Search NASA⌕ Search

SEARCH · Search NASA

Results for “approximate 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 793 records · Page 44

Accelerating multilevel Markov Chain Monte Carlo using machine learning models

Here, this work presents an efficient approach for accelerating multilevel Markov Chain Monte Carlo (MCMC) sampling for large-scale problems using low-fidelity machine learning models. While conventional techniques for large-scale Bayesian inference often substitute computationally expensive high-fidelity models with machine learning models, thereby introducing approximation errors, our approach offers a computationally efficient alternative by augmenting high-fidelity models with low-fidelity ones within a hierarchical framework. The multilevel approach utilizes the low-fidelity machine learning model (MLM) for inexpensive evaluation of proposed samples thereby improving the acceptance of samples by the high-fidelity model. The hierarchy in our multilevel algorithm is derived from geometric multigrid hierarchy. We utilize an MLM to accelerate the coarse level sampling. Training machine learning model for the coarsest level significantly reduces the computational cost associated with generating training data and training the model. We present an MCMC algorithm to accelerate the coarsest level sampling using MLM and account for the approximation error introduced. We provide theoretical proofs of detailed balance and demonstrate that our multilevel approach constitutes a consistent MCMC algorithm. Additionally, we derive the expression for cost reduction due to machine learning model to facilitate cost analysis of the hierarchical sampling algorithm. Our technique is demonstrated on a standard benchmark inference problem in groundwater flow, where we estimate the probability density of a quantity of interest using a four-level MCMC algorithm. Our proposed algorithm accelerates multilevel sampling by a factor of two while achieving similar accuracy compared to sampling using the standard multilevel algorithm.

97 MATHEMATICS AND COMPUTING↗

Note: A pure-sampling quantum Monte Carlo algorithm with independent Metropolis

Recently, Ospadov and Rothstein published a pure-sampling quantum Monte Carlo algorithm (PSQMC) that features an auxiliary Path Z that connects the midpoints of the current and proposed Paths X and Y, respectively. When sufficiently long, Path Z provides statistical independence of Paths X and Y. Under those conditions, the Metropolis decision used in PSQMC is done without any approximation, i.e., not requiring microscopic reversibility and without having to introduce any G(x → x′; τ) factors into its decision function. This is a unique feature that contrasts with all competing reptation algorithms in the literature. An example illustrates that dependence of Paths X and Y has adverse consequences for pure sampling.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Tau leaping of stiff stochastic chemical systems via local central limit approximation

Stiffness manifests in stochastic dynamic systems in a more complex manner than in deterministic systems; it is not only important for a time-stepping method to remain stable but it is also important for the method to capture the asymptotic variances accurately. In the context of stochastic chemical systems, time stepping methods are known as tau leaping. Well known existing tau leaping methods have shortcomings in this regard. The implicit tau method is far more stable than the trapezoidal tau method but underestimates the asymptotic variance. On the other hand, the trapezoidal tau method which estimates the asymptotic variance exactly for linear systems suffers from the fact that the transients of the method do not decay fast enough in the context of very stiff systems. We propose a tau leaping method that possesses the same stability properties as the implicit method while it also captures the asymptotic variance with reasonable accuracy at least for the test system S{sub 1}↔S{sub 2}. The proposed method uses a central limit approximation (CLA) locally over the tau leaping interval and is referred to as the LCLA-τ. The CLA predicts the mean and covariance as solutions of certain differential equations (ODEs) and for efficiency we solve these using a single time step of a suitable low order method. We perform a mean/covariance stability analysis of various possible low order schemes to determine the best scheme. Numerical experiments presented show that LCLA-τ performs favorably for stiff systems and that the LCLA-τ is also able to capture bimodal distributions unlike the CLA itself. The proposed LCLA-τ method uses a split implicit step to compute the mean update. We also prove that any tau leaping method employing a split implicit step converges in the fluid limit to the implicit Euler method as applied to the fluid limit differential equation.

97 MATHEMATICS AND COMPUTING↗

Effect of magnetic field on the donor impurity in CdTe/Cd{sub 1-x}Mn{sub x}Te quantum well wire

The donor impurity binding energy in CdTe / Cd{sub 1-x}Mn{sub x}Te QWW with square well confinement along x – direction and parabolic confinement along y – direction under the influence of externally applied magnetic field has been computed using variational principle in the effective mass approximation. The spin polaronic shift has also been computed. The results are presented and discussed.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Divergence of Solutions to Perturbation-Based Advection-Dispersion Moment Equations

Moment equation methods are popular and powerful tools for modeling transport processes in randomly heterogeneous porous media, but the application of these methods to advection-dispersion equations often leads to erroneous oscillations. Perturbative methods, required to close systems of moment equations, become inaccurate for large perturbations; however, little quantitative theory exists for determining when this occurs for advection-dispersion equations. We consider three different methods (asymptotic approximation, Eulerian truncation, and iterative solution) for closing and solving advection-dispersion moment equations describing transport in stratified porous media with random permeability. We obtain approximate analytical expressions for time above which the asymptotic approximation to the mean diverges, in particular quantifying the impact that dispersion has on delaying---but not eliminating---divergence. We demonstrate that Eulerian truncation and iterative solution methods do not eliminate divergent behavior either. Our divergence criteria provide \emph{a priori} estimates that signal a warning to the practitioner of stochastic advection-dispersion equations to carefully consider whether to apply perturbative approaches.

58 GEOSCIENCES↗

An efficient Bayesian data-worth analysis using a multilevel Monte Carlo method

Improving the understanding of subsurface systems and thus reducing prediction uncertainty requires collection of data. As the collection of subsurface data is costly, it is important that the data collection scheme is cost-effective. Design of a cost-effective data collection scheme, i.e., data-worth analysis, requires quantifying model parameter, prediction, and both current and potential data uncertainties. Assessment of these uncertainties in large-scale stochastic subsurface hydrological model simulations using standard Monte Carlo (MC) sampling or surrogate modeling is extremely computationally intensive, sometimes even infeasible. In this paper, we propose an efficient Bayesian data-worth analysis using a multilevel Monte Carlo (MLMC) method. Compared to the standard MC that requires a significantly large number of high-fidelity model executions to achieve a prescribed accuracy in estimating expectations, the MLMC can substantially reduce computational costs using multifidelity approximations. Since the Bayesian data-worth analysis involves a great deal of expectation estimation, the cost saving of the MLMC in the assessment can be outstanding. While the proposed MLMC-based data-worth analysis is broadly applicable, we use it for a highly heterogeneous two-phase subsurface flow simulation to select an optimal candidate data set that gives the largest uncertainty reduction in predicting mass flow rates at four production wells. Finally, the choices made by the MLMC estimation are validated by the actual measurements of the potential data, and consistent with the standard MC estimation. But compared to the standard MC, the MLMC greatly reduces the computational costs.

97 MATHEMATICS AND COMPUTING↗

Ab initio multiple cloning algorithm for quantum nonadiabatic molecular dynamics

We present a new algorithm for ab initio quantum nonadiabatic molecular dynamics that combines the best features of ab initio Multiple Spawning (AIMS) and Multiconfigurational Ehrenfest (MCE) methods. In this new method, ab initio multiple cloning (AIMC), the individual trajectory basis functions (TBFs) follow Ehrenfest equations of motion (as in MCE). However, the basis set is expanded (as in AIMS) when these TBFs become sufficiently mixed, preventing prolonged evolution on an averaged potential energy surface. We refer to the expansion of the basis set as “cloning,” in analogy to the “spawning” procedure in AIMS. This synthesis of AIMS and MCE allows us to leverage the benefits of mean-field evolution during periods of strong nonadiabatic coupling while simultaneously avoiding mean-field artifacts in Ehrenfest dynamics. We explore the use of time-displaced basis sets, “trains,” as a means of expanding the basis set for little cost. We also introduce a new bra-ket averaged Taylor expansion (BAT) to approximate the necessary potential energy and nonadiabatic coupling matrix elements. The BAT approximation avoids the necessity of computing electronic structure information at intermediate points between TBFs, as is usually done in saddle-point approximations used in AIMS. The efficiency of AIMC is demonstrated on the nonradiative decay of the first excited state of ethylene. The AIMC method has been implemented within the AIMS-MOLPRO package, which was extended to include Ehrenfest basis functions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Expectation propagation for nonlinear inverse problems – with an application to electrical impedance tomography

In this paper, we study a fast approximate inference method based on expectation propagation for exploring the posterior probability distribution arising from the Bayesian formulation of nonlinear inverse problems. It is capable of efficiently delivering reliable estimates of the posterior mean and covariance, thereby providing an inverse solution together with quantified uncertainties. Some theoretical properties of the iterative algorithm are discussed, and the efficient implementation for an important class of problems of projection type is described. The method is illustrated with one typical nonlinear inverse problem, electrical impedance tomography with complete electrode model, under sparsity constraints. Numerical results for real experimental data are presented, and compared with that by Markov chain Monte Carlo. The results indicate that the method is accurate and computationally very efficient.

97 MATHEMATICS AND COMPUTING↗

Vibrations of Elastic Shells of Revolution Partially Filled with Ideal Liquid

We propose an algorithm for the numerical analyses of vibrations of elastic shells of revolution partially filled with ideal incompressible liquids. In the solution of this problem, the wave motions of liquid on its free surface are taken into account. The solution of the problem of hydroelasticity is based on the application of the method of decomposition of the domain of integration of equations of the theory of shells with the use of the variational statement of the problem and on the approximate construction of the operator inverse to the operator of the hydrodynamic part of the problem. We construct a generalized functional of displacements of the shell for which the role of conditions of matching for the solutions in different subdomains is played by the natural boundary conditions. The obtained numerical results are compared with the existing exact solutions of the analyzed problem for a shell in the form of a straight circular cylinder.

42 ENGINEERING↗

Solution of stochastic media transport problems using a numerical quadrature-based method

We present a new conceptual framework for analyzing transport problems in random media. We decompose such problems into stratified subproblems according to the number of material pseudo-interfaces within realizations. For a given subproblem we assign pseudo-interface locations in each realization according to product quadrature rules, which allows us to deterministically generate a fixed number of realizations. Quadrature integration of the solutions of these realizations thus approximately solves each subproblem; the weighted superposition of solutions of the subproblems approximately solves the general stochastic media transport problem. We revisit some benchmark problems to determine the accuracy and efficiency of this approach in comparison to randomly generated realizations. We find that this method is very accurate and fast when the number of pseudo-interfaces in a problem is generally low, but that these advantages quickly degrade as the number of pseudo-interfaces increases. (authors)

97 MATHEMATICS AND COMPUTING↗

On linearization and preconditioning for radiation diffusion coupled to material thermal conduction equations

Jacobian-free Newton–Krylov (JFNK) method is an effective algorithm for solving large scale nonlinear equations. One of the most important advantages of JFNK method is that there is no necessity to form and store the Jacobian matrix of the nonlinear system when JFNK method is employed. However, an approximation of the Jacobian is needed for the purpose of preconditioning. In this paper, JFNK method is employed to solve a class of non-equilibrium radiation diffusion coupled to material thermal conduction equations, and two preconditioners are designed by linearizing the equations in two methods. Numerical results show that the two preconditioning methods can improve the convergence behavior and efficiency of JFNK method.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Universality versus material dependence of fluctuation forces between metallic wires

We calculate the Casimir interaction between two parallel wires and between a wire and a metal plate. The dielectric properties of the objects are described by the plasma, Drude, and perfect metal models. We find that at asymptotically large separation interactions involving plasma, wires and/or plates are independent of the material properties but depend on the dc conductivity {sigma} for Drude wires. Counterintuitively, at intermediate separations the interaction involving Drude wires can become independent of {sigma}. At smaller separations, we compute the interaction numerically and observe an approach to the proximity approximation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Sign Problem in Tensor-Network Contraction

We investigate how the computational difficulty of contracting tensor networks depends on the sign structure of the tensor entries. Using results from computational complexity, we observe that the approximate contraction of tensor networks with only positive entries has lower computational complexity as compared to tensor networks with general real or complex entries. This raises the question of how this transition in computational complexity manifests itself in the hardness of different tensor-network-contraction schemes. We pursue this question by studying random tensor networks with varying bias toward positive entries. First, we consider contraction via Monte Carlo sampling and find that the transition from hard to easy occurs when the tensor entries become predominantly positive; this can be understood as a tensor-network manifestation of the well-known negative-sign problem in quantum Monte Carlo. Second, we analyze the commonly used contraction based on boundary tensor networks. The performance of this scheme is governed by the number of correlations in contiguous parts of the tensor network (which by analogy can be thought of as entanglement). Remarkably, we find that the transition from hard to easy—i.e., from a volume-law to a boundary-law scaling of entanglement—already occurs for a slight bias of the tensor entries toward a positive mean, scaling inversely with the bond dimension D , and thus the problem becomes easy the earlier the larger D occurs. This is in contrast both to expectations and to the behavior found in Monte Carlo contraction, where the hardness at fixed bias increases with the bond dimension. To provide insight into this early breakdown of computational hardness and the accompanying entanglement transition, we construct an effective classical statistical-mechanical model that predicts a transition at a bias of the tensor entries of 1 / D , confirming our observations. We conclude by investigating the computational difficulty of computing expectation values of tensor-network wave functions (projected entangled-pair states, PEPSs) and find that in this setting, the complexity of entanglement-based contraction always remains low. We explain this by providing a local transformation that maps PEPS expectation values to a positive-valued tensor network. This not only provides insight into the origin of the observed boundary-law entanglement scaling but also suggests new approaches toward PEPS contraction based on positive decompositions. Published by the American Physical Society 2025

Chen, Jielun (ORCID:0000000178411545)↗

Electric Power Distribution System Model Simplification Using Segment Substitution

Quasi-static time-series (QSTS) simulation is used to simulate the behavior of distribution systems over long periods of time (typically hours to years). The technique involves repeatedly solving the load-flow problem for a distribution system model and is useful for distributed energy resource (DER) planning. When a QSTS simulation has a small time step and a long duration, the computational burden of the simulation can be a barrier to integration into utility workflows. One way to relieve the computational burden is to simplify the system model. The segment substitution method of simplifying distribution system models introduced in this paper offers model bus reduction of up to 98% with a simplification error as low as 0.2% (0.002 pu voltage). In contrast to existing methods of distribution system model simplification, which rely on topological inspection and linearization, the segment substitution method uses black-box segment data and an assumed simplified topology.

Reiman, Andrew P. (ORCID:0000000302042960)↗

COMBINE7.1 - A Portable ENDF/B-VII.0 Based Neutron Spectrum and Cross-Section Generation Program

COMBINE7.1 is a FORTRAN 90 computer code that generates multigroup neutron constants for use in the deterministic diffusion and transport theory neutronics analysis. The cross-section database used by COMBINE7.1 is derived from the Evaluated Nuclear Data Files (ENDF/B-VII.0). The neutron energy range covered is from 20 MeV to 1.0E-5 eV. The Los Alamos National Laboratory NJOY code is used as the processing code to generate a 167 fine-group cross-section library in MATXS format for Bondarenko self-shielding treatment. Resolved resonance parameters are extracted from ENDF/B-VII.0 File 2 for a separate library to be used in an alternate Nordheim self-shielding treatment in the resolved resonance energy range. The equations solved for energy dependent neutron spectrum in the 167 fine-group structure are the B-3 or B-1 approximations to the transport equation. The fine group cross sections needed for the spectrum calculation are first prepared by Bondarenko self-shielding interpolation in terms of background cross section and temperature. The geometric lump effect, when present, is accounted for by augmenting the background cross section. Nordheim self-shielded fine group cross sections for a material having resolved resonance parameters overwrite correspondingly the existing self-shielded fine group cross sections when this option is used. The fine group cross sections in the thermal energy range are replaced by those self-shielded with the Amouyal/Benoist/Horowitz method in the three region geometry when this option is requested. COMBINE7.1 coalesces fine group cross sections into broad group macroscopic and microscopic constants. The coalescing is performed by utilizing fine-group fluxes and/or currents obtained by spectrum calculation as the weighting functions. The multigroup constant may be output in any of several standard formats including ANISN 14** free format, CCCC ISOTXS format, and AMPX working library format. ANISN-PC, a one-dimensional, discrete-ordinate transport code, is incoprated into COMBINE7.1. As an option, the 167 fine-group constants generated by COMBINE portion in the program can be used to cacluate regionwise spectra in the ANISN portion, all internally to reflect the one-dimensional transport correction. Results for the criticality validation calculations are included as a part of verification and validation.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

COMBINE7.1 - A Portable ENDF/B-VII.0 Based Neutron Spectrum and Cross-Section Generation Program

COMBINE7.1 is a FORTRAN 90 computer code that generates multigroup neutron constants for use in the deterministic diffusion and transport theory neutronics analysis. The cross-section database used by COMBINE7.1 is derived from the Evaluated Nuclear Data Files (ENDF/B-VII.0). The neutron energy range covered is from 20 MeV to 1.0E-5 eV. The Los Alamos National Laboratory NJOY code is used as the processing code to generate a 167 fine-group cross-section library in MATXS format for Bondarenko self-shielding treatment. Resolved resonance parameters are extracted from ENDF/B-VII.0 File 2 for a separate library to be used in an alternate Nordheim self-shielding treatment in the resolved resonance energy range. The equations solved for energy dependent neutron spectrum in the 167 fine-group structure are the B3 or B1 zero-dimensional approximations to the transport equation. The fine group cross sections needed for the spectrum calculation are first prepared by Bondarenko self-shielding interpolation in terms of background cross section and temperature. The geometric lump effect, when present, is accounted for by augmenting the background cross section. Nordheim self-shielded fine group cross sections for a material having resolved resonance parameters overwrite correspondingly the existing self-shielded fine group cross sections when this option is used. COMBINE7.1 coalesces fine group cross sections into broad group macroscopic and microscopic constants. The coalescing is performed by utilizing fine-group fluxes and/or currents obtained by spectrum calculation as the weighting functions. The multigroup constants may be output in any of several standard formats including INL format, ANISN 14** free format, CCCC ISOTXS format, and AMPX working library format. ANISN-PC, a one-dimensional (1-D) discrete-ordinate transport code, is incorporated into COMBINE7.1. As an option, the 167 fine-group constants generated by zero-dimensional COMBINE portion in the program can be used to calculate regionwise spectra in the 1-D ANISN portion, all internally to reflect the 1-D transport correction. The regionwise spectra are then used to generate mutigroup regionwise neutron constants. The 1-D neutron transport can be performed up to three stages, e.g., from a TRISO fuel to PEBBLE to 1-D full core wedge. In addition, COMBINE7.1 has now the capability of adjoint flux calculation through the 1-D ANISN transport. Photon transport capability is also added. For this, a photon production and photo-atomic cross section library, MATNG.LIB, was generated in MATXS format through NJOY code. The photon production cross section matrix is of 167 neutron - 18 photon groups. Photo-atomic cross sections, including heating, are in 18 energy groups.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Noise effects on Padé approximants and conformal maps

Here, we analyze the properties of Padé and conformal map approximants for functions with branch points, in the situation where the expansion coefficients are only known with finite precision or are subject to noise. We prove that there is a universal scaling relation between the strength of the noise and the expansion order at which Padé or the conformal map breaks down. We illustrate this behavior with some physically relevant model test functions and with two non-trivial physical examples where the relevant Riemann surface has complicated structure.

97 MATHEMATICS AND COMPUTING↗

Efficient and Flexible Hierarchical Data Layouts for a Unified Encoding of Scalar Field Precision and Resolution

To address the problem of ever-growing scientific data sizes making data movement a major hindrance to analysis, we introduce a novel encoding for scalar fields: a unified tree of resolution and precision, specifically constructed so that valid cuts correspond to sensible approximations of the original field in the precision-resolution space. Furthermore, we introduce a highly flexible encoding of such trees that forms a parameterized family of data hierarchies. We discuss how different parameter choices lead to different trade-offs in practice, and show how specific choices result in known data representation schemes such as zfp[52], idx[58], and jpeg2000 [76]. Lastly, we provide system-level details and empirical evidence on how such hierarchies facilitate common approximate queries with minimal data movement and time, using real-world data sets ranging from a few gigabytes to nearly a terabyte in size. Experiments suggest that our new strategy of combining reductions in resolution and precision is competitive with state-of-the-art compression techniques with respect to data quality, while being significantly more flexible and orders of magnitude faster, and requiring significantly reduced resources.

97 MATHEMATICS AND COMPUTING↗