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 1,099 records · Page 61

Learning algorithms for stack filter classifiers

Stack Filters define a large class of increasing filter that is used widely in image and signal processing. The motivations for using an increasing filter instead of an unconstrained filter have been described as: (1) fast and efficient implementation, (2) the relationship to mathematical morphology and (3) more precise estimation with finite sample data. This last motivation is related to methods developed in machine learning and the relationship was explored in an earlier paper. In this paper we investigate this relationship by applying Stack Filters directly to classification problems. This provides a new perspective on how monotonicity constraints can help control estimation and approximation errors, and also suggests several new learning algorithms for Boolean function classifiers when they are applied to real-valued inputs.

97 MATHEMATICS AND COMPUTING↗

Extension of PENELOPE to protons: Simulation of nuclear reactions and benchmark with Geant4

Purpose: Describing the implementation of nuclear reactions in the extension of the Monte Carlo code (MC) PENELOPE to protons (PENH) and benchmarking with Geant4.Methods: PENH is based on mixed-simulation mechanics for both elastic and inelastic electromagnetic collisions (EM). The adopted differential cross sections for EM elastic collisions are calculated using the eikonal approximation with the Dirac–Hartree–Fock–Slater atomic potential. Cross sections for EM inelastic collisions are computed within the relativistic Born approximation, using the Sternheimer–Liljequist model of the generalized oscillator strength. Nuclear elastic and inelastic collisions were simulated using explicitly the scattering analysis interactive dialin database for {sup 1}H and ICRU 63 data for {sup 12}C, {sup 14}N, {sup 16}O, {sup 31}P, and {sup 40}Ca. Secondary protons, alphas, and deuterons were all simulated as protons, with the energy adapted to ensure consistent range. Prompt gamma emission can also be simulated upon user request. Simulations were performed in a water phantom with nuclear interactions switched off or on and integral depth–dose distributions were compared. Binary-cascade and precompound models were used for Geant4. Initial energies of 100 and 250 MeV were considered. For cases with no nuclear interactions simulated, additional simulations in a water phantom with tight resolution (1 mm in all directions) were performed with FLUKA. Finally, integral depth–dose distributions for a 250 MeV energy were computed with Geant4 and PENH in a homogeneous phantom with, first, ICRU striated muscle and, second, ICRU compact bone.Results: For simulations with EM collisions only, integral depth–dose distributions were within 1%/1 mm for doses higher than 10% of the Bragg-peak dose. For central-axis depth–dose and lateral profiles in a phantom with tight resolution, there are significant deviations between Geant4 and PENH (up to 60%/1 cm for depth–dose distributions). The agreement is much better with FLUKA, with deviations within 3%/3 mm. When nuclear interactions were turned on, agreement (within 6% before the Bragg-peak) between PENH and Geant4 was consistent with uncertainties on nuclear models and cross sections, whatever the material simulated (water, muscle, or bone).Conclusions: A detailed and flexible description of nuclear reactions has been implemented in the PENH extension of PENELOPE to protons, which utilizes a mixed-simulation scheme for both elastic and inelastic EM collisions, analogous to the well-established algorithm for electrons/positrons. PENH is compatible with all current main programs that use PENELOPE as the MC engine. The nuclear model of PENH is realistic enough to give dose distributions in fair agreement with those computed by Geant4.

61 RADIATION PROTECTION AND DOSIMETRY↗

High speed inlet calculations with real gas effects

A 2-D steady-state Navier-Stokes solver has been upgraded to include the effects of frozen and equilibrium air chemistry for applications to high speed flight vehicles. To provide a computationally economical first order approximation to the high temperature physics, variable thermodynamic data is used for the chemically frozen mode to allow for a variation with temperature of the air specific heats and enthalpy. For calculations involving air in chemical equilibrium, a specially modified version of the NASA Lewis Chemical Equilibrium Code, CEC, is used to compute the chemical composition and resultant thermochemical properties. The upgraded solver is demonstrated by comparing results from calorically perfect (C sub p=constant), thermally perfect (frozen) and equilibrium air calculations for a variety of geometries, and flight Mach numbers.

Coirier, William J.↗

High speed inlet calculations with real gas effects

A 2-D steady-state Navier-Stokes solver has been upgraded to include the effects of frozen and equilibrium air chemistry for applications to high speed flight vehicles. To provide a computationally economical first order approximation to the high temperature physics, variable thermodynamic data is used for the chemically frozen mode to allow for a variation with temperature of the air specific heats and enthalpy. For calculations involving air in chemical equilibrium, a specially modified version of the NASA Lewis Chemical Equilibrium Code, CEC, is used to compute the chemical composition and resultant thermochemical properties. The upgraded solver is demonstrated by comparing results from calorically perfect (C sub p=constant), thermally perfect (frozen) and equilibrium air calculations for a variety of geometries, and flight Mach numbers.

Coirier, William J.↗

Multi-plane moment-of-fluid interface reconstruction in 3D

Moment-of-fluid (MOF) methods for interface reconstruction approximate the region occupied by material in each mesh element only through reference to its geometric moments. Here, we present a 3D MOF method that represents the material (POM) in each cell as the convex intersection of the cell and multiple half-spaces, each selected to minimize the least-squares error between computed moments of the approximated material and provided reference moments. This optimization problem is highly non-linear and non-convex, making the numerical result very sensitive to the initial guess. To create an effective initial guess in each cell, we construct an ellipsoid from 0th–2nd order reference moments such that its shape corresponds with that of the POM. Within this ellipsoid we inscribe a polyhedron, and initialize the minimization problem with the half-spaces defined by each of its faces. The inscribed polyhedron has minimally 4 faces, and using up to 3rd order moments permits optimization over up to 20 unknown values. We therefore define MOF methods that utilize 4, 5, or 6 half-spaces, correspondingly initialized with the faces of a single inscribed tetrahedron, triangular prism, or hexahedron. Stability of the non-linear optimization is further improved with a prepossessing step that normalizes the reference moments according to the axes of the reference ellipsoid. Using this approach, the non-linear least-squares solver reliably converges to a near-global minimum from a single initial guess. We demonstrate accuracy and robustness using single-cell and multi-cell examples over a wide spectrum of geometry. In particular, we demonstrate our ability to exactly reproduce several important and complex features defined by up to four half-spaces, such as corners, filaments, filament tips, and embedded material in the cell.

3D interface reconstruction↗

Scalar relativistic computations of nuclear magnetic shielding and g-shifts with the zeroth-order regular approximation and range-separated hybrid density functionals

Density functional theory (DFT) calculations of NMR chemical shifts and molecular g-tensors with Gaussian-type orbitals are implemented via second-order energy derivatives within the scalar relativistic zeroth order regular approximation (ZORA) framework. Nonhybrid functionals, standard (global) hybrids, and range-separated (Coulomb-attenuated, long-range corrected) hybrid functionals are tested. Origin invariance of the results is ensured by use of gauge-including atomic orbital (GIAO) basis functions. The new implementation in the NWChem quantum chemistry package is verified by calculations of nuclear shielding constants for the heavy atoms in HX (X=F, Cl, Br, I, At) and H2X (X = O, S, Se, Te, Po), and Te chemical shifts in a number of tellurium compounds. The basis set and functional dependence of g-shifts is investigated for 14 radicals with light and heavy atoms. The problem of accurately predicting F NMR shielding in UF6-nCln, n = 1 to 6, is revisited. The results are sensitive to approximations in the density functionals, indicating a delicate balance of DFT self-interaction vs. correlation. For the uranium halides, the results with the range-separated functionals are mixed.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Classical symmetries and the Quantum Approximate Optimization Algorithm

Here, we study the relationship between the Quantum Approximate Optimization Algorithm (QAOA) and the underlying symmetries of the objective function to be optimized. Our approach formalizes the connection between quantum symmetry properties of the QAOA dynamics and the group of classical symmetries of the objective function. The connection is general and includes but is not limited to problems defined on graphs. We show a series of results exploring the connection and highlight examples of hard problem classes where a nontrivial symmetry subgroup can be obtained efficiently. In particular, we show how classical objective function symmetries lead to invariant measurement outcome probabilities across states connected by such symmetries, independent of the choice of algorithm parameters or number of layers. To illustrate the power of the developed connection, we apply machine learning techniques toward predicting QAOA performance based on symmetry considerations. We provide numerical evidence that a small set of graph symmetry properties suffices to predict the minimum QAOA depth required to achieve a target approximation ratio on the MaxCut problem, in a practically important setting where QAOA parameter schedules are constrained to be linear and hence easier to optimize.

97 MATHEMATICS AND COMPUTING↗

Numerical Computation of Sensitivities and the Adjoint Approach

We discuss the numerical computation of sensitivities via the adjoint approach in optimization problems governed by differential equations. We focus on the adjoint problem in its weak form. We show how one can avoid some of the problems with the adjoint approach, such as deriving suitable boundary conditions for the adjoint equation. We discuss the convergence of numerical approximations of the costate computed via the weak form of the adjoint problem and show the significance for the discrete adjoint problem.

Lewis, Robert Michael↗

Conservative Analytical Collision Probability for Design of Orbital Formations

The literature offers a number of approximations for analytically and/or efficiently computing the probability of collision between two space objects. However, only one of these techniques is a completely analytical approximation that is suitable for use in the preliminary design phase, when it is more important to quickly analyze a large segment of the trade space than it is to precisely compute collision probabilities. Unfortunately, among the types of formations that one might consider, some combine a range of conditions for which this analytical method is less suitable. This work proposes a simple, conservative approximation that produces reasonable upper bounds on the collision probability in such conditions. Although its estimates are much too conservative under other conditions, such conditions are typically well suited for use of the existing method.

Carpenter, J. Russell↗

Conservative Analytical Collision Probabilities for Orbital Formation Flying

The literature offers a number of approximations for analytically and/or efficiently computing the probability of collision between two space objects. However, only one of these techniques is a completely analytical approximation that is suitable for use in the preliminary design phase, when it is more important to quickly analyze a large segment of the trade space than it is to precisely compute collision probabilities. Unfortunately, among the types of formations that one might consider, some combine a range of conditions for which this analytical method is less suitable. This work proposes a simple, conservative approximation that produces reasonable upper bounds on the collision probability in such conditions. Although its estimates are much too conservative under other conditions, such conditions are typically well suited for use of the existing method.

Carpenter, J. Russell↗

A new class of ensemble conserving algorithms for approximate quantum dynamics: Theoretical formulation and model problems

We develop two classes of quasi-classical dynamics that are shown to conserve the initial quantum ensemble when used in combination with the Feynman-Kleinert approximation of the density operator. These dynamics are used to improve the Feynman-Kleinert implementation of the classical Wigner approximation for the evaluation of quantum time correlation functions known as Feynman-Kleinert linearized path-integral. As shown, both classes of dynamics are able to recover the exact classical and high temperature limits of the quantum time correlation function, while a subset is able to recover the exact harmonic limit. A comparison of the approximate quantum time correlation functions obtained from both classes of dynamics is made with the exact results for the challenging model problems of the quartic and double-well potentials. It is found that these dynamics provide a great improvement over the classical Wigner approximation, in which purely classical dynamics are used. In a special case, our first method becomes identical to centroid molecular dynamics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A unified and constructive framework for the universality of neural networks

Abstract One of the reasons why many neural networks are capable of replicating complicated tasks or functions is their universal approximation property. Though the past few decades have seen tremendous advances in theories of neural networks, a single constructive and elementary framework for neural network universality remains unavailable. This paper is an effort to provide a unified and constructive framework for the universality of a large class of activation functions including most of the existing ones. At the heart of the framework is the concept of neural network approximate identity (nAI). The main result is as follows: any nAI activation function is universal in the space of continuous functions on compacta. It turns out that most of the existing activation functions are nAI, and thus universal. The framework induces several advantages over the contemporary counterparts. First, it is constructive with elementary means from functional analysis, probability theory, and numerical analysis. Second, it is one of the first unified and constructive attempts that is valid for most of the existing activation functions. Third, it provides new proofs for most activation functions. Fourth, for a given activation and error tolerance, the framework provides precisely the architecture of the corresponding one-hidden neural network with a predetermined number of neurons and the values of weights/biases. Fifth, the framework allows us to abstractly present the first universal approximation with a favorable non-asymptotic rate. Sixth, our framework also provides insights into the developments, and hence providing constructive derivations, of some of the existing approaches.

97 MATHEMATICS AND COMPUTING↗

Eikonal Corrections for Nucleus-Nucleus Reaction Cross Sections - Paper 90

The goal of this research is to modify a Fortran computer code, which uses an optical potential approximation to quantum mechanical multiple-scattering theory to compute total and reaction cross sections for nucleon-nucleus and nucleus-nucleus collisions for use in space radiation protection and shielding applications. These nuclear collision cross sections are utilized in space radiation transport codes that describe the passage of solar and galactic cosmic radiation fields through matter. In the code being modified by this work, an Eikonal approximation, in analogy with the complex index of refraction used to describe absorption in optics, is used to formulate the nuclear scattering amplitudes that describe the total and reaction cross sections in the formalism. Previous work found that the cross sections calculated with this model sometimes underestimated the experimental data. It has been noted that the Eikonal phase functions, which are used to calculate these cross sections and which contain all of the scattering information in the model, can be improved by incorporating higher order corrections In this work, a previously-developed first-order correction to the Eikonal phase function is coded in Fortran and incorporated into the optical model computer code that calculates reaction cross sections for nucleon-nucleus and nucleus-nucleus collisions at energies relevant to space radiation protection and shielding. The main thrust of this work is to test the use of this higher order correction by calculating reaction cross sections for a variety of nuclear collisions and comparing the results to experimental data. (authors)

61 RADIATION PROTECTION AND DOSIMETRY↗

Least-squares solutions of generalized inverse eigenvalue problem over Hermitian–Hamiltonian matrices with a submatrix constraint

In this paper, a gradient-based iterative algorithm is proposed for finding the least-squares solutions of the following constrained generalized inverse eigenvalue problem: given X∈C{sup n×m}, Λ=diag(λ{sub 1},λ{sub 2},…,λ{sub m})∈C{sup m×m}, find A{sup ∗},B{sup ∗}∈C{sup n×n}, such that ∥AX−BXΛ∥ is minimized, where A{sup ∗},B{sup ∗} are Hermitian–Hamiltonian except for a special submatrix. For any initial constrained matrices, a solution pair (A{sup ∗},B{sup ∗}) can be obtained in finite iteration steps by this iterative algorithm in the absence of roundoff errors. The least-norm solution can be obtained by choosing a special kind of initial matrix pencil. In addition, the unique optimal approximation solution to a given matrix pencil in the solution set of the above problem can also be obtained. A numerical example is given to show the efficiency of the proposed algorithm.

97 MATHEMATICS AND COMPUTING↗

Potential of pin-by-pin SPN calculations as an industrial reference

This paper aims at analysing the potential of pin-by-pin SP{sub n} calculations to compute the neutronic flux in PWR cores as an alternative to the diffusion approximation. As far as pin-by-pin calculations are concerned, a SPH equivalence is used to preserve the reactions rates. The use of SPH equivalence is a common practice in core diffusion calculations. In this paper, a methodology to generalize the equivalence procedure in the SP{sub n} equations context is presented. In order to verify and validate the equivalence procedure, SP{sub n} calculations are compared to 2D transport reference results obtained with the APOLL02 code. The validation cases consist in 3x3 analytical assembly color sets involving burn-up heterogeneities, UOX/MOX interfaces, and control rods. Considering various energy discretizations (up to 26 groups) and flux development orders (up to 7) for the SP{sub n} equations, results show that 26-group SP{sub 3} calculations are very close to the transport reference (with pin production rates discrepancies < 1%). This proves the high interest of pin-by-pin SP{sub n} calculations as an industrial reference when relying on 26 energy groups combined with SP{sub 3} flux development order. Additionally, the SP{sub n} results are compared to diffusion pin-by-pin calculations, in order to evaluate the potential benefit of using a SP{sub n} solver as an alternative to diffusion. Discrepancies on pin-production rates are less than 1.6% for 6-group SP{sub 3} calculations against 3.2% for 2-group diffusion calculations. This shows that SP{sub n} solvers may be considered as an alternative to multigroup diffusion. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Two-phase viscoelastic jetting

A coupled finite difference algorithm on rectangular grids is developed for viscoelastic ink ejection simulations. The ink is modeled by the Oldroyd-B viscoelastic fluid model. The coupled algorithm seamlessly incorporates several things: (1) a coupled level set-projection method for incompressible immiscible two-phase fluid flows; (2) a higher-order Godunov type algorithm for the convection terms in the momentum and level set equations; (3) a simple first-order upwind algorithm for the convection term in the viscoelastic stress equations; (4) central difference approximations for viscosity, surface tension, and upper-convected derivative terms; and (5) an equivalent circuit model to calculate the inflow pressure (or flow rate) from dynamic voltage.

97 MATHEMATICS AND COMPUTING↗

The Spacelab Experiment Interface Device (SEID)

The Spacelab experiment interface device (SEID) is described which simulates the electrical, and logical connections of the Spacelab remote acquisition unit (RAU), the interface functions of the Spacelab experiment computer software, and the electrical aspects of the high rate multiplexer. Simulated RAU interfaces include PCM serial channels (up to four), user time clock (UTC), flexible inputs (up to 128) and discrete outputs (up to 64). Connectors are logically compatible with the rau. be instructed to execute sequences of input/output commands to the RAU, similar to those performed during flight. This approximation of the experiment computer software is adequate to detect interface problems and prevent these from occurring during Spacelab payload integration. The SEID hardware is a microprocessor based system, utilizing an 8085 microprocessor with 6K of PROM and 4K of static RAM. The basic unit is augmented with an LSI-11 microcomputer to provide disk storage and a more dynamic environment for generation of control data. A simulation of the general monitoring loop is provided to emulate Spacelab system timing.

Kallus, R.↗

Shape optimization of a sheet swimming over a thin liquid layer

Motivated by the propulsion mechanisms adopted by gastropods, annelids and other invertebrates, we consider shape optimization of a flexible sheet that moves by propagating deformation waves along its body. The self-propelled sheet is separated from a rigid substrate by a thin layer of viscous Newtonian fluid. We use a lubrication approximation to model the dynamics and derive the relevant Euler-Lagrange equations to simultaneously optimize swimming speed, efficiency and fluid loss. We find that as the parameters controlling these quantities approach critical values, the optimal solutions become singular in a self-similar fashion and sometimes leave the realm of validity of the lubrication model. We explore these singular limits by computing higher order corrections to the zeroth order theory and find that wave profiles that develop cusp-like singularities are appropriately penalized, yielding non-singular optimal solutions. These corrections are themselves validated by comparison with finite element solutions of the full Stokes equations, and, to the extent possible, using recent rigorous a-priori error bounds.

97 MATHEMATICS AND COMPUTING↗