Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “computational costs”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Beyond Fisher forecasting for cosmology

The planning and design of future experiments rely heavily on forecasting to assess the potential scientific value provided by a hypothetical set of measurements. The Fisher information matrix, due to its convenient properties and low computational cost, provides an especially useful forecasting tool. However, the Fisher matrix only provides a reasonable approximation to the true likelihood when data are nearly Gaussian distributed and observables have nearly linear dependence on the parameters of interest. Also, Fisher forecasting techniques alone cannot be used to assess their own validity. Thorough sampling of the exact or mock likelihood can definitively determine whether a Fisher forecast is valid, though such sampling is often prohibitively expensive. Here we propose a simple test, based on the Derivative Approximation for likelihoods (DALI) technique, to determine whether the Fisher matrix provides a good approximation to the exact likelihood. We show that the Fisher matrix becomes a poor approximation to the true likelihood in regions where two-dimensional slices of level surfaces of the DALI approximation to the likelihood differ from two-dimensional slices of level surfaces of the Fisher approximation to the likelihood. We demonstrate that our method accurately predicts situations in which the Fisher approximation deviates from the true likelihood for various cosmological models and several data combinations, with only a modest increase in computational cost compared to standard Fisher forecasts.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Robust, Accurate, and Efficient: Quantum Embedding Using the Huzinaga Level-Shift Projection Operator for Complex Systems

Using wave function (WF) in density functional theory (DFT) embedding methods provides a framework for performing localized, high-accuracy WF calculations on a system, while not incurring the full computational cost of the WF calculation on the full system. To effectively partition a system into localized WF and DFT subsystems, we utilize the Huzinaga level-shift projection operator within an absolutely localized basis. In this work, we study the ability of the absolutely localized Huzinaga level-shift projection operator method to study complex WF and DFT partitions, including partitions between multiple covalent bonds, a double bond, and transition-metal–ligand bonds. We find that our methodology can accurately describe all of these complex partitions. Additionally, we study the robustness of this method with respect to the WF method, specifically where the embedded systems were described using a multiconfigurational WF method. We found that the method is systematically improvable with respect to both the number of atoms in the WF region and the size of the basis set used, with energy errors less than 1 kcal/mol. Additionally, we calculated the adsorption energy of H2 to a model of an iron metal–organic framework (Fe-MOF-74) to within 1 kcal/mol compared to CASPT2 calculations performed on the full model while incurring only a small fraction of the full computational cost. In conclusion, this work demonstrates that the absolutely localized Huzinaga level-shift projection operator method is applicable to very complex systems with difficult electronic structures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS2013

Methods and Results of a Search for Gravitational Waves Associated with Gamma-Ray Bursts Using the GEO 600, LIGO, and Virgo Detectors

In this paper we report on a search for short-duration gravitational wave bursts in the frequency range 64 Hz-1792 Hz associated with gamma-ray bursts (GRBs), using data from GEO600 and one of the LIGO or Virgo detectors. We introduce the method of a linear search grid to analyze GRB events with large sky localization uncertainties such as the localizations provided by the Fermi Gamma-ray Burst Monitor (GBM). Coherent searches for gravitational waves (GWs) can be computationally intensive when the GRB sky position is not well-localized, due to the corrections required for the difference in arrival time between detectors. Using a linear search grid we are able to reduce the computational cost of the analysis by a factor of O(10) for GBM events. Furthermore, we demonstrate that our analysis pipeline can improve upon the sky localization of GRBs detected by the GBM, if a high-frequency GW signal is observed in coincidence. We use the linear search grid method in a search for GWs associated with 129 GRBs observed satellite-based gamma-ray experiments between 2006 and 2011. The GRBs in our sample had not been previously analyzed for GW counterparts. A fraction of our GRB events are analyzed using data from GEO600 while the detector was using squeezed-light states to improve its sensitivity; this is the first search for GWs using data from a squeezed-light interferometric observatory. We find no evidence for GW signals, either with any individual GRB in this sample or with the population as a whole. For each GRB we place lower bounds on the distance to the progenitor, assuming a fixed GW emission energy of 10(exp -2)Stellar Mass sq c, with a median exclusion distance of 0.8 Mpc for emission at 500 Hz and 0.3 Mpc at 1 kHz. The reduced computational cost associated with a linear search grid will enable rapid searches for GWs associated with Fermi GBM events in the Advanced detector era.

methods↗
DOE OSTI2023

Thermofield Theory for Finite-Temperature Electronic Structure

Wave function methods have offered a robust, systematically improvable means to study ground-state properties in quantum many-body systems. Theories like coupled cluster and their derivatives provide highly accurate approximations to the energy landscape at a reasonable computational cost. Analogues of such methods to study thermal properties, though highly desirable, have been lacking because evaluating thermal properties involve a trace over the entire Hilbert space, which is a formidable task. Besides, excited-state theories are generally not as well studied as ground-state ones. In this mini-review, we present an overview of a finite-temperature wave function formalism based on thermofield dynamics to overcome these difficulties. Thermofield dynamics allows us to map the equilibrium thermal density matrix to a pure state, i.e., a single wave function, albeit in an expanded Hilbert space. Ensemble averages become expectation values over this so-called thermal state. Around this thermal state, we have developed a procedure to generalize ground-state wave function theories to finite temperatures. As explicit examples, we highlight formulations of mean-field, configuration interaction, and coupled cluster theories for thermal properties of Fermions in the grand-canonical ensemble. To assess the quality of these approximations, we also show benchmark studies for the one-dimensional Hubbard model, while comparing against exact results. We will see that the thermal methods perform similarly to their ground-state counterparts, while merely adding a prefactor to the asymptotic computational cost. Furthermore, they also inherit all the properties, good or bad, from the ground-state methods, signifying the robustness of our formalism and the scope for future development.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Efficient Parameterization of Density Functional Tight-Binding for 5 f -Elements: A Th–O Case Study

Density functional tight binding (DFTB) models for f-element species are challenging to parametrize owing to the large number of adjustable parameters. The explicit optimization of the terms entering the semiempirical DFTB Hamiltonian related to f orbitals is crucial to generating a reliable parametrization for f-block elements, because they play import roles in bonding interactions. However, since the number of parameters grows quadratically with the number of orbitals, the computational cost for parameter optimization is much more expensive for the f-elements than for the main group elements. In this work we present a set of efficient approaches for mitigating the hurdle imposed by the large size of the parameter space. A novel group-by-orbital correction functions for two-center bond integrals was developed. With this approach the number of parameters is reduced, and it grows linearly with the number of elements, maintaining the accuracy and the number of parameters, in the case of f elements, by more than 40%. The parameter optimization step was accelerated by means of the mini-batch BFGS method. This method allows parameter optimizations with much larger training sets than other single batch methods. A stochastic optimizer was employed that helped overcome shallow local minima in the objective function. The proposed algorithm was used to parametrize the DFTB Hamiltonian for the Th–O system, which was subsequently applied to the study of ThO 2 nanoparticles. The training set consisted of 6322 unique structures, which is barely feasible with conventional optimization methods. The optimized parameter set, LANL-ThO, displays good agreement with DFT-calculated properties such as energies, forces, and structures for both clusters and bulk ThO 2 . Benefiting from the fewer number of parameters and lower computational costs for objective function evaluations, this new approach shows its potential applications in DFTB parametrization for elements with high angular momentum, which present a challenge to conventional methods.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

An evaluation of the E3SMv1 Arctic ocean and sea-ice regionally refined model

The Energy Exascale Earth System Model (E3SM) is a state-of-the-science Earth system model (ESM) with the ability to focus horizontal resolution of its multiple components in specific areas. Regionally refined global ESMs are motivated by the need to explicitly resolve, rather than parameterize, relevant physics within the regions of refined resolution, while offering significant computational cost savings relative to the respective cost of configurations with high-resolution (HR) everywhere on the globe. In this paper, we document results from the first Arctic regionally refined E3SM configuration for the ocean and sea-ice components (E3SM-Arctic-OSI), while employing data-based atmosphere, land, and hydrology components. Our aim is an improved representation of the Arctic coupled ocean and sea-ice state, its variability and trends, and the exchanges of mass and property fluxes between the Arctic and the sub-Arctic. We find that E3SM-Arctic-OSI increases the realism of simulated Arctic ocean and sea-ice conditions compared to a similar low-resolution E3SM simulation without the Arctic regional refinement in ocean and sea-ice components (E3SM-LR-OSI). In particular, exchanges through the main Arctic gateways are greatly improved with respect to E3SM-LR-OSI. Other aspects, such as the Arctic freshwater content variability and sea-ice trends, are also satisfactorily simulated. Yet, other features, such as the upper-ocean stratification and the sea-ice thickness distribution, need further improvements, involving either more advanced parameterizations, model tuning, or additional grid refinements. Overall, E3SM-Arctic-OSI offers an improved representation of the Arctic system relative to E3SM-LR-OSI, at a fraction (15 %) of the computational cost of comparable global high-resolution configurations, while permitting exchanges with the lower-latitude oceans that cannot be directly accounted for in Arctic regional models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Automated ICRF heating surrogate modeling via machine learning

This work introduces automated machine learning workflows that address critical bottlenecks in surrogate model development for Ion Cyclotron Range of Frequencies (ICRF) heating applications. The automated framework includes data analysis tools that transform raw datasets into actionable insights in seconds, replacing weeks of manual exploratory effort and ensuring consistent, reproducible dataset characterization. By integrating advanced hyperparameter optimization (HPO) methods including Bayesian optimization via BoTorch and Tree-structured Parzen Estimators (TPE), the framework significantly reduces model development time from weeks to hours, decreasing computational cost and required expertise, while enabling high-accuracy surrogate models. Compared to traditional hyperparameter scanning (HPS) techniques such as methodical, randomized, and grid searches, HPO methods achieve superior convergence and predictive performance, even when compared to already well-tuned reference models. On NSTX High Harmonic Fast Wave (HHFW) heating datasets, both Random Forest Regressor (RFR) and neural network surrogates demonstrate improved accuracy, achieving R 2 values beyond 0.97 and 0.98, respectively. The results show that while HPO gains are modest for robust architectures like RFR, they become essential for more sensitive models such as neural networks, highlighting the trade-offs across optimization strategies. Through automated workflows that eliminate manual hyperparameter tuning and require minimal ML expertise, this work enables widespread adoption of high-fidelity surrogate models across the fusion community for real-time plasma control, uncertainty quantification, rapid experimental scenario development, and integrated system optimization.

Sanchez-Villar, Alvaro [Princeton Plasma Physics L↗
NASA NTRS2012

Extension of the Time-Spectral Approach to Overset Solvers for Arbitrary Motion

Forced periodic flows arise in a broad range of aerodynamic applications such as rotorcraft, turbomachinery, and flapping wing configurations. Standard practice involves solving the unsteady flow equations forward in time until the initial transient exits the domain and a statistically stationary flow is achieved. It is often required to simulate through several periods to remove the initial transient making unsteady design optimization prohibitively expensive for most realistic problems. An effort to reduce the computational cost of these calculations led to the development of the Harmonic Balance method [1, 2] which capitalizes on the periodic nature of the solution. The approach exploits the fact that forced temporally periodic flow, while varying in the time domain, is invariant in the frequency domain. Expanding the temporal variation at each spatial node into a Fourier series transforms the unsteady governing equations into a steady set of equations in integer harmonics that can be tackled with the acceleration techniques afforded to steady-state flow solvers. Other similar approaches, such as the Nonlinear Frequency Domain [3,4,5], Reduced Frequency [6] and Time-Spectral [7, 8, 9] methods, were developed shortly thereafter. Additionally, adjoint-based optimization techniques can be applied [10, 11] as well as frequency-adaptive methods [12, 13, 14] to provide even more flexibility to the method. The Fourier temporal basis functions imply spectral convergence as the number of harmonic modes, and correspondingly number of time samples, N, is increased. Some elect to solve the equations in the frequency domain directly, while others choose to transform the equations back into the time domain to simplify the process of adding this capability to existing solvers, but each harnesses the underlying steady solution in the frequency domain. These temporal projection methods will herein be collectively referred to as Time-Spectral methods. Time-Spectral methods have demonstrated marked success in reducing the computational costs associated with simulating periodic forced flows, but have yet to be fully applied to overset or Cartesian solvers for arbitrary motion with dynamic hole-cutting. Overset and Cartesian grid methodologies are versatile techniques capable of handling complex geometry configurations in practical engineering applications, and the combination of the Time-Spectral approach with this general capability potentially provides an enabling new design and analysis tool. In an arbitrary moving-body scenario for these approaches, a Lagrangian body moves through a fixed Eulerian mesh and mesh points in the Eulerian mesh interior to the solid body are removed (cut or blanked), leaving a hole in the Eulerian mesh. During the dynamic motion some gridpoints in the domain are blanked and do not have a complete set of time-samples preventing a direct implementation of the Time-Spectral method. Murman[6] demonstrated the Time-Spectral approach for a Cartesian solver with a rigid domain motion, wherein the hole cutting remains constant. Similarly, Custer et al. [15, 16] used the NASA overset OVERFLOW solver and limited the amount of relative motion to ensure static hole-cutting and interpolation. Recently, Mavriplis and Mundis[17] demonstrated a qualitative method for applying the Time-Spectral approach to an unstructured overset solver for arbitrary motion. The goal of the current work is to develop a robust and general method for handling arbitrary motion with the Time-Spectral approach within an overset or Cartesian mesh method, while still approaching the spectral convergence rate of the original Time-Spectral approach. The viscous OVERFLOW solver will be augmented with the new Time-Spectral algorithm and the capability of the method for benchmark problems in rotorcraft and turbomachinery will be demonstrated. This abstract begins with a brief synopsis of the Time-Spectral approach for overset grids and provides details of e current approach to allow for arbitrary motion. Model problem results in one and two dimensions are included to demonstrate the viability of the method and the convergence properties. Section IV briefly outlines the implementation into the OVERFLOW solver, and the abstract closes with a description of the benchmark test cases which will be included in the final paper.

Leffell, Joshua Isaac↗
DOE OSTI2020

Assessing Uncertainties and Approximations in Solar Heating of the Climate System

In calculating solar radiation, climate models make many simplifications, in part to reduce computational cost and enable climate modeling, and in part from lack of understanding of critical atmospheric information. Whether known errors or unknown errors, the community's concern is how these could impact the modeled climate. The simplifications are well known and most have published studies evaluating them, but with individual studies it is difficult to compare. Here, we collect a wide range of such simplifications in either radiative transfer modeling or atmospheric conditions and assess potential errors within a consistent framework on climate-relevant scales. We build benchmarking capability around a solar heating code (Solar-J) that doubles as a photolysis code for chemistry and can be readily adapted to consider other errors and uncertainties. The broad classes here include: use of broad wavelength bands to integrate over spectral features; scattering approximations that alter phase function and optical depths for clouds and gases; uncertainty in ice-cloud optics; treatment of fractional cloud cover including overlap; and variability of ocean surface albedo. We geographically map the errors in W m -2 using a full climate re-creation for January 2015 from a weather forecasting model. For many approximations assessed here, mean errors are ~2 W m -2 with greater latitudinal biases and are likely to affect a model’s ability to match the current climate state. Combining this work with previous studies, we make priority recommendations for fixing these simplifications based on both the magnitude of error and the ease or computational cost of the fix.

58 GEOSCIENCES↗
DOE OSTI2025

A rheological model for loose sands with insights from DEM

A rheological model for loose granular media is developed to capture both solid-like and fluid-like responses during shearing. The proposed model is built by following the mathematical structure of an extended Kelvin–Voigt model, where an elastic spring and plastic slider act in parallel to a viscous damper. This arrangement requires the partition of the total stress into rate-independent and rate-dependent stress components. To model the solid-like behavior, a simple frictional plasticity model is adopted without modifications, thus contributing to the rate-independent stress. Instead, the fluid-like or rate-dependent stress is further decomposed into deviatoric and volumetric parts, by proposing a new formulation based on a combination of the μ(I) relation, originally developed under pressure-controlled shear, with a pressure-shear rate relation derived under volume-controlled shear. The proposed formulation allows the model to capture both the increase in the friction coefficient and the enhanced dilation at high shear rates. High-fidelity simulation data, obtained from discrete element method and multiscale modelling, are used to evaluate the performance of the proposed constitutive model. The model provides accurate results under both drained and undrained simple shear paths across a wide range of shear rates. Furthermore, it successfully reproduces at much lower computational cost the flowslide mobility computed through multiscale simulations, which is primarily regulated by the shear rate dependence of the material properties during the dynamic runout stage.

Elasticity↗
DOE OSTI2022

Global Sensitivity Analysis Using the Ultra‐Low Resolution Energy Exascale Earth System Model

Abstract For decades, Arctic temperatures have increased twice as fast as average global temperatures. As a first step toward quantifying parametric uncertainty in Arctic climate, we performed a variance‐based global sensitivity analysis (GSA) using a fully coupled, ultra‐low resolution (ULR) configuration of version 1 of the U.S. Department of Energy's Energy Exascale Earth System Model (E3SMv1). Specifically, we quantified the sensitivity of six quantities of interests (QOIs), which characterize changes in Arctic climate over a 75 year period, to uncertainties in nine model parameters spanning the sea ice, atmosphere, and ocean components of E3SMv1. Sensitivity indices for each QOI were computed with a Gaussian process emulator using 139 random realizations of the random parameters and fixed preindustrial forcing. Uncertainties in the atmospheric parameters in the Cloud Layers Unified by Binormals (CLUBB) scheme were found to have the most impact on sea ice status and the larger Arctic climate. Our results demonstrate the importance of conducting sensitivity analyses with fully coupled climate models. The ULR configuration makes such studies computationally feasible today due to its low computational cost. When advances in computational power and modeling algorithms enable the tractable use of higher‐resolution models, our results will provide a baseline that can quantify the impact of model resolution on the accuracy of sensitivity indices. Moreover, the confidence intervals provided by our study, which we used to quantify the impact of the number of model evaluations on the accuracy of sensitivity estimates, have the potential to inform the computational resources needed for future sensitivity studies.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Programmatic Advantages of Linear Equivalent Seismic Models

Underground explosions nonlinearly deform the surrounding earth material and can interact with the free surface to produce spall. However, at typical seismological observation distances the seismic wavefield can be accurately modeled using linear approximations. Although nonlinear algorithms can accurately simulate very near field ground motions, they are computationally expensive and potentially unnecessary for far field wave simulations. Conversely, linearized seismic wave propagation codes are orders of magnitude faster computationally and can accurately simulate the wavefield out to typical observational distances. Thus, devising a means of approximating a nonlinear source in terms of a linear equivalent source would be advantageous both for scenario modeling and for interpretation of seismic source models that are based on linear, far-field approximations. This allows fast linear seismic modeling that still incorporates many features of the nonlinear source mechanics built into the simulation results so that one can have many of the advantages of both types of simulations without the computational cost of the nonlinear computation. In this report we first show the computational advantage of using linear equivalent models, and then discuss how the near-source (within the nonlinear wavefield regime) environment affects linear source equivalents and how well we can fit seismic wavefields derived from nonlinear sources.

58 GEOSCIENCES↗
NASA NTRS1993

Active control of asymmetric conical flow using spinning and rotatory oscillations

The effectiveness of active control on asymmetric flows around circular cones is investigated computationally using cone spinning and rotatory oscillation around its axis. The investigation uses the time-accurate solution of the unsteady, compressible, full Navier-Stokes equations with the implicit, upwind, flux-difference splitting, finite-volume scheme. The present solutions are obtained under the locally-conical-flow assumption in order to understand the flow physics using very fine grids for reasonable flow resolution at low computational cost. For all the computational solutions, a grid of 241 x 81 x 2 points in the wrap-around, normal and axial directions, respectively, is used. The grid is spinning or oscillating rigidly with the cone according to its motion and the kinematical and dynamical boundary conditions are modified accordingly. The computational applications include the effects of uniform spinning rates and periodic rotatory oscillations at different amplitudes and frequencies on the flow asymmetry.

Kandil, Osama A.↗
NASA NTRS2017

Aeroelastic Wingbox Stiffener Topology Optimization

This work considers an aeroelastic wingbox model seeded with run-out blade stiffeners along the skins. Topology optimization is conducted within the shell webs of the stiffeners, in order to add cutouts and holes for mass reduction. This optimization is done with a global-local approach in order to moderate the computational cost: aeroelastic loads are computed at the wing-level, but the topology and sizing optimization is conducted at the panel-level. Each panel is optimized separately under stress, buckling, and adjacency constraints, and periodically reassembled to update the trimmed aeroelastic loads. The resulting topology is baselined against a design with standard full-depth solid stiffener blades, and found to weigh 7.43% less.

Stanford, Bret K.↗
NASA NTRS1993

Low Cost Spacecraft Computers: Oxymoron or Future Trend?

This paper describes how future trends in commercial computer technology will simplify (or hinder) selection of computer technology for spacecraft control applications, and what spacecraft electronic system designers can do to circumvent design constraints.

spacecraft↗
NASA NTRS2012

AdapChem

AdapChem software enables high efficiency, low computational cost, and enhanced accuracy on computational fluid dynamics (CFD) numerical simulations used for combustion studies. The software dynamically allocates smaller, reduced chemical models instead of the larger, full chemistry models to evolve the calculation while ensuring the same accuracy to be obtained for steady-state CFD reacting flow simulations. The software enables detailed chemical kinetic modeling in combustion CFD simulations. AdapChem adapts the reaction mechanism used in the CFD to the local reaction conditions. Instead of a single, comprehensive reaction mechanism throughout the computation, a dynamic distribution of smaller, reduced models is used to capture accurately the chemical kinetics at a fraction of the cost of the traditional single-mechanism approach.

Oluwole, Oluwayemisi O.↗
DOE OSTI2022

SMALE: Enhancing Scalability of Machine Learning Algorithms on Extreme-Scale Computing Platforms

Deployment and execution of machine learning tasks on extreme-scale computing platforms face several significant technical challenges: 1) High computing cost incurred by dense networks – The computing workload of deep networks with densely-connected topology increases rapidly with the network size, imposing a non-scalable computing model of extreme-scale computing platforms; 2) Non-optimized workload distribution – Many advanced deep learning algorithms, e.g., sparsification and irregular net-work topology, produce very unbalanced workload distribution on extreme-scale computing platforms. The computation efficiency is greatly hindered by the incurred data and computation redundancies as well as long tails of the node with extensive workload; 3) Constraints in data movement and I/O bottle-neck – Inter-node data movement in extreme-scale computing platforms are associated with high energy and latency costs, and subject to the constraints of I/O bandwidth; and 4) Generalization of algorithm realization and acceleration on computing platforms – The large varieties of machine learning algorithms and structures of extreme-scale computing platforms make the derivation of a generalized algorithm realization and acceleration method very challenging, which, however, is the requirement by domain scientists and interested users. We call the above challenges Smale’s Problems in Machine Learning and Understanding for High-Performance Computing Scientific Discovery. The objective of our three-year research project is to develop a holistic innovation set at structure, assembly, and acceleration layers of machine learning algorithms to address the above challenges in algorithm deployment and execution. Three tasks are particularly performed, including: At the algorithm structure level, we investigate the techniques that can structurally sparsify on the topology of deep networks for computing workload reduction. We also study clustering and pruning techniques that can optimize the workload distributions over the extreme-scale computing platforms; At the algorithm assembly level, we derive a unified learning framework for unsupervised transfer learning and dynamic growing capabilities. Novel training methods are also exploited to enhance the training efficiency of the proposed framework; At the algorithm acceleration level, we will develop a series of techniques that can accelerate the computation of sparse matrix operations, which are one of the core executions in deep learning and optimize memory access of the concerned platforms. Our proposed techniques attack the fundamental problems in machine learning algorithms running on extreme-scale computing platforms by vertically integrating the solutions at three closely entangled layers, paving the long-term scaling path of machine learning applications under DOE context. Three tasks corresponding to the above respective research orientations are performed during the three-year project period with our collaborators at ORNL. The outcome of the proposed project is anticipated to form a holistic solution set of novel algorithms and network topologies, efficient training techniques, and fast acceleration methods to promote the computing scalability of the machine learning applications of particular interest to DOE.

97 MATHEMATICS AND COMPUTING↗