanning, sleep, human papillomavirus vaccine receipt, and obesity, stratified by sex, age, and race/ethnicity. Results: The prevalence of some risk factors varied by age and race/ethnicity. Obesity (one in four people) and insufficient sleep (one in three people) were common among men and women. Physical inactivity (one in five men, one in four women); binge drinking (one in four men, one in eight women); cigarette smoking (one in five men, one in seven women); and frequent consumption of red meat (one in four men, one in six women) also were common. More than half of the population of adults aged 18–44 years consumed sugar-sweetened beverages daily and processed meat at least once a week. Most young adults had never had the human papillomavirus vaccine. Conclusions: Findings can be used to target evidence-based environmental and policy interventions to reduce the prevalence of cancer risk factors among young adults and prevent the development of future cancers.

60 APPLIED LIFE SCIENCES↗

Scale effects on core design, fuel costs, and spent fuel volume of pressurized water reactors

The desire to improve the economic competitiveness and deployment pace of nuclear energy through modularization, manufacturing, and series production had led to the development of smaller size reactors. As the standard 17x17 fuel technology is mainly maintained in the pressurized water reactors (PWRs) category, this translates into a lower number of fuel assemblies in the core and sometimes a reduced fuel height. To assess the impact of such scale change in core design on fuel cycle cost and spent fuel volume, a scoping analysis tool is developed based on infinite lattice calculations, leakage, fuel management reduced models, and levelized unit cost of electricity (LCOE) estimate. As such, cost dynamics driven by fuel specific power, burnup, core leakage, feed, cycle length, fuel assembly height as well as uranium market data are captured with consistent set of assumptions and analysis methods. A selection of 5 reactor designs representative of leading PWR developers is assessed and compared. Pursuing higher specific powers and optimal burnups are highlighted as the main fuel cost reduction drivers, nevertheless, practical limitations and opportunities must be evaluated to establish the feasibility of such enhanced fuel operation. In consequence, a detailed core design is performed using SIMULATE3 code for 5 PWR variations including natural and forced coolant circulation modes, two reactor scales, power densities of 73, 112, and 123 kW/l and higher discharge burnups. Design and optimization are performed at the lattice level, for the reflector, and at the core loading level. Satisfactory steady-state operation including power distribution, coolant operating limits, and reactivity requirements are analyzed and reported in this paper. The fuel economics of the detailed core designs confirm the scoping analysis findings. Despite the unlocked power uprates in small PWRs, the achievable burnup for a given fuel specific power requires more enrichment and shorter fuel height results in higher fabrication costs per mass of fuel, which makes scaling down core size a more expensive endeavor on the fuel cycle front. Spent fuel volumes are reported for the PWRs designed in this paper. Furthermore, these volumes are driven by the core average discharge burnup regardless of the scale in consideration. Additional cost and core performance aspects related to heavy reflector gains, fuel-reflector substitution, and disposal cost policy in the U.S. are examined.

42 ENGINEERING↗

Overcoming the Memory Bottleneck in Auxiliary Field Quantum Monte Carlo Simulations with Interpolative Separable Density Fitting

We investigate the use of interpolative separable density fitting (ISDF) as a means to reduce the memory bottleneck in auxiliary field quantum Monte Carlo (AFQMC) simulations of real materials in Gaussian basis sets. Here, we find that ISDF can reduce the memory scaling of AFQMC simulations from O(M 4 ) to O(M 2 ) . We test these developments by computing the structural properties of carbon in the diamond phase, comparing to results from existing computational methods and experiment. (LLNL-JRNL-759024)

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Perspectives on Systematic Cloud Microphysics Scheme Development With Machine Learning

Cloud microphysics—the collection of processes that govern the small‐scale formation, evolution, and interactions of liquid droplets and ice crystals in clouds and precipitation—remains a major source of uncertainty in weather and climate models. Although too small in scale to be explicitly resolved in any large‐eddy simulation, weather, or climate model, the representation of cloud microphysical processes has significant impact at the climate scale. Current microphysical schemes are limited by both parametric uncertainty, linked to uncertainty in physical parameter values, and structural uncertainty, arising from incomplete physical understanding of the processes at play or approximations made for computational efficiency. Recent advances in the application of machine learning (ML) to the physical sciences show significant potential for minimizing these limitations by leveraging high‐fidelity simulations and observations. Here we outline the challenges that must be addressed to apply ML toward cloud microphysics scheme development. This perspectives paper synthesizes recent progress in using data‐driven methods, including ML, to improve cloud microphysics parameterizations and highlights opportunities to address key uncertainties. We discuss the roles of aleatoric (irreducible, or statistical) and epistemic (reducible, or systematic) errors in contributing to microphysics parameterization uncertainty. ML can leverage observations to improve microphysical schemes via bottom‐up and top‐down constraints. Methods such as differentiable programming and ML‐enhanced sampling strategies and the creation of large scale benchmark data sets promise to bridge the gap between observations and models and to improve the consistency of cloud microphysical representation across temporal and spatial scales.

Lamb, Kara D. [Columbia Univ., New York, NY (Unite↗

Calculation of excitation energies from the CC2 linear response theory using Cholesky decomposition

A new implementation of the approximate coupled cluster singles and doubles CC2 linear response model is reported. It employs a Cholesky decomposition of the two-electron integrals that significantly reduces the computational cost and the storage requirements of the method compared to standard implementations. Our algorithm also exploits a partitioning form of the CC2 equations which reduces the dimension of the problem and avoids the storage of doubles amplitudes. We present calculation of excitation energies of benzene using a hierarchy of basis sets and compare the results with conventional CC2 calculations. The reduction of the scaling is evaluated as well as the effect of the Cholesky decomposition parameter on the quality of the results. The new algorithm is used to perform an extrapolation to complete basis set investigation on the spectroscopically interesting benzylallene conformers. A set of calculations on medium-sized molecules is carried out to check the dependence of the accuracy of the results on the decomposition thresholds. Moreover, CC2 singlet excitation energies of the free base porphin are also presented.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Gaussian process based optimization of molecular geometries using statistically sampled energy surfaces from quantum Monte Carlo

Optimization of atomic coordinates and lattice parameters remains a significant challenge to the wide use of stochastic electronic structure methods such as quantum Monte Carlo (QMC). Measurements of the forces and stress tensor by these methods contain statistical errors, challenging conventional gradient-based numerical optimization methods that assume deterministic results. Additionally, forces are not yet available for some methods, wavefunctions, and basis sets and when available may be expensive to compute to sufficiently high statistical accuracy near energy minima, where the energy surfaces are flat. Here, we explore the use of Gaussian process based techniques to sample the energy surfaces and reduce sensitivity to the statistical nature of the problem. We utilize Latin hypercube sampling, with the number of sampled energy points scaling quadratically with the number of optimized parameters. Furthermore, we show these techniques may be successfully applied to systems consisting of tens of parameters, demonstrating QMC optimization of a benzene molecule starting from a randomly perturbed, broken symmetry geometry.

74 ATOMIC AND MOLECULAR PHYSICS↗

Filtered region of interest cone-beam rotational angiography

Purpose: Cone-beam rotational angiography (CBRA) is widely used in the modern clinical settings. In a number of procedures, the area of interest is often considerably smaller than the field of view (FOV) of the detector, subjecting the patient to potentially unnecessary x-ray dose. The authors therefore propose a filter-based method to reduce the dose in the regions of low interest, while supplying high image quality in the region of interest (ROI). Methods: For such procedures, the authors propose a method of filtered region of interest (FROI)-CBRA. In the authors' approach, a gadolinium filter with a circular central opening is placed into the x-ray beam during image acquisition. The central region is imaged with high contrast, while peripheral regions are subjected to a substantial lower intensity and dose through beam filtering. The resulting images contain a high contrast/intensity ROI, as well as a low contrast/intensity peripheral region, and a transition region in between. To equalize the two regions' intensities, the first projection of the acquisition is performed with and without the filter in place. The equalization relationship, based on Beer's law, is established through linear regression using corresponding filtered and nonfiltered data. The transition region is equalized based on radial profiles. Results: Evaluations in 2D and 3D show no visible difference between conventional FROI-CBRA projection images and reconstructions in the ROI. CNR evaluations show similar image quality in the ROI, with a reduced CNR in the reconstructed peripheral region. In all filtered projection images, the scatter fraction inside the ROI was reduced. Theoretical and experimental dose evaluations show a considerable dose reduction; using a ROI half the original FOV reduces the dose by 60% for the filter thickness of 1.29 mm. Conclusions: These results indicate the potential of FROI-CBRA to reduce the dose to the patient while supplying the physician with the desired image detail inside the ROI.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Finite element analysis of multilayer coextrusion.

Multilayer coextrusion has become a popular commercial process for producing complex polymeric products from soda bottles to reflective coatings. A numerical model of a multilayer coextrusion process is developed based on a finite element discretization and two different free-surface methods, an arbitrary-Lagrangian-Eulerian (ALE) moving mesh implementation and an Eulerian level set method, to understand the moving boundary problem associated with the polymer-polymer interface. The goal of this work is to have a numerical capability suitable for optimizing and troubleshooting the coextrusion process, circumventing flow instabilities such as ribbing and barring, and reducing variability in layer thickness. Though these instabilities can be both viscous and elastic in nature, for this work a generalized Newtonian description of the fluid is used. Models of varying degrees of complexity are investigated including stability analysis and direct three-dimensional finite element free surface approaches. The results of this work show how critical modeling can be to reduce build test cycles, improve material choices, and guide mold design.

36 MATERIALS SCIENCE↗

Data from Zhang et al. (2019, JAMES, DOI: 10.1029/2019MS001817)

This data set includes the EAM model output used to generate figures presented in the following paper: Zhang, S., Wan, H., Rasch, P. J., Balwinder, S., Larson, V. E., and Woodward, C. S. (2019): An objective and efficient method for assessing the impact of reduced-precision calculations on solution correctness, Journal of Advances in Modeling Earth Systems, https://doi.org/10.1029/2019MS001817

54 ENVIRONMENTAL SCIENCES↗

A metrological large range atomic force microscope with improved performance

A metrological large range atomic force microscope (Met. LR-AFM) has been set up and improved over the past years at Physikalisch-Technische Bundesanstalt (PTB). Being designed as a scanning sample type instrument, the sample is moved in three dimensions by a mechanical ball bearing stage in combination with a compact z-piezostage. Its topography is detected by a position-stationary AFM head. The sample displacement is measured by three embedded miniature homodyne interferometers in the x, y, and z directions. The AFM head is aligned in such a way that its cantilever tip is positioned on the sample surface at the intersection point of the three interferometer measurement beams for satisfying the Abbe measurement principle. In this paper, further improvements of the Met. LR-AFM are reported. A new AFM head using the beam deflection principle has been developed to reduce the influence of parasitic optical interference phenomena. Furthermore, an off-line Heydemann correction method has been applied to reduce the inherent interferometer nonlinearities to less than 0.3 nm (p-v). Versatile scanning functions, for example, radial scanning or local AFM measurement functions, have been implemented to optimize the measurement process. The measurement software is also improved and allows comfortable operations of the instrument via graphical user interface or script-based command sets. The improved Met. LR-AFM is capable of measuring, for instance, the step height, lateral pitch, line width, nanoroughness, and other geometrical parameters of nanostructures. Calibration results of a one-dimensional grating and a set of film thickness standards are demonstrated, showing the excellent metrological performance of the instrument.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Machine-Learning Assisted Identification of Accurate Battery Lifetime Models with Uncertainty

Reduced-order battery lifetime models, which consist of algebraic expressions for various aging modes, are widely utilized for extrapolating degradation trends from accelerated aging tests to real-world aging scenarios. Identifying models with high accuracy and low uncertainty is crucial for ensuring that model extrapolations are believable, however, it is difficult to compose expressions that accurately predict multivariate data trends; a review of cycling degradation models from literature reveals a wide variety of functional relationships. Here, a machine-learning assisted model identification method is utilized to fit degradation in a stand-out LFP-Gr aging data set, with uncertainty quantified by bootstrap resampling. The model identified in this work results in approximately half the mean absolute error of a human expert model. Models are validated by converting to a state-equation form and comparing predictions against cells aging under varying loads. Parameter uncertainty is carried forward into an energy storage system simulation to estimate the impact of aging model uncertainty on system lifetime. The new model identification method used here reduces life-prediction uncertainty by more than a factor of three (86% ± 5% relative capacity at 10 years for human-expert model, 88.5% ± 1.5% for machine-learning assisted model), empowering more confident estimates of energy storage system lifetime.

25 ENERGY STORAGE↗

Coordinated PEV Charging for Distribution System Management

Plug-in electric vehicles (PEVs) can help to reduce worldwide dependence on petroleum and carbon emissions. In the past a few years, PEVs have received considerable attention as an eco-friendly and cost-effective alternative to conventional gasoline vehicles. PEVs consume higher power from the grid during charging compared to conventional residential loads. Therefore, the emerging fleet of PEVs will introduce a considerable amount of additional load on power system. On the other hand, majority of PEVs are parked for more than 90% of the time, making them ideal for providing various services through Electric Vehicle Supply Equipment (EVSE), a.k.a. charging stations which are PEVs’ connection points to the power system. In this project, coordinated PEV charging are developed to provide distribution system services, considering PEV characteristics, realistic travel pattern, charging behaviors, and EVSE power rating and availability. The main efforts and contributions are summarized as follows. • Mobility models and EVSE charging capability/availability are indispensable for evaluating PEV charging coordination strategies and developing PEV charging control. Many existing studies utilize simplified mobility models assuming that the entire vehicle fleet returns home in the evening and is parked at home until the next morning. Some other PEV models better represent diversified home arrival and departure time, but cannot capture varying charging flexibility and capability at different locations, and therefore are not appropriate to study the impacts of public EVSE on PEV load and utilization. In this work, a mobility and charging flexibility model is proposed to better represent the temporal availability and varying charging capability from PEV onboard charger and EVSE. • Based on the mobility model, a generalized optimization method is proposed to evaluate different PEV charging coordination strategies. In existing literature, different algorithms and methods need to be designed to evaluate each charging strategy from both vehicle owners’ and system’s perspective. With the optimization method proposed in this work, different charging control strategies can be studied and compared by only updating objective functions. Moreover, optimization tricks are provided to convert the optimization problems to equivalent linear programming problems, which can be efficiently solved with existing solvers. • An innovative scheduling and control framework is proposed to enable smart PEV charging for grid services while meeting PEV owners’ travel needs. In the proposed framework, each set of EVSE is equipped with a controller that estimates charging power and energy flexibility based on vehicle characteristics, EVSE power rating, battery energy state, and upcoming trip information. With the simplified charging flexibility model received from each EVSE controller, the central coordinator determines the optimal power allocation for a look-ahead time window for the given grid services. The proposed charging control can help reduce the computational complexity and communication requirement compared with existing methods. It is also scalable to the expanding PEV fleet and robust to uncertainties in upcoming vehicle trips and future system condition. • The proposed charging strategy evaluation and smart charging control methods are applied to one of the prototypical feeders developed by the Pacific Northwest National Laboratory (PNNL) for case studies. Detailed trip information extracted from the National Household Travel Survey (NHTS) is used to represent travel and parking patterns. Different scenarios are simulated to evaluate the performance of the proposed methods and understand how PEV mobility and public EVSE availability affect PEV load and potential grid services.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Data Summarization and Inference at Scale

This is the final report for the DOE ASCR grant SC-0022260, Data Summarization and Inference at Scale, PI: Alex Pothen, Purdue University. The goal of the project was to solve data-intensive and compute-intensive problems in the physical sciences, engineering, information science, data science, etc. by designing and implementing new algorithms that could work with a subset of the data. The four subgoals were: (a) The solution of problems where the data is too large to be stored in the memory of a computer. In this streaming model of computation, the data arrives as a stream of elements to the computer, each element is processed as it arrives, and a decision is made to discard the data or to store it; only a small subset of the data proportional to the size of the output solution is stored, and when all the data has been streamed, a solution to the problem is computed from the stored subset. (b) The use of machine learning methods to compute solutions to data-intensive problems. The use of GPUs is critical to obtain high performance on machine learning tasks, but their memory sizes are smaller relative to that of CPUs. For large-scale problems, the data is sampled many times, and small samples are used with repetition, for robustness, to compute solutions to inference tasks. This sampling reduces the memory required to solve the problem, but attention is needed to avoid slow convergence to the solutions, and reduced accuracy of inference. We propose submodular optimization, Large Language Models, and physics-informed neural networks to enable GPU computations here. (c) Modeling and visualization of high-dimensional data using interpretable features. Clinical proteomic data sets from immunology for the detection of cancer and other diseases are temporal and high-dimensional, and algorithms for visualizing these data sets using clinically interpretable features are lacking. We propose methods that compute distances based on the optimal transportation problem and graph edit distances to address this problem. We also propose the use of optimal transport-based distances, spatial statistics, and network structure to classify image data sets, We apply these algorithms to electron micrographs of the peripheral nervous system in the digestive tract. (d) The design of data-intensive algorithms on emerging architectures, specifically, noisy, intermediate-scale quantum (NISQ) devices. Quantum computers offer the possibility of exploring large solution spaces due to the principle of superposition, but current quantum computers are limited by few qubits, short coherence times due to noise, poor interconections among the qubits, etc. We propose the use of the divide and conquer paradigm to solve large-scale problems, wherein collections of small subproblems are solved on the quantum devices, and the solutions to the subproblems are integrated into a solution for the original problem on a classical computer.

97 MATHEMATICS AND COMPUTING↗