Engineering Papers⌕ Search

SEARCH · Engineering Papers

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 55 records · Page 3
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Randomized Sketching Algorithms for Low-Memory Dynamic Optimization

This paper develops a novel limited-memory method to solve dynamic optimization problems. The memory requirements for such problems often present a major obstacle, particularly for problems with PDE constraints such as optimal flow control, full waveform inversion, and optical tomography. In these problems, PDE constraints uniquely determine the state of a physical system for a given control; the goal is to find the value of the control that minimizes an objective. While the control is often low dimensional, the state is typically more expensive to store. This paper suggests using randomized matrix approximation to compress the state as it is generated and shows how to use the compressed state to reliably solve the original dynamic optimization problem. Concretely, the compressed state is used to compute approximate gradients and to apply the Hessian to vectors. The approximation error in these quantities is controlled by the target rank of the sketch. This approximate first- and second-order information can readily be used in any optimization algorithm. As an example, we develop a sketched trust-region method that adaptively chooses the target rank using a posteriori error information and provably converges to a stationary point of the original problem. Numerical experiments with the sketched trust-region method show promising performance on challenging problems such as the optimal control of an advection-reaction-diffusion equation and the optimal control of fluid flow past a cylinder.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

HD-Bind: Encoding of Molecular Structure with Low Precision, Hyperdimensional Binary Representations

Publicly available collections of drug-like molecules have grown to comprise tens of billions of compounds due to advances in combinatorial chemistry. Traditional methods for identifying "hit" molecules from a large collection of potential drug-like candidates have relied on biophysical theory to compute approximations to the Gibbs free energy of the binding interaction between the drug and its protein target. These approaches have the major drawback that they require exceptional computing capabilities for even relatively small collections of molecules. Hyperdimensional Computing (HDC) is a recently-proposed learning paradigm that represents data with high-dimension binary vectors; this allows the use of low-precision binary vector arithmetic to create models of the data that can be learned without the need for the gradient-based optimization required in many conventional machine learning and deep learning methods. This algorithmic simplicity allows for acceleration in hardware that has been previously demonstrated in a range of application areas. We consider existing HDC approaches for molecular property classification and introduce two novel encodings of a commonly-used molecular representation, the extended connectivity fingerprint (ECFP). We show that HDC-based inference methods are as much as 91 times more efficient than traditional machine learning methods, and achieve an acceleration of nearly nine orders of magnitude compared to molecular docking. Our results show that HDC accelerated methods retain competitive accuracy on a number of well-studied tasks such as molecular property predictions using the MoleculeNet dataset, and bind/no-bind activity classification using the DUD-E and LIT-PCBA datasets. Our work thus motivates further investigation into molecular representation learning to develop ultraefficient pre-screening tools.

Jones, William↗
NASA NTRS1990

Verification of floating-point software

Floating point computation presents a number of problems for formal verification. Should one treat the actual details of floating point operations, or accept them as imprecisely defined, or should one ignore round-off error altogether and behave as if floating point operations are perfectly accurate. There is the further problem that a numerical algorithm usually only approximately computes some mathematical function, and we often do not know just how good the approximation is, even in the absence of round-off error. ORA has developed a theory of asymptotic correctness which allows one to verify floating point software with a minimum entanglement in these problems. This theory and its implementation in the Ariel C verification system are described. The theory is illustrated using a simple program which finds a zero of a given function by bisection. This paper is presented in viewgraph form.

Hoover, Doug N.↗
DOE OSTI2024

HDBind: encoding of molecular structure with hyperdimensional binary representations

Traditional methods for identifying “hit” molecules from a large collection of potential drug-like candidates rely on biophysical theory to compute approximations to the Gibbs free energy of the binding interaction between the drug and its protein target. These approaches have a significant limitation in that they require exceptional computing capabilities for even relatively small collections of molecules. Increasingly large and complex state-of-the-art deep learning approaches have gained popularity with the promise to improve the productivity of drug design, notorious for its numerous failures. However, as deep learning models increase in their size and complexity, their acceleration at the hardware level becomes more challenging. Hyperdimensional Computing (HDC) has recently gained attention in the computer hardware community due to its algorithmic simplicity relative to deep learning approaches. The HDC learning paradigm, which represents data with high-dimension binary vectors, allows the use of low-precision binary vector arithmetic to create models of the data that can be learned without the need for the gradient-based optimization required in many conventional machine learning and deep learning methods. This algorithmic simplicity allows for acceleration in hardware that has been previously demonstrated in a range of application areas (computer vision, bioinformatics, mass spectrometery, remote sensing, edge devices, etc.). To the best of our knowledge, our work is the first to consider HDC for the task of fast and efficient screening of modern drug-like compound libraries. We also propose the first HDC graph-based encoding methods for molecular data, demonstrating consistent and substantial improvement over previous work. We compare our approaches to alternative approaches on the well-studied MoleculeNet dataset and the recently proposed LIT-PCBA dataset derived from high quality PubChem assays. We demonstrate our methods on multiple target hardware platforms, including Graphics Processing Units (GPUs) and Field Programmable Gate Arrays (FPGAs), showing at least an order of magnitude improvement in energy efficiency versus even our smallest neural network baseline model with a single hidden layer. Our work thus motivates further investigation into molecular representation learning to develop ultra-efficient pre-screening tools. We make our code publicly available at https://github.com/LLNL/hdbind.

59 BASIC BIOLOGICAL SCIENCES↗
NASA NTRS1976

The role of fracture mechanics in the design of fuel tanks in space vehicles

With special reference to design of fuel tanks in space vehicles, the principles of fracture mechanics are reviewed. An approximate but extremely simple relationship is derived among the operating stress level, the length of crack, and the number of cycles of failure. Any one of the variables can be computed approximately from the knowledge of the other two, if the loading schedule (mission of the tank) is not greatly altered. Two sample examples illustrating the procedures of determining the allowable safe operating stress corresponding to a set of assumed loading schedule are included. The selection of sample examples is limited by the relatively meager available data on the candidate material for various stress ratios in the cycling.

Denton, S. J.↗
NASA NTRS1985

Modelling Equilibrium and Fractional Crystallization in the System MgO-FeO-CaO-Al2O3-SiO2

A mathematical modelling technique for use in petrogenesis calculations in the system MgO-FeO-CaO-Al2O3-SiO2 is reported. Semiempirical phase boundary and elemental distribution information was combined with mass balance to compute approximate equilibrium crystallization paths for arbitrary system compositions. The calculation is applicable to a range of system compositions and fractionation calculations are possible. The goal of the calculation is the computation of the composition and quantity of each phase present as a function of the degree of solidification. The degree of solidification is parameterized by the heat released by the solidifying phases. The mathematical requirement for the solution of this problem is: (1) An equation constraining the composition of the magma for each solid phase in equilibrium with the liquidus phase, and (2) an equation for each solid phase and each component giving the distribution of that element between that phase and the magma.

Herbert, F.↗
NASA NTRS1993

Proceedings of the Third International Workshop on Neural Networks and Fuzzy Logic, volume 2

Papers presented at the Neural Networks and Fuzzy Logic Workshop sponsored by the National Aeronautics and Space Administration and cosponsored by the University of Houston, Clear Lake, held 1-3 Jun. 1992 at the Lyndon B. Johnson Space Center in Houston, Texas are included. During the three days approximately 50 papers were presented. Technical topics addressed included adaptive systems; learning algorithms; network architectures; vision; robotics; neurobiological connections; speech recognition and synthesis; fuzzy set theory and application, control and dynamics processing; space applications; fuzzy logic and neural network computers; approximate reasoning; and multiobject decision making.

Culbert, Christopher J.↗
NASA NTRS1993

Proceedings of the Third International Workshop on Neural Networks and Fuzzy Logic, volume 1

Documented here are papers presented at the Neural Networks and Fuzzy Logic Workshop sponsored by the National Aeronautics and Space Administration and cosponsored by the University of Houston, Clear Lake. The workshop was held June 1-3, 1992 at the Lyndon B. Johnson Space Center in Houston, Texas. During the three days approximately 50 papers were presented. Technical topics addressed included adaptive systems; learning algorithms; network architectures; vision; robotics; neurobiological connections; speech recognition and synthesis; fuzzy set theory and application, control, and dynamics processing; space applications; fuzzy logic and neural network computers; approximate reasoning; and multiobject decision making.

Culbert, Christopher J.↗
NASA NTRS1972

Methods and approximations for the computation of transmission profiles in the v4 band of methane in the atmosphere of Jupiter.

This note discusses the validity of certain band models and scaling approximations for computing transmissions in the v4 band of methane along inhomogeneous paths in the atmosphere of Jupiter. It is shown that Goody's random band model approximates the results of a rigorous numerical line-by-line calculation of the transmission profile of a Jovian model atmosphere.

Taylor, F. W.↗
NASA NTRS1943

Experimental Investigation of Impact in Landing on Water

The extent of agreement of the theoretical impact computations with the actual phenomenon has not as yet been fully clarified. There is on the one hand a certain imperfection in the theory (simplifying assumptions made) and on the other an insufficiency in the experimental data available. The object of our present paper is to show how far test results agree with the available approximate computation methods, to investigate in greater detail the physical nature of impact on water, and to perfect the experimental method of studying the phenomenon.

Kreps, R. L.↗
DOE OSTI2022

Estimates of Quantum Tunneling Effects for Hydrogen Diffusion in PuO 2

We detail the estimation of activation energies and quantum nuclear vibrational tunneling effects for hydrogen diffusion in PuO 2 based on Density Functional Theory calculations and a quantum double well approximation. We find that results are relatively insensitive to choice of exchange correlation functional. In addition, the representation of spin in the system and use of an extended Hubbard U correction has only a small effect on hydrogen point defect formation energies when the PuO 2 lattice is held fixed at the experimental density. We then compute approximate activation energies for transitions between hydrogen interstitial sites seeded by a semi-empirical quantum model and determine the quantum tunneling enhancement relative to classical kinetic rates. Our model indicates that diffusion rates in H/PuO 2 systems could be enhanced by more than one order of magnitude at ambient conditions and that these effects persist at high temperature. The method we propose here can be used as a fast screening tool for assessing possible quantum nuclear vibrational effects in any number of condensed phase materials and surfaces, where hydrogen hopping tends to follow well defined minimum energy pathways.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
NASA NTRS2006

The Effect of Cumulus Cloud Field Anisotropy on Domain-Averaged Solar Fluxes and Atmospheric Heating Rates

Cumulus clouds can become tilted or elongated in the presence of wind shear. Nevertheless, most studies of the interaction of cumulus clouds and radiation have assumed these clouds to be isotropic. This paper describes an investigation of the effect of fair-weather cumulus cloud field anisotropy on domain-averaged solar fluxes and atmospheric heating rate profiles. A stochastic field generation algorithm was used to produce twenty three-dimensional liquid water content fields based on the statistical properties of cloud scenes from a large eddy simulation. Progressively greater degrees of x-z plane tilting and horizontal stretching were imposed on each of these scenes, so that an ensemble of scenes was produced for each level of distortion. The resulting scenes were used as input to a three-dimensional Monte Carlo radiative transfer model. Domain-average transmission, reflection, and absorption of broadband solar radiation were computed for each scene along with the average heating rate profile. Both tilt and horizontal stretching were found to significantly affect calculated fluxes, with the amount and sign of flux differences depending strongly on sun position relative to cloud distortion geometry. The mechanisms by which anisotropy interacts with solar fluxes were investigated by comparisons to independent pixel approximation and tilted independent pixel approximation computations for the same scenes. Cumulus anisotropy was found to most strongly impact solar radiative transfer by changing the effective cloud fraction, i.e., the cloud fraction when the field is projected on a surface perpendicular to the direction of the incident solar beam.

Hinkelman, Laura M.↗
NASA NTRS1986

An approximation formula for a class of fault-tolerant computers

An approximation formula is derived for the probability of failure for fault-tolerant process-control computers. These computers use redundancy and reconfiguration to achieve high reliability. Finite-state Markov models capture the dynamic behavior of component failure and system recovery, and the approximation formula permits an estimation of system reliability by an easy examination of the model.

White, A. L.↗
NASA NTRS1980

Calculation of laminar heating rates on three-dimensional configurations using the axisymmetric analogue

A theoretical method was developed for computing approximate laminar heating rates on three dimensional configurations at angle of attack. The method is based on the axisymmetric analogue which is used to reduce the three dimensional boundary layer equations along surface streamlines to an equivalent axisymmetric form by using the metric coefficient which describes streamline divergence (or convergence). The method was coupled with a three dimensional inviscid flow field program for computing surface streamline paths, metric coefficients, and boundary layer edge conditions.

Hamilton, H. H., II↗
DOE OSTI2026

Near-Optimal Performance of Stochastic Model Predictive Control

Here, this article presents a regret analysis for stochastic model predictive control (SMPC) in linear systems with quadratic performance index and additive and multiplicative uncertainties. Under a finite support assumption, the problem can be cast as a finite-dimensional quadratic program, but the problem becomes quickly intractable as the problem size grows exponentially in the horizon length. SMPC aims to compute approximate solutions by solving a sequence of problems with truncated prediction horizons and committing the solution in a receding-horizon fashion. Although this approach is widely used in practice, its performance relative to the optimal solution is not well understood. This article reports for the first time a rigorous near-optimal performance guarantee of SMPC: under stabilizability and detectability conditions, the regret of SMPC is exponentially small in the prediction horizon length, allowing SMPC to achieve near-optimal performance at a substantially reduced computational expense.

93E20, 93B45↗
DOE OSTI2024

Infinite quantum signal processing

Quantum signal processing (QSP) represents a real scalar polynomial of degree d using a product of unitary matrices of size 2 × 2 , parameterized by ( d + 1 ) real numbers called the phase factors. This innovative representation of polynomials has a wide range of applications in quantum computation. When the polynomial of interest is obtained by truncating an infinite polynomial series, a natural question is whether the phase factors have a well defined limit as the degree d → ∞ . While the phase factors are generally not unique, we find that there exists a consistent choice of parameterization so that the limit is well defined in the ℓ 1 space. This generalization of QSP, called the infinite quantum signal processing, can be used to represent a large class of non-polynomial functions. Our analysis reveals a surprising connection between the regularity of the target function and the decay properties of the phase factors. Our analysis also inspires a very simple and efficient algorithm to approximately compute the phase factors in the ℓ 1 space. The algorithm uses only double precision arithmetic operations, and provably converges when the ℓ 1 norm of the Chebyshev coefficients of the target function is upper bounded by a constant that is independent of d . This is also the first numerically stable algorithm for finding phase factors with provable performance guarantees in the limit d → ∞ .

Dong, Yulong [Department of Mathematics, Universit↗
DOE OSTI2024

MCNPs Easy Sources for (α,n) (MESA):Verification with Sources4c

MESA is intended to be a direct re-implementation of the algorithms documented in the Sources manual in C++. However, some fundamental changes to the computational structure and logic were made to avoid the use of goto, require consistency in input definition, and reduce computational approximations in later steps of three-layer problems. In addition, MESA uses α-decay information from ISC libraries (based on ENDF). In most cases, differences between Sources4c and MESA are dominated by differences in decay energy spectra and intensities. To verify this we have reproduced 6 problems that are documented examples or samples in Sources4c. These problems cover the three types of problems MESA currently supports: homogeneous, interface, and three-layer.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗