Search NASA⌕ Search

SEARCH · Search NASA

Results for “Ensemble computing.”

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 91 records · Page 5

GMRES with embedded ensemble propagation for the efficient solution of parametric linear systems in uncertainty quantification of computational models

In a previous work, embedded ensemble propagation was proposed to improve the efficiency of sampling-based uncertainty quantification methods of computational models on emerging computational architectures. It consists of simultaneously evaluating the model for a subset of samples together, instead of evaluating them individually. A first approach introduced to solve parametric linear systems with ensemble propagation is ensemble reduction. In Krylov methods for example, this reduction consists in coupling the samples together using an inner product that sums the sample contributions. Ensemble reduction has the advantages of being able to use optimized implementations of BLAS functions and having a stopping criterion which involves only one scalar. However, the reduction potentially decreases the rate of convergence due to the gathering of the spectra of the samples. In this paper, we investigate a second approach: ensemble propagation without ensemble reduction in the case of GMRES. This second approach solves each sample simultaneously but independently to improve the convergence compared to ensemble reduction. This raises two new issues which are solved in this paper: the fact that optimized implementations of BLAS functions cannot be used anymore and that ensemble divergence, whereby individual samples within an ensemble must follow different code execution paths, can occur. We tackle those issues by implementing a high-performing ensemble GEMV and by using masks. The proposed ensemble GEMV leads to a similar cost per GMRES iteration for both approaches, i.e. with and without reduction. For illustration, we study the performances of the new linear solver in the context of a mesh tying problem. Furthermore, this example demonstrates improved ensemble propagation speed-up without reduction.

BLAS↗

State-to-state, multi-collision, energy transfer in H–H{sub 2} gas ensembles

We use our recently developed computational model of energy flow in gas ensembles to study translation-to-internal energy conversion in an ensemble consisting of H{sub 2}(0; 0) in a bath of H atoms. This mixture is found in plasmas of industrial importance and also in interstellar clouds. The storage of energy of relative motion as rovibrational energy of H{sub 2} represents a potential mechanism for cooling translation. This may have relevance in astrophysical contexts such as the post-recombination epoch of the early universe when hydrogenic species dominated and cooling was a precondition for the formation of structured objects. We find that conversion of translational motion to H{sub 2} vibration and rotation is fast and, in our closed system, is complete within around 100 cycles of ensemble collisions. Large amounts of energy become stored as H{sub 2} vibration and a tentative mechanism for this unequal energy distribution is suggested. The “structured dis-equilibrium” we observe is found to persist through many collision cycles. In contrast to the rapidity of excitation, the relaxation of H{sub 2}(6; 10) in H is very slow and not complete after 10{sup 5} collision cycles. The quasi-equilibrium modal temperatures of translation, rotation, and vibration are found to scale linearly with collision energy but at different rates. This may be useful in estimating the partitioning of energy within a given H + H{sub 2} ensemble.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A Polynomial-Time Classical Algorithm for Noisy Quantum Circuits

We provide a polynomial-time classical algorithm for noisy quantum circuits. The algorithm computes the expectation value of any observable for any circuit, with a small average error over input states drawn from an ensemble (e.g., the computational basis). Our approach is based upon the intuition that noise exponentially damps nonlocal correlations relative to local correlations. This enables one to classically simulate a noisy quantum circuit by keeping track of only the dynamics of local quantum information. Our algorithm also enables sampling from the output distribution of a circuit in quasipolynomial time, so long as the distribution anticoncentrates. A number of implications are discussed, including a fundamental limit on the efficacy of noise mitigation strategies: For constant noise rates, any quantum circuit for which error mitigation succeeds in polynomial-time on most input states can also be classically simulated in polynomial-time on most input states. Our algorithms scale exponentially in the inverse noise rate, which is fundamental and makes them impractical for current quantum devices.

decoherence↗

Classical harmonic generation in rare gases

The classical microcanonical ensemble approach to high-harmonic generation (HHG) in rare gases subjected to intense laser fields is studied. We show that the ensemble spectrum is a ''sampled'' version of the single trajectory spectrum. Unlike the radiation of the single ensemble member, the total ensemble radiation possesses all the basic HHG features: odd laser harmonics, plateau, and cutoff. The sampling theorem for uniform grids is used to explain why the ensemble spectrum can be computed accurately with a very small number of ensemble members compared to the Monte Carlo method. Furthermore, The phase space relevant to harmonic generation is found to be significantly smaller than the field free microcanonical ensemble. In addition we demonstrate the seeding effect that was predicted and observed in quantum simulation. For circular polarization, we verify that the harmonic generation is highly suppressed even when the argument of the three-step model does not apply. All the findings are numerically calculated for the xenon atom.

74 ATOMIC AND MOLECULAR PHYSICS↗

Huge ensembles – Part 1: Design of ensemble weather forecasts using spherical Fourier neural operators

Abstract. Simulating low-likelihood high-impact extreme weather events in a warming world is a significant and challenging task for current ensemble forecasting systems. While these systems presently use up to 100 members, larger ensembles could enrich the sampling of internal variability. They may capture the long tails associated with climate hazards better than traditional ensemble sizes. Due to computational constraints, it is infeasible to generate huge ensembles (comprised of 1000–10 000 members) with traditional, physics-based numerical models. In this two-part paper, we replace traditional numerical simulations with machine learning (ML) to generate hindcasts of huge ensembles. In Part 1, we construct an ensemble weather forecasting system based on spherical Fourier neural operators (SFNOs), and we discuss important design decisions for constructing such an ensemble. The ensemble represents model uncertainty through perturbed-parameter techniques, and it represents initial condition uncertainty through bred vectors, which sample the fastest-growing modes of the forecast. Using the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (IFS) as a baseline, we develop an evaluation pipeline composed of mean, spectral, and extreme diagnostics. With large-scale, distributed SFNOs with 1.1 billion learned parameters, we achieve calibrated probabilistic forecasts. As the trajectories of the individual members diverge, the ML ensemble mean spectra degrade with lead time, consistent with physical expectations. However, the individual ensemble members' spectra stay constant with lead time. Therefore, these members simulate realistic weather states during the rollout, and the ML ensemble passes a crucial spectral test in the literature. The IFS and ML ensembles have similar extreme forecast indices, and we show that the ML extreme weather forecasts are reliable and discriminating. These diagnostics ensure that the ensemble can reliably simulate the time evolution of the atmosphere, including low-likelihood high-impact extremes. In Part 2, we generate a huge ensemble initialized each day in summer 2023, and we characterize the simulations of extremes.

Mahesh, Ankur↗

Parallel solution of closely coupled systems

An odd-even permutation and a nested dissection technique were used to circumvent the strong seriality of a system of closely coupled equations. The effect of transforming the n x n Hermitian definite positive matrix coefficient on the topology of Cholesky factors is discussed. A series of directed graphs is constructed in order to show the computational steps required for the odd-even permutation. Numerical expressions for the speed-up and efficiency of parallel N-processing techniques and sequential processing by a single computer are derived. Similar expressions are derived for the case of insufficient processing capacity. The application of the odd-even permutation to the ensemble class of computer architectures is demonstrated.

Utku, S.↗

Computer system with moving target defenses against vulnerability attacks

A computer system includes an ensemble moving target defense architecture that protects the computer system against attack using one or more composable protection layers that change each churn cycle, thereby requiring an attacker to acquire information needed for an attack (e.g., code and pointers) and successfully deploy the attack, before the layers have changed state. Each layer may deploy a respective attack information asset protection providing multiple respective attack protections each churn cycle, wherein the respective attack information asset protections may differ.

Austin, Todd↗

Quantifying uncertainty for deep learning based forecasting and flow-reconstruction using neural architecture search ensembles

Classical problems in computational physics such as data-driven forecasting and signal reconstruction from sparse sensors have recently seen an explosion in deep neural network (DNN) based algorithmic approaches. However, most DNN models do not provide uncertainty estimates, which are crucial for establishing the trustworthiness of these techniques in downstream decision making tasks and scenarios. In recent years, ensemble-based methods have achieved significant success for the uncertainty quantification in DNNs on a number of benchmark problems. However, their performance on real-world applications remains under-explored. In this work, we present an automated approach to DNN discovery and demonstrate how this may also be utilized for ensemble-based uncertainty quantification. Specifically, we propose the use of a scalable neural and hyperparameter architecture search for discovering an ensemble of DNN models for complex dynamical systems. We highlight how the proposed method not only discovers high-performing neural network ensembles for our tasks, but also quantifies uncertainty seamlessly. This is achieved by using genetic algorithms and Bayesian optimization for sampling the search space of neural network architectures and hyperparameters. Subsequently, a model selection approach is used to identify candidate models for an ensemble set construction. Afterwards, a variance decomposition approach is used to estimate the uncertainty of the predictions from the ensemble. We demonstrate the feasibility of this framework for two tasks — forecasting from historical data and flow reconstruction from sparse sensors for the sea-surface temperature. In conclusion, we demonstrate superior performance from the ensemble in contrast with individual high-performing models and other benchmarks.

Deep ensembles↗

Statistical perspective on embrittling potency for intergranular fracture

Embrittling potency is a thermodynamic metric that assesses the influence of solute segregation to a grain boundary (GB) on intergranular fracture. Historically, authors of studies have reported embrittling potency as a single scalar value, assuming a single segregation site of importance at a GB and a particular cleavage plane. However, the topography of intergranular fracture surfaces is not generally known a priori. Accordingly, we, in this study, present a statistical ensemble approach to compute embrittling potency, where many free surface (FS) permutations are systematically considered to model fracture of a GB. The result is a statistical description of the thermodynamics of GB embrittlement. As a specific example, embrittling potency distributions are presented for Cr segregation to sites at two Ni $\langle 111 \rangle$ symmetric tilt GBs using atomistic simulations. We show that the average embrittling potency for a particular GB site, considering an ensemble of FS permutations, is not equal to the embrittling potency computed using the lowest energy pair of FSs. A mean GB embrittlement is proposed, considering both the likelihood of formation of a particular FS and the probability of solute occupancy at each GB site, to compare the relative embrittling behavior of two distinct GBs.

36 MATERIALS SCIENCE↗

Simulating excited-state complex ensembles: Fluorescence and solvatochromism in amine-arene exciplexes

Exciplexes are excited-state complexes formed as a result of partial charge transfer from the donor to the acceptor species when one moiety of the donor–acceptor pair is electronically excited. The arene–amine exciplex formed between oligo-(p-phenylene) (OPP) and triethylamine (TEA) is of interest in the catalytic photoreduction of CO2 because it can compete with complete electron transfer to the OPP catalyst. Therefore, formation of the exciplex can hinder the generation of a radical anion OPP·− necessary for subsequent CO2 reduction. We report an implementation of a workflow automating quantum-chemistry calculations that generate and characterize an ensemble of structures to represent this exciplex state. We use FireWorks, Pymatgen, and Custodian Python packages for high-throughput ensemble generation. The workflow includes time-dependent density functional theory optimization, verification of excited-state minima, and exciplex characterization with natural transition orbitals, exciton analysis, excited-state Mulliken charges, and energy decomposition analysis. Fluorescence spectra computed for these ensembles using Boltzmann-weighted contributions of each structure agree better with experiment than our previous calculations based on a single representative exciplex structure [Kron et al., J. Phys. Chem. A 126, 2319–2329 (2022)]. The ensemble description of the exciplex state also reproduces an experimentally observed red shift of the emission spectrum of [OPP-4–TEA]* relative to [OPP-3–TEA]*. The workflow developed here streamlines otherwise labor-intensive calculations that would require significant user involvement and intervention.

Chemistry↗

Categorization of Computing Education Resources into the ACM Computing Classification System

The Ensemble Portal harvests resources from multiple heterogonous federated collections. Managing these dynamically increasing collections requires an automatic mechanism to categorize records in to corresponding topics. We propose an approach to use existing ACM DL metadata to build classifiers for harvested resources in the Ensemble project. We also present our experience on utilizing the Amazon Mechanical Turk platform to build ground truth training data sets from Ensemble collections.

99 GENERAL AND MISCELLANEOUS↗

Gradient Flow and Scale Setting on MILC HISQ Ensembles

We report on a scale determination with gradient-flow techniques on the N f = 2 + 1 + 1 HISQ ensembles generated by the MILC collaboration. The ensembles include four lattice spacings, ranging from approximately 0.15 to 0.06 fm, and both physical and unphysical values of the quark masses. The scales p √t 0 /a and w 0 /a and their tree-level improvements,√t 0;imp and w 0;imp , are computed on each ensemble using Symanzik ow and the cloverleaf definition of the energy density E. Using a combination of continuum chiral perturbation theory and a Taylor-series ansatz for the lattice-spacing and strong-coupling dependence, the results are simultaneously extrapolated to the continuum and interpolated to physical quark masses. We also determine the scales p t 0 = 0:1416( +8 -5 ) fm and w 0 = 0:1717( +12 -11 ) fm, where the errors are sums, in quadrature, of statistical and all systematic errors. The precision of w 0 and √t 0 is comparable to or more precise than the best previous estimates, respectively. We also find the continuum mass-dependence of w 0 that will be useful for estimating the scales of other ensembles. Furthermore, we estimate the integrated autocorrelation length of . For long flow times, the autocorrelation length of appears to be comparable to or smaller than that of the topological charge.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Quantification of regional net CO 2 flux errors in the Orbiting Carbon Observatory-2 (OCO-2) v10 model intercomparison project (MIP) ensemble using airborne measurements

Inverse model intercomparison projects (MIPs) provide a chance to assess the uncertainties in inversion estimates arising from various sources. However, accurately quantifying ensemble CO 2 flux errors remains challenging and often relies on the ensemble spread. This study proposes a method for quantifying the errors in regional net surface–atmosphere CO 2 flux estimates from models taken from the Orbiting Carbon Observatory-2 (OCO-2) v10 MIP by using independent airborne CO 2 measurements for the period 2015–2017. We first calculate the root mean square error (RMSE) between the ensemble mean of posterior CO 2 concentrations and airborne observations and then isolate the CO 2 concentration errors caused solely by the ensemble mean of posterior net fluxes by subtracting the observation, representation, and transport errors from seven regions. Our analysis reveals that the flux errors projected onto CO 2 space account for 55 %–85 % of the regional average RMSE over the 3 years, ranging from 0.88 to 1.91 ppm. In five regions, the error estimates based on observations exceed those computed from the ensemble spread of posterior fluxes by a factor of 1.33–1.93, implying an underestimation of the actual flux errors, while their magnitudes are comparable in two regions. The adjoint sensitivity analysis identifies that the underestimation of flux errors is prominent where the magnitudes of fossil fuel emissions exceed those of terrestrial-biosphere fluxes by a factor of 3–31 over the 3 years. This suggests the presence of systematic biases in the inversion estimates associated with errors in the prescribed fossil fuel emissions common to all models. Our study emphasizes the value of airborne measurements for quantifying regional errors in ensemble net CO 2 flux estimates.

54 ENVIRONMENTAL SCIENCES↗

Constant-pH Molecular Dynamics Simulations for Large Biomolecular Systems

We report that an increasingly important endeavor is to develop computational strategies that enable molecular dynamics (MD) simulations of biomolecular systems with spontaneous changes in protonation states under conditions of constant pH. The present work describes our efforts to implement the powerful constant-pH MD simulation method, based on a hybrid nonequilibrium MD/Monte Carlo (neMD/MC) technique within the highly scalable program NAMD. The constant-pH hybrid neMD/MC method has several appealing features; it samples the correct semigrand canonical ensemble rigorously, the computational cost increases linearly with the number of titratable sites, and it is applicable to explicit solvent simulations. The present implementation of the constant-pH hybrid neMD/MC in NAMD is designed to handle a wide range of biomolecular systems with no constraints on the choice of force field. Furthermore, the sampling efficiency can be adaptively improved on-the-fly by adjusting algorithmic parameters during the simulation. Finally, illustrative examples emphasizing medium- and large-scale applications on next-generation supercomputing architectures are provided.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Studying Rate of Fire Spread Using Landscape Statistics and FIRETEC

Wildfires are multi-system and multi-scale processes that are influenced by the land and atmosphere. Here, we study the influence of land, in terms of topography and vegetation, on uphill and downhill fire spread using data of the state of California as a testbed. The California landscape is first quantified by reclassifying its dominant vegetation types into grass, trees, and shrubs. Next, statistics of slope, vegetation height, and live fuel moisture percentage for each dominant vegetation type are gathered to create ensembles of idealized computational fluid dynamics (CFD) model simulations representative of the landscape. Rate of spread values for each CFD simulation are recorded to observe the differences in perturbing slope and fuel moisture for each vegetation type. We have found that all vegetation types prefer slopes of less than 30 degrees. Trees in the California data are 10 to 40 meters high, whereas shrub heights are less than 3 meters and grass heights are less than 1 meter. The live fuel moisture percentage of both trees and shrubs fluctuates around approximately 100% over one year (2019-2020). Using these parameters as inputs into the CFD model, fire spread rate is found to be fastest in the grass ensemble (maximum 12 m s -1 ), followed by shrubs (maximum 10 m s -1 ) and lastly by trees (maximum 7 m s -1 ). Increasing slope increases the spread rate of the uphill fire but the downhill spread decreases for slopes greater than 15 degrees. Increasing fuel moisture decreases the spread rate both up and down the hill. This study highlights the impacts that topography and vegetation have on fire spread.

54 ENVIRONMENTAL SCIENCES↗

A new method for simulating atmospheric turbulence for rotorcraft applications

Simulation of atmospheric turbulence as seen by a rotating blade element involves treatment of cyclostationary processes. Conventional filtering techniques do not lend themselves well to the generation of such turbulence sample functions as are required in rotorcraft flight dynamics simulation codes. A method to generate sample functions containing second order statistics of mean and covariance is presented. Compared to ensemble averaging involving excessive computer time, the novelty is to exploit cycloergodicity and thereby, replace ensemble averaging by averaging over a single path sample function of long duration. The method is validated by comparing its covariance results with the analytical and ensemble averaged results for a widely used 1-D turbulence approximation.

J. Riaz↗

A new method for simulating atmospheric turbulence for rotorcraft applications

Simulation of atmospheric turbulence as seen by a rotating blade element involves treatment of cyclostationary processes. Conventional filtering techniques do not lend themselves well to the generation of such turbulence sample functions as are required in rotorcraft flight dynamics simulation codes. A method to generate sample functions containing second-order statistics of mean and covariance is presented. Compared to ensemble averaging involving excessive computer time, the novelty is to exploit cycloergodicity and thereby, replace ensemble averaging by averaging over a single-path sample function of long duration. The method is validated by comparing its covariance results with the analytical and ensemble-averaged results for a widely used one-dimensional turbulence approximation.

Prasad, J. V. R.↗

CoSim: A Simulator for Co-Scheduling of Batch and On-Demand Jobs in HPC Datacenters

The increasing scale and complexity of scientific applications are rapidly transforming the ecosystem of tools, methods, and workflows adopted by the high-performance computing (HPC) community. Big data analytics and deep learning are gaining traction as essential components in this ecosystem in a variety of scenarios, such as, steering of experimental instruments , acceleration of high-fidelity simulations through surrogate computations, and guided ensemble searches. In this context, the batch job model traditionally adopted by the supercomputing in-frastructures needs to be complemented with support to schedule opportunistic on-demand analytics jobs, leading to the problem of efficient preemption of batch jobs with minimum loss of progress. In this paper, we design and implement a simulator, CoSim, that enables on-the-fly analysis of the trade-offs arising between delaying the start of opportunistic on-demand jobs, which leads to longer analytics latency, and loss of progress due to preemption of batch jobs, which is necessary to make room for on-demand jobs. To this end, we propose an algorithm based on dynamic programming with predictable performance and scalability that enables supercomputing infrastructure schedulers to analyze the aforementioned trade-off and take decisions in near real-time. Compared with other state-of-art approaches using traces of the Theta pre-Exascale machine, our approach is capable of finding the optimal solution, while achieving high performance and scalability.

batch job preemption↗