Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “sparse data”

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 649 records · Page 36
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

Multithreaded sparse matrix-matrix multiplication for many-core and GPU architectures

Sparse matrix-matrix multiplication is a key kernel that has applications in several domains such as scientific computing and graph analysis. Several algorithms have been studied in the past for this foundational kernel. In this work, we develop parallel algorithms for sparse matrix-matrix multiplication with a focus on performance portability across different high performance computing architectures. The performance of these algorithms depend on the data structures used in them. We compare different types of accumulators in these algorithms and demonstrate the performance difference between these data structures. Furthermore, we develop a meta-algorithm, kkSpGEMM, to choose the right algorithm and data structure based on the characteristics of the problem. We show performance comparisons on three architectures and demonstrate the need for the community to develop two phase sparse matrix-matrix multiplication implementations for efficient reuse of the data structures involved.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Distributed out-of-memory NMF on CPU/GPU architectures

We propose an efficient distributed out-of-memory implementation of the non-negative matrix factorization (NMF) algorithm for heterogeneous high-performance-computing systems. The proposed implementation is based on prior work on NMFk, which can perform automatic model selection and extract latent variables and patterns from data. In this work, we extend NMFk by adding support for dense and sparse matrix operation on multi-node, multi-GPU systems. The resulting algorithm is optimized for out-of-memory problems where the memory required to factorize a given matrix is greater than the available GPU memory. Memory complexity is reduced by batching/tiling strategies, and sparse and dense matrix operations are significantly accelerated with GPU cores (or tensor cores when available). Input/output latency associated with batch copies between host and device is hidden using CUDA streams to overlap data transfers and compute asynchronously, and latency associated with collective communications (both intra-node and inter-node) is reduced using optimized NVIDIA Collective Communication Library (NCCL) based communicators. Benchmark results show significant improvement, from 32X to 76x speedup, with the new implementation using GPUs over the CPU-based NMFk. Good weak scaling was demonstrated on up to 4096 multi-GPU cluster nodes with approximately 25,000 GPUs when decomposing a dense 340 Terabyte-size matrix and an 11 Exabyte-size sparse matrix of density 10 -6 .

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2013

Scoping assessments of ATF impact on late–stage accident progression including molten core-concrete interaction

Simple scoping models that can be used to evaluate ATF performance under severe accident conditions have been developed. The methodology provides a fundamental technical basis (a.k.a. metric) based on the thermodynamic boundary for evaluating performance relative to that of traditional Zr-based claddings. The initial focus in this study was on UO 2 fuel with the advanced claddings 310 SS, D9, FeCrAl, and SiC. The evaluation considered only energy release with concurrent combustible gas production from fuel–cladding–coolant interactions and, separately, molten core–concrete interactions at high temperatures. Other important phenomenological effects that can influence the rate and extent of cladding decomposition (e.g., eutectic interactions, degradation of other core constituents) were not addressed. For the cladding types addressed, potential combustible gas production under both in-vessel and ex-vessel conditions was similar to that for Zr. However, exothermic energy release from cladding oxidation was substantially less for iron-based alloys (by at least a factor of 4), and modestly less (by ~20%) for SiC. Data on SiC-clad UO 2 fuel performance under severe accident conditions are sparse in the literature; thus, assumptions on the nature of the cladding decomposition process were made in order to perform this initial screening evaluation. Furthermore, experimental data for this system under severe accident conditions is needed for a proper evaluation and comparison to iron-based claddings.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2017

Continuously amplified warming in the Alaskan Arctic: Implications for estimating global warming hiatus: SPATIAL COVERAGE AND BIAS IN TREND

Historically, in-situ measurements have been notoriously sparse over the Arctic. As a consequence, the existing gridded data of Surface Air Temperature (SAT) may have large biases in estimating the warming trend in this region. Using data from an expanded monitoring network with 31 stations in the Alaskan Arctic, we demonstrate that the SAT has increased by 2.19 °C in this region, or at a rate of 0.23 °C/decade during 1921-2015. Mean- while, we found that the SAT warmed at 0.71 °C/decade over 1998-2015, which is two to three times faster than the rate established from the gridded datasets. Focusing on the "hiatus" period 1998-2012 as identied by the Intergovernmental Panel on Climate Change (IPCC) report, the SAT has increased at 0.45 °C/decade, which captures more than 90% of the regional trend for 1951- 2012. We suggest that sparse in-situ measurements are responsible for underestimation of the SAT change in the gridded datasets. It is likely that enhanced climate warming may also have happened in the other regions of the Arctic since the late 1990s but left undetected because of incomplete observational coverage.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Distributed-Memory Sparse Deep Neural Network Inference Using Global Arrays

Partitioned Global Address Space (PGAS) models exhibit tremendous promise in developing efficient and productive distributed-memory parallel applications. They have been used extensively in scientific computations due to conveniently offering a ``shared-memory''-like model and convenient interfaces that separate communication with synchronization. Traditionally, PGAS communication models have been applied to dense/contiguously distributed data, but most modern applications depict varied levels of sparsity. Existing PGAS models require certain adaptations to support distributed sparse computations, since associated computations often require matrix arithmetic, in addition to data movement. The Global Arrays toolkit from Pacific Northwest National Laboratory (PNNL) is one of the earliest PGAS models to combine one-sided data communication and distributed matrix operations and is still used in the popular NWChem quantum chemistry suite. Recently, we have expanded the Global Arrays toolkit to support common sparse operations, like sparse matrix-dense matrix multiplies (SpMM), sparse matrix-sparse matrix multiplication (SpGEMM) and Sampled Dense-Dense Matrix Multiplication (SDDMM). As it turns out, these operations are the bedrock of sparse Deep Learning (DL); sparse deep neural networks and Graph Neural Networks (GNNs) have gained increasing attention recently in achieving speedups on training and inference with reduced memory footprints. Unlike scientific applications in High Performance Computing (HPC), modern (distributed-memory capable) DL toolkits often rely on non-standardized and closed-source vendor software optimizations, creating challenges in software-hardware co-design at scale. Our goal is to support a variety of distributed-memory sparse matrix operations and helper functions in the newly created Sparse Global Arrays (SGA), such that it is possible to build portable and productive Machine Learning scenarios for algorithm/software and hardware codesign purposes. Contemporary data-parallel schemes for training/inference are undergoing a major overhaul since model replication limits scalability and causes resource inefficiencies. As such, we have adopted tensor parallelism in decomposing the model and inputs, to mitigate memory issues. Current implementation is built on top of MPI and uses CPUs to maximize the portability across the platforms.

Distributed computing, machine learning↗
DOE OSTI2024

Asymptotic consistency of the WSINDy algorithm in the limit of continuum data

In this work we study the asymptotic consistency of the weak-form sparse identification of nonlinear dynamics algorithm (WSINDy) in the identification of differential equations from noisy samples of solutions. We prove that the WSINDy estimator is unconditionally asymptotically consistent for a wide class of models that includes the Navier–Stokes, Kuramoto–Sivashinsky and Sine–Gordon equations. We thus provide a mathematically rigorous explanation for the observed robustness to noise of weak-form equation learning. Conversely, we also show that, in general, the WSINDy estimator is only conditionally asymptotically consistent, yielding discovery of spurious terms with probability one if the noise level exceeds a critical threshold σ c . We provide explicit bounds on σ c in the case of Gaussian white noise and we explicitly characterize the spurious terms that arise in the case of trigonometric and/or polynomial libraries. Furthermore, we show that, if the data is suitably denoised (a simple moving average filter is sufficient), then asymptotic consistency is recovered for models with locally-Lipschitz, polynomial-growth nonlinearities. Our results reveal important aspects of weak-form equation learning, which may be used to improve future algorithms. We demonstrate our findings numerically using the Lorenz system, the cubic oscillator, a viscous Burgers-growth model and a Kuramoto–Sivashinsky-type high-order PDE.

asymptotic consistency↗
DOE OSTI2023

Feeder Power Disaggregation: A Data-Efficient Matrix Completion Approach: Preprint

This paper presents a data-driven algorithm for the feeder power disaggregation problem in distribution systems. Leveraging spatio-temporal power patterns in residential homes, residential power is discomposed into three components: sparse-switching loads, periodic loads, and photovoltaic generation, using two sparse matrices and a rank-one matrix. The matrix completion process is data-efficient because of the matrix sparsity and low rankness, along with the use of power system models. The proposed approach is tested using real-world residential datasets on a 33-bus distribution system, demonstrating accurate power disaggregation with efficient matrix completion.

distribution system↗
DOE OSTI2020

Three-dimensional fracture continuum characterization aided by surface time-domain electromagnetics and hydrogeophysical joint inversion—proof-of-concept

Efficient and safe production of hydraulically fractured reservoirs benefits from the prediction of their geometrical attributes. Geophysical methods have the potential to provide data that are sensitive to fracture geometries, alleviating the typically sparse nature of in situ reservoir observations. Moreover, surface-based methods can be logistically and economically attractive since they avoid operational interference with the injection well infrastructure. Here, we investigate the potential of the surface-based time-domain electromagnetic (EM) method. EM methods can play an important role owing to their sensitivity to injection-induced fluid property changes. Two other advantageous factors are the EM signal-enhancing effect of vertical steel-cased wells and the fact that injected proppants can be enhanced to produce a stronger electrical conductivity contrast with the reservoir’s connate fluid. Nevertheless, an optimal fracture characterization will no doubt require the integration of EM and reservoir injection and production data. We hence carry out our investigations within a hydrogeophysical parameter estimation framework where EM data and injection flow rates are combined in a fully coupled way. Given the interdisciplinary nature of coupled hydrogeophysical inverse modeling, we dedicate one section to laying out key aspects in a didactic manner.

58 GEOSCIENCES↗
DOE OSTI2022

Preventing Reverse Engineering of Critical Industrial Data with DIOD

Business analytics augmented by artificial intelligence and machine learning (AI/ML) have revolutionized the role of data in the modern world. In recent years, businesses have incorporated data into their decision-making process for better prediction, risk-assessment, content creation, etc. While such businesses often seek to leverage the full use of their data through third-party AI/ML services, they are often hampered by the risks of data leaks, reverse-engineering, stolen technology, etc. that often have disastrous consequences for businesses and their stakeholders alike. Thus, there arises a need for data masking prior to its transmission that obfuscates proprietary information while preserving the information relevant for AI/ML applications. In order to meet the needs of industrial data which are significantly different from those of data warehouses, previous work proposed an efficient time and space-scalable data masking paradigm known as the deceptive infusion of data (DIOD) methodology. The present work expands upon this work by leveraging existing reverse-engineering capabilities to facilitate the decomposition of industrial data into its proprietary and AI/ML-relevant parts, referred to as fundamental and inference metadata respectively. Both sets of metadata are further obfuscated in accordance with the DIOD methodology to create the DIOD rendition of the industrial data, which is rendered immune to reverse-engineering by discarding proprietary information and only preserving AI/ML-relevant information. Additionally, constraints of the original DIOD manuscript are relaxed using mutual information by configuring the methodology to the target AI/ML application to unlock the full potential of the DIOD methodology. As an example, data from a nuclear reactor is transformed into that from a nonlinear spring-mass system with different levels of data masking as required by the generic system and the target application.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

A Novel Sparse Compositional Technique Reveals Microbial Perturbations

ABSTRACT The central aims of many host or environmental microbiome studies are to elucidate factors associated with microbial community compositions and to relate microbial features to outcomes. However, these aims are often complicated by difficulties stemming from high-dimensionality, non-normality, sparsity, and the compositional nature of microbiome data sets. A key tool in microbiome analysis is beta diversity, defined by the distances between microbial samples. Many different distance metrics have been proposed, all with varying discriminatory power on data with differing characteristics. Here, we propose a compositional beta diversity metric rooted in a centered log-ratio transformation and matrix completion called robust Aitchison PCA. We demonstrate the benefits of compositional transformations upstream of beta diversity calculations through simulations. Additionally, we demonstrate improved effect size, classification accuracy, and robustness to sequencing depth over the current methods on several decreased sample subsets of real microbiome data sets. Finally, we highlight the ability of this new beta diversity metric to retain the feature loadings linked to sample ordinations revealing salient intercommunity niche feature importance. IMPORTANCE By accounting for the sparse compositional nature of microbiome data sets, robust Aitchison PCA can yield high discriminatory power and salient feature ranking between microbial niches. The software to perform this analysis is available under an open-source license and can be obtained at https://github.com/biocore/DEICODE ; additionally, a QIIME 2 plugin is provided to perform this analysis at https://library.qiime2.org/plugins/q2-deicode .

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Recursive Dynamic Compressive Sensing in Smart Distribution Systems

With a transition to a smarter grid, we are witnessing a significant growth in smart metering infrastructure and sensor deployment in the distribution system. The underlying communication infrastructure is stressed due to the large volume of data that is generated by the smart meters/sensors. Furthermore, real time operations such as state estimation and control are impaired due to the lack of reliable aggregation of the data. In this paper, we exploit the underlying sparsity in grid data to implement two recursive dynamic compressive sensing (CS) approaches-streaming modified weighted-`1 CS and Kalman filtered CS. These approaches aim to reconstruct the sparse signal using the current underdetermined measurements and the prior information about the sparse signal and its support set. Slow signal and support change is in distribution grid data is validated using Pecan Street data. Both the IEEE 34 node test feeder system and PecanStreet data are considered as two examples to validate the superior performance of the two recursive CS techniques relative to classic CS.

Karimi, Hazhar Sufi↗
DOE OSTI2016

Plug-and-Play Priors for Bright Field Electron Tomography and Sparse Interpolation

Many material and biological samples in scientific imaging are characterized by nonlocal repeating structures. These are studied using scanning electron microscopy and electron tomography. Sparse sampling of individual pixels in a two-dimensional image acquisition geometry, or sparse sampling of projection images with large tilt increments in a tomography experiment, can enable high speed data acquisition and minimize sample damage caused by the electron beam. In this paper, we present an algorithm for electron tomographic reconstruction and sparse image interpolation that exploits the nonlocal redundancy in images. We adapt a framework, termed plug-and-play priors, to solve these imaging problems in a regularized inversion setting. The power of the plug-and-play approach is that it allows a wide array of modern denoising algorithms to be used as a “prior model” for tomography and image interpolation. We also present sufficient mathematical conditions that ensure convergence of the plug-and-play approach, and we use these insights to design a new nonlocal means denoising algorithm. Finally, we demonstrate that the algorithm produces higher quality reconstructions on both simulated and real electron microscope data, along with improved convergence properties compared to other methods.

47 OTHER INSTRUMENTATION↗
DOE OSTI2015

A new protocol to accurately determine microtubule lattice seam location

Microtubules (MTs) are cylindrical polymers of αβ-tubulin that display pseudo-helical symmetry due to the presence of a lattice seam of heterologous lateral contacts. The structural similarity between α- and β-tubulin makes it difficult to computationally distinguish them in the noisy cryo-EM images, unless a marker protein for the tubulin dimer, such as kinesin motor domain, is present. We have developed a new data processing protocol that can accurately determine αβ-tubulin register and seam location for MT segments. Our strategy can handle difficult situations, where the marker protein is relatively small or the decoration of marker protein is sparse. Using this new seam-search protocol, combined with movie processing for data from a direct electron detection camera, we were able to determine the cryo-EM structures of MT at 3.5. Å resolution in different functional states. The successful distinction of α- and β-tubulin allowed us to visualize the nucleotide state at the E-site and the configuration of lateral contacts at the seam.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2016

MO-DE-207A-11: Sparse-View CT Reconstruction Via a Novel Non-Local Means Method

Purpose: Sparse-view computed tomography (CT) reconstruction is an effective strategy to reduce the radiation dose delivered to patients. Due to its insufficiency of measurements, traditional non-local means (NLM) based reconstruction methods often lead to over-smoothness in image edges. To address this problem, an adaptive NLM reconstruction method based on rotational invariance (RIANLM) is proposed. Methods: The method consists of four steps: 1) Initializing parameters; 2) Algebraic reconstruction technique (ART) reconstruction using raw projection data; 3) Positivity constraint of the image reconstructed by ART; 4) Update reconstructed image by using RIANLM filtering. In RIANLM, a novel similarity metric that is rotational invariance is proposed and used to calculate the distance between two patches. In this way, any patch with similar structure but different orientation to the reference patch would win a relatively large weight to avoid over-smoothed image. Moreover, the parameter h in RIANLM which controls the decay of the weights is adaptive to avoid over-smoothness, while it in NLM is not adaptive during the whole reconstruction process. The proposed method is named as ART-RIANLM and validated on Shepp-Logan phantom and clinical projection data. Results: In our experiments, the searching neighborhood size is set to 15 by 15 and the similarity window is set to 3 by 3. For the simulated case with a resolution of 256 by 256 Shepp-Logan phantom, the ART-RIANLM produces higher SNR (35.38dB<24.00dB) and lower MAE (0.0006<0.0023) reconstructed image than ART-NLM. The visual inspection demonstrated that the proposed method could suppress artifacts or noises more effectively and preserve image edges better. Similar results were found for clinical data case. Conclusion: A novel ART-RIANLM method for sparse-view CT reconstruction is presented with superior image. Compared to the conventional ART-NLM method, the SNR and MAE from ART-RIANLM increases 47% and decreases 74%, respectively.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2015

WE-EF-207-07: Dual Energy CT with One Full Scan and a Second Sparse-View Scan Using Structure Preserving Iterative Reconstruction (SPIR)

Purpose: Conventional dual energy CT (DECT) reconstructs CT and basis material images from two full-size projection datasets with different energy spectra. To relax the data requirement, we propose an iterative DECT reconstruction algorithm using one full scan and a second sparse-view scan by utilizing redundant structural information of the same object acquired at two different energies. Methods: We first reconstruct a full-scan CT image using filtered-backprojection (FBP) algorithm. The material similarities of each pixel with other pixels are calculated by an exponential function about pixel value differences. We assume that the material similarities of pixels remains in the second CT scan, although pixel values may vary. An iterative method is designed to reconstruct the second CT image from reduced projections. Under the data fidelity constraint, the algorithm minimizes the L2 norm of the difference between pixel value and its estimation, which is the average of other pixel values weighted by their similarities. The proposed algorithm, referred to as structure preserving iterative reconstruction (SPIR), is evaluated on physical phantoms. Results: On the Catphan600 phantom, SPIR-based DECT method with a second 10-view scan reduces the noise standard deviation of a full-scan FBP CT reconstruction by a factor of 4 with well-maintained spatial resolution, while iterative reconstruction using total-variation regularization (TVR) degrades the spatial resolution at the same noise level. The proposed method achieves less than 1% measurement difference on electron density map compared with the conventional two-full-scan DECT. On an anthropomorphic pediatric phantom, our method successfully reconstructs the complicated vertebra structures and decomposes bone and soft tissue. Conclusion: We develop an effective method to reduce the number of views and therefore data acquisition in DECT. We show that SPIR-based DECT using one full scan and a second 10-view scan can provide high-quality DECT images and accurate electron density maps as conventional two-full-scan DECT.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Deep learning-accelerated 3D carbon storage reservoir pressure forecasting based on data assimilation using surface displacement from InSAR

Fast forecasting of the reservoir pressure distribution during geologic carbon storage (GCS) by assimilating monitoring data is a challenging problem. Due to high drilling cost, GCS projects usually have spatially sparse measurements from few wells, leading to high uncertainties in reservoir pressure prediction. To address this challenge, we use low-cost Interferometric Synthetic-Aperture Radar (InSAR) data as monitoring data to infer reservoir pressure build up. We develop a deep learning-accelerated workflow to assimilate surface displacement maps interpreted from InSAR and to forecast dynamic reservoir pressure. Employing an Ensemble Smoother Multiple Data Assimilation (ES-MDA) framework, the workflow updates three-dimensional (3D) geologic properties and predicts reservoir pressure with quantified uncertainties. We use a synthetic commercial-scale GCS model with bimodally distributed permeability and porosity to demonstrate the efficacy of the workflow. A two-step CNN-PCA approach is employed to parameterize the bimodal fields. The computational efficiency of the workflow is boosted by two residual U-Net based surrogate models for surface displacement and reservoir pressure predictions, respectively. The workflow can complete data assimilation and reservoir pressure forecasting in half an hour on a personal computer.

25 ENERGY STORAGE↗
DOE OSTI2025

sparse_bias

This is a python package used to fit an unknown function from data that potential contains systematic biases related to metadata. The model fits the unknown function and uses a Bayesian horseshoe prior model to impose sparsity on the bias terms. This code has been generalized from research code developed for AIACHNE into a package that should have more general application in a wider class of statistical models.

Walton, Noah↗
DOE OSTI2018

Multi-threaded Sparse Matrix Sparse Matrix Multiplication for Many-Core and GPU Architectures.

Sparse Matrix-Matrix multiplication is a key kernel that has applications in several domains such as scientific computing and graph analysis. Several algorithms have been studied in the past for this foundational kernel. In this paper, we develop parallel algorithms for sparse matrix- matrix multiplication with a focus on performance portability across different high performance computing architectures. The performance of these algorithms depend on the data structures used in them. We compare different types of accumulators in these algorithms and demonstrate the performance difference between these data structures. Furthermore, we develop a meta-algorithm, kkSpGEMM, to choose the right algorithm and data structure based on the characteristics of the problem. We show performance comparisons on three architectures and demonstrate the need for the community to develop two phase sparse matrix-matrix multiplication implementations for efficient reuse of the data structures involved.

97 MATHEMATICS AND COMPUTING↗