Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Network”

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 109 records · Page 6

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↗

FES Science Network Requirements - Report of the Fusion Energy Sciences Network Requirements Workshop Conducted March 13 and 14, 2008

The Energy Sciences Network (ESnet) is the primary provider of network connectivity for the U.S. Department of Energy Office of Science, the single largest supporter of basic research in the physical sciences in the United States of America. In support of the Office of Science programs, ESnet regularly updates and refreshes its understanding of the networking requirements of the instruments, facilities, scientists, and science programs that it serves. This focus has helped ESnet to be a highly successful enabler of scientific discovery for over 20 years. In March 2008, ESnet and the Fusion Energy Sciences (FES) Program Office of the DOE Office of Science organized a workshop to characterize the networking requirements of the science programs funded by the FES Program Office. Most sites that conduct data-intensive activities (the Tokamaks at GA and MIT, the supercomputer centers at NERSC and ORNL) show a need for on the order of 10 Gbps of network bandwidth for FES-related work within 5 years. PPPL reported a need for 8 times that (80 Gbps) in that time frame. Estimates for the 5-10 year time period are up to 160 Mbps for large simulations. Bandwidth requirements for ITER range from 10 to 80 Gbps. In terms of science process and collaboration structure, it is clear that the proposed Fusion Simulation Project (FSP) has the potential to significantly impact the data movement patterns and therefore the network requirements for U.S. fusion science. As the FSP is defined over the next two years, these changes will become clearer. Also, there is a clear and present unmet need for better network connectivity between U.S. FES sites and two Asian fusion experiments--the EAST Tokamak in China and the KSTAR Tokamak in South Korea. In addition to achieving its goal of collecting and characterizing the network requirements of the science endeavors funded by the FES Program Office, the workshop emphasized that there is a need for research into better ways of conducting remote collaboration with the control room of a Tokamak running an experiment. This is especially important since the current plans for ITER assume that this problem will be solved.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Learning generative neural networks with physics knowledge

Deep generative neural networks have enabled modeling complex distributions, but incorporating physics knowledge into the neural networks is still challenging and is at the core of current physics-based machine learning research. To this end, we propose a physics generative neural network (PhysGNN), a new class of generative neural networks for learning unknown distributions in a physical system described by partial differential equations (PDE). PhysGNN couples PDE systems with generative neural networks. It is a fully differentiable model that allows back-propagation of gradients through both numerical PDE solvers and generative neural networks, and is trained by minimizing the discrete Wasserstein distance between generated and observed probability distributions of the PDE outputs using the stochastic gradient descent method. Moreover, PhysGNN does not require adversarial training like standard generative neural networks, which offers better stability than adversarial training. We show that PhysGNN can learn complex distributions in stochastic inverse problems, where conventional methods such as maximum likelihood estimation and momentum matching methods may be inapplicable when little knowledge is known about the form of unknown distributions or the physical model is too complex. Furthermore, our method allows physics-based generative neural network training for learning complex distributions in the context of differential equations.

97 MATHEMATICS AND COMPUTING↗

Accelerating network layouts using graph neural networks

Graph layout algorithms used in network visualization represent the first and the most widely used tool to unveil the inner structure and the behavior of complex networks. Current network visualization software relies on the force-directed layout (FDL) algorithm, whose high computational complexity makes the visualization of large real networks computationally prohibitive and traps large graphs into high energy configurations, resulting in hard-to-interpret “hairball” layouts. Here we use Graph Neural Networks (GNN) to accelerate FDL, showing that deep learning can address both limitations of FDL: it offers a 10 to 100 fold improvement in speed while also yielding layouts which are more informative. We analytically derive the speedup offered by GNN, relating it to the number of outliers in the eigenspectrum of the adjacency matrix, predicting that GNNs are particularly effective for networks with communities and local regularities. Finally, we use GNN to generate a three-dimensional layout of the Internet, and introduce additional measures to assess the layout quality and its interpretability, exploring the algorithm’s ability to separate communities and the link-length distribution. The novel use of deep neural networks can help accelerate other network-based optimization problems as well, with applications from reaction-diffusion systems to epidemics.

97 MATHEMATICS AND COMPUTING↗

High Energy Physics Network Requirements Review: Final Report, July 2024–December 2024

The world-class research infrastructure at the US Department of Energy (DOE) Office of Science (SC) provides the research community with premier observational, experimental, computational, and network capabilities. Each user facility is designed to provide unique capabilities to advance the core DOE mission in science and technology for its SC program to stimulate rich scientific discoveries and enhance its innovation ecosystem. Research communities gather and flourish around each user facility, bringing together new and enhanced perspectives. The continual reinvention of the practice of science — as users and staff forge novel approaches expressed in research workflows — unlocks new discoveries and propels scientific progress. Within this research ecosystem, the high-performance computing (HPC) and networking user facilities stewarded by the SC’s Advanced Scientific Computing Research (ASCR) program play a dynamic cross-cutting role, enabling complex workflows demanding high-performance data, networking, and computing solutions. The ASCR facilities enterprise seeks to understand and meet the needs and requirements across SC and DOE domain science programs and priority efforts, highlighted by the formal requirements review methodology. Between July and December 2024, the Energy Sciences Network (ESnet) and the Office of High Energy Physics (HEP) of the DOE-SC organized an ESnet requirements review of HEP-supported program activities. Preparation for these events included identification of key stakeholders: program and facility management, research groups, and technology providers. Each stakeholder group was asked to prepare formal case study documents about its relationship to the HEP program to build a complete understanding of the current, near-term, and long-term status, expectations, and processes that will support the science going forward.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Multi-Omics Driven Metabolic Network Reconstruction and Analysis of Lignocellulosic Carbon Utilization in Rhodosporidium toruloides

An oleaginous yeast Rhodosporidium toruloides is a promising host for converting lignocellulosic biomass to bioproducts and biofuels. In this work, we performed multi-omics analysis of lignocellulosic carbon utilization in R. toruloides and reconstructed the genome-scale metabolic network of R. toruloides . High-quality metabolic network models for model organisms and orthologous protein mapping were used to build a draft metabolic network reconstruction. The reconstruction was manually curated to build a metabolic model using functional annotation and multi-omics data including transcriptomics, proteomics, metabolomics, and RB-TDNA sequencing. The multi-omics data and metabolic model were used to investigate R. toruloides metabolism including lipid accumulation and lignocellulosic carbon utilization. The developed metabolic model was validated against high-throughput growth phenotyping and gene fitness data, and further refined to resolve the inconsistencies between prediction and data. We believe that this is the most complete and accurate metabolic network model available for R. toruloides to date.

09 BIOMASS FUELS↗

Network coding for improving throughput in WDM optical networks with dedicated protection

Network coding (NC) has been recently emerging as an excellent technique for leveraging network efficiency. In optical networks, NC and protection appear to be a good match and indeed, NC could be effectively exploited for greater protection efficiency thanks to the opportunity of combining signals rather than duplication in the traditional approach. Motivated by this observation, we investigate the application of a simple and practical NC technique based on XOR operation to improve the network throughput in survivable optical networks. To this end, we propose the NC-based protection whose speed recovery is on a par with the conventional 1+1 dedicated protection while the capacity usage could be more efficient, paving the way for improving throughput in constrained capacity conditions. A novel mixed integer linear programming model is developed for survivable optical networks design employing the proposed NC-based protection. The model attempts to optimally determine the demands to serve, the routing for served demands and NC assignment among demands aiming at maximizing the network throughput in the constraint of bandwidth limitation. Extensive numerical evaluations on realistic topologies are provided to highlight the advantages of our NC-based protection proposal compared with the conventional dedicated protection in improving the network throughput. Encouraging results are observed when the new protection proposal based on NC could accommodate up to 30% more traffic than the conventional case.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Synthesis and Structural Characterization of Magnesium Based Coordination Networks in Different Solvents

Three magnesium based metal-organic frameworks, Mg{sub 3}(3,5-PDC){sub 3}(DMF){sub 3} {center_dot} DMF [1], Mg(3,5-PDC)(H{sub 2}O) {center_dot} (H{sub 2}O) [3], and Mg4(3,5-PDC)4(DMF){sub 2}(H{sub 2}O){sub 2} {center_dot} 2DMF {center_dot} 4.5H{sub 2}O [4], and a 2-D coordination polymer, [Mg(3,5-PDC)(H{sub 2}O){sub 2}] [2] [PDC = pyridinedicarboxylate], were synthesized using a combination of DMF, methanol, ethanol, and water. Compound 1 [space group P2{sub 1}/n, a = 12.3475(5) {angstrom}, b = 11.1929(5) {angstrom}, c = 28.6734(12) {angstrom}, {beta} = 98.8160(10){sup o}, V = 3916.0(3) {angstrom}{sup 3}] consists of a combination of isolated and corner-sharing magnesium octahedra connected by the organic linkers to form a 3-D network with a 12.2 {angstrom} x 4.6 {angstrom} 1-D channel. The channel contains coordinated and free DMF molecules. In compound 2 [space group C2/c, a = 9.964(5) {angstrom}, b = 12.0694(6) {angstrom}, c = 7.2763(4) {angstrom}, {beta} = 106.4970(6){sup o}, V = 836.70(6) {angstrom}{sup 3}], PDC connects isolated seven coordinated magnesium polyhedra into a layered structure. Compound 3 [space group P6{sub 1}22, a = 11.479(1) {angstrom}, c = 14.735(3) {angstrom}, V = 1681.7(4) {angstrom}{sup 3}] (previously reported) contains isolated magnesium octahedra connected by the organic linker with each other forming a 3D network. Compound 4 [space group P2{sub 1}/c, a = 13.7442(14) {angstrom}, b = 14.2887(15) {angstrom}, c = 14.1178(14) {angstrom}, {beta} = 104.912(2){sup o}, V = 2679.2(5) {angstrom}{sup 3}] also exhibits a 3D network based on isolated magnesium octahedra with square cavities containing both disordered DMF and water molecules. The structural topologies originate due to the variable coordination ability of solvent molecules with the metal center. Water molecules coordinate with the magnesium metal centers preferably over other polar solvents (DMF, methanol, ethanol) used to synthesize the coordination networks. Despite testing multiple desolvation routes, we were unable to measure BET surface areas greater than 51.9 m{sup 2}/g for compound 1.

Banerjee, Debasis↗

Identifying gaps in grid middleware on fast networks with the advanced networking initiative

As of 2012, a number of US Department of Energy (DOE) National Laboratories have access to a 100 Gb/s wide-area network backbone. The ESnet Advanced Networking Initiative (ANI) project is intended to develop a prototype network, based on emerging 100 Gb/s Ethernet technology. The ANI network will support DOE's science research programs. A 100 Gb/s network test bed is a key component of the ANI project. The test bed offers the opportunity for early evaluation of 100Gb/s network infrastructure for supporting the high impact data movement typical of science collaborations and experiments. In order to make effective use of this advanced infrastructure, the applications and middleware currently used by the distributed computing systems of large-scale science need to be adapted and tested within the new environment, with gaps in functionality identified and corrected. As a user of the ANI test bed, Fermilab aims to study the issues related to end-to-end integration and use of 100 Gb/s networks for the event simulation and analysis applications of physics experiments. In this paper we discuss our findings from evaluating existing HEP Physics middleware and application components, including GridFTP, Globus Online, etc. in the high-speed environment. These will include possible recommendations to the system administrators, application and middleware developers on changes that would make production use of the 100 Gb/s networks, including data storage, caching and wide area access.

Dykstra, Dave↗

Context-Enriched Identification of Particles with a Convolutional Network for Neutrino Events

Particle detectors record the interactions of subatomic particles and their passage through matter. The identification of these particles is necessary for in-depth physics analysis. While particles can be identified by their individual behavior as they travel through matter, the full context of the interaction in which they are produced can aid the classification task substantially. We have developed the first convolutional neural network for particle identification which uses context information. This is also the first implementation of a four-tower siamese-type architecture both for separation of independent inputs and inclusion of context information. The network classifies clusters of energy deposits from the NOvA neutrino detectors as electrons, muons, photons, pions, and protons with an overall efficiency and purity of 83.3% and 83.5%, respectively. We show that providing the network with context information improves performance by comparing our results with a network trained without context information.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Networked HIL Simulation System for Modeling Large-scale Power Systems

This paper presents a networked hardware-in-the-loop (HIL) simulation system for modeling large-scale power systems. Researchers have developed many HIL test systems for power systems in recent years. Those test systems can model both microsecond-level dynamic responses of power electronic systems and millisecond-level transients of transmission and distribution grids. By integrating individual HIL test systems into a network of HIL test systems, we can create large-scale power grid digital twins with flexible structures at required modeling resolution that fits for a wide range of system operating conditions. This will not only significantly reduce the need for field tests when developing new technologies but also greatly shorten the model development cycle. In this paper, we present a networked OPAL-RT based HIL test system for developing transmission-distribution coordinative Volt-VAr regulation technologies as an example to illustrate system setups, communication requirements among different HIL simulation systems, and system connection mechanisms. Impacts of communication delays, information exchange cycles, and computing delays are illustrated. Simulation results show that the performance of a networked HIL test system is satisfactory.

Xie, Fuhong↗

DEFINING THE PLAYERS IN HIGHER-ORDER NETWORKS: PREDICTIVE MODELING FOR REVERSE ENGINEERING FUNCTIONAL INFLUENCE NETWORKS

A difficult problem that is currently growing rapidly due to the sharp increase in the amount of high-throughput data available for many systems is that of determining useful and informative causative influence networks. These networks can be used to predict behavior given observation of a small number of components, predict behavior at a future time point, or identify components that are critical to the functioning of the system under particular conditions. In these endeavors incorporating observations of systems from a wide variety of viewpoints can be particularly beneficial, but has often been undertaken with the objective of inferring networks that are generally applicable. The focus of the current work is to integrate both general observations and measurements taken for a particular pathology, that of ischemic stroke, to provide improved ability to produce useful predictions of systems behavior. A number of hybrid approaches have recently been proposed for network generation in which the Gene Ontology is used to filter or enrich network links inferred from gene expression data through reverse engineering methods. These approaches have been shown to improve the biological plausibility of the inferred relationships determined, but still treat knowledge-based and machine-learning inferences as incommensurable inputs. In this paper, we explore how further improvements may be achieved through a full integration of network inference insights achieved through application of the Gene Ontology and reverse engineering methods with specific reference to the construction of dynamic models of transcriptional regulatory networks. We show that integrating two approaches to network construction, one based on reverse-engineering from conditional transcriptional data, one based on reverse-engineering from in situ hybridization data, and another based on functional associations derived from Gene Ontology, using probabilities can improve results of clustering as evaluated by a predictive model of transcriptional expression levels.

59 BASIC BIOLOGICAL SCIENCES↗

Evolving Deep Networks Using HPC

While a large number of deep learning networks have been studiedand published that produce outstanding results on natural imagedatasets, these datasets only make up a fraction of those to whichdeep learning can be applied. These datasets include text data, audiodata, and arrays of sensors that have very different characteristicsthan natural images. As these “best” networks for natural imageshave been largely discovered through experimentation and cannotbe proven optimal on some theoretical basis, there is no reasonto believe that they are the optimal network for these drasticallydifferent datasets. Hyperparameter search is thus often a very im-portant process when applying deep learning to a new problem. Inthis work we present an evolutionary approach to searching thepossible space of network hyperparameters and construction thatcan scale to 18, 000 nodes. This approach is applied to datasets ofvarying types and characteristics where we demonstrate the abilityto rapidly find best hyperparameters in order to enable practitionersto quickly iterate between idea and result.

Young, Steven↗

A Comparison of Geographic Information Systems, Complex Networks, and Other Models for Analyzing Transportation Network Topologies

This report reviews six classes of models that are used for studying transportation network topologies. The report is motivated by two main questions. First, what can the "new science" of complex networks (scale-free, small-world networks) contribute to our understanding of transport network structure, compared to more traditional methods? Second, how can geographic information systems (GIS) contribute to studying transport networks? The report defines terms that can be used to classify different kinds of models by their function, composition, mechanism, spatial and temporal dimensions, certainty, linearity, and resolution. Six broad classes of models for analyzing transport network topologies are then explored: GIS; static graph theory; complex networks; mathematical programming; simulation; and agent-based modeling. Each class of models is defined and classified according to the attributes introduced earlier. The paper identifies some typical types of research questions about network structure that have been addressed by each class of model in the literature.

Alexandrov, Natalia↗

Rates and Reversibilities in Interconnected Reaction Networks

Mathematical relations prescribing unidirectional forward and reverse rates originally derived based on single-path reaction sequences do not apply to interconnected reaction networks. The presence of branches in reaction networks, leading to alternative stable products, decreases unidirectional rates in reference to those calculated by single-path functional forms, as shown by simulated isotopic exchange rates, but impacts the unidirectional forward and reverse rates equally such that the functional form of effective reversibility remains unchanged. Regardless of stoichiometric numbers and network connectivity, the application of the pseudo-steady-state hypothesis on reactive intermediates in conjunction with consideration of unidirectional rates toward the product of interest and all alternative stable products results in mathematical expressions that accurately reflect simulated isotopic exchange rates. Further analyses based on kinetic resistance, a property akin to electrical resistance, illustrate the manifestation of nodal resistances in addition to the single-path kinetic resistance for interconnected reaction networks. The generalized formalism for assessing rates and reversibilities in interconnected networks derived herein enables us to demonstrate that unidirectional rates cannot be assessed solely from effective reversibilities and net rates of generation of stable species in such networks. Furthermore, isotopic exchange rates, under the condition that each elementary step in the overall reaction sequence forms a unique reactive intermediate that is consumed solely by the subsequent step, can be utilized to determine unidirectional rates and can serve to validate postulated reaction pathways in highly interconnected reaction networks (e.g., CO x hydrogenation).

42 ENGINEERING↗

Hierarchical Network Partitioning for Solution of Potential-Driven, Steady-State Nonlinear Network Flow Equations

The solution of potential-driven steady-state flow in large networks is a task which manifests in various engineering applications, such as transport of natural gas or water through pipeline networks. The resultant system of nonlinear equations depends on the network topology, and in general, there is no numerical algorithm that offers guaranteed convergence to the solution (assuming a solution exists). Some methods offer guarantees in cases where the network topology satisfies certain assumptions, but these methods fail for larger networks. On the other hand, the Newton-Raphson algorithm offers a convergence guarantee if the starting point lies close to the (unknown) solution. It would be advantageous to compute the solution of the large nonlinear system through the solution of smaller nonlinear sub-systems wherein the solution algorithms (Newton-Raphson or otherwise) are more likely to succeed. Here, this letter proposes and describes such a procedure, a hierarchical network partitioning algorithm that enables the solution of large nonlinear systems corresponding to potential-driven steady-state network flow equations.

42 ENGINEERING↗

xGFabric: Coupling Sensor Networks and HPC Facilities with Private 5G Wireless Networks for Real-Time Digital Agriculture

Advanced scientific applications require coupling distributed sensor networks with centralized high-performance computing facilities. Citrus Under Protective Screening (CUPS) exemplifies this need in digital agriculture, where citrus research facilities are instrumented with numerous sensors monitoring environmental conditions and detecting protective screening damage. CUPS demands access to computational fluid dynamics codes for modeling environmental conditions and guiding real-time interventions like water application or robotic repairs. These computing domains have contrasting properties: sensor networks provide low-performance, limited-capacity, unreliable data access, while high-performance facilities offer enormous computing power through high-latency batch processing. Private 5G networks present novel capabilities addressing this challenge by providing low latency, high throughput, and reliability necessary for near-real-time coupling of edge sensor networks with HPC simulations. This work presents xGFabric, an end-to-end system coupling sensor networks with HPC facilities through Private 5G networks. The prototype connects remote sensors via 5G network slicing to HPC systems, enabling real-time digital agriculture simulation.

Digital Agriculture↗