SEARCH · Search NASA

Results for “algorithmic differentiation”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 829 records · Page 46

All sources NASA NTRS DOE OSTI

DOE OSTI2017

Speeding up N -body simulations of modified gravity: chameleon screening models

We describe and demonstrate the potential of a new and very efficient method for simulating certain classes of modified gravity theories, such as the widely studied f ( R ) gravity models. High resolution simulations for such models are currently very slow due to the highly nonlinear partial differential equation that needs to be solved exactly to predict the modified gravitational force. This nonlinearity is partly inherent, but is also exacerbated by the specific numerical algorithm used, which employs a variable redefinition to prevent numerical instabilities. The standard Newton-Gauss-Seidel iterative method used to tackle this problem has a poor convergence rate. Our new method not only avoids this, but also allows the discretised equation to be written in a form that is analytically solvable. We show that this new method greatly improves the performance and efficiency of f ( R ) simulations. For example, a test simulation with 512{sup 3} particles in a box of size 512 Mpc/ h is now 5 times faster than before, while a Millennium-resolution simulation for f ( R ) gravity is estimated to be more than 20 times faster than with the old method. Our new implementation will be particularly useful for running very high resolution, large-sized simulations which, to date, are only possible for the standard model, and also makes it feasible to run large numbers of lower resolution simulations for covariance analyses. We hope that the method will bring us to a new era for precision cosmological tests of gravity.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2019

The Monarch Initiative in 2019: an integrative data and analytic platform connecting phenotypes to genotypes across species

In biology and biomedicine, relating phenotypic outcomes with genetic variation and environmental factors remains a challenge: patient phenotypes may not match known diseases, candidate variants may be in genes that haven’t been characterized, research organisms may not recapitulate human or veterinary diseases, environmental factors affecting disease outcomes are unknown or undocumented, and many resources must be queried to find potentially significant phenotypic associations. The Monarch Initiative (https://monarchinitiative.org) integrates information on genes, variants, genotypes, phenotypes and diseases in a variety of species, and allows powerful ontology-based search. We develop many widely adopted ontologies that together enable sophisticated computational analysis, mechanistic discovery and diagnostics of Mendelian diseases. Our algorithms and tools are widely used to identify animal models of human disease through phenotypic similarity, for differential diagnostics and to facilitate translational research. Launched in 2015, Monarch has grown with regards to data (new organisms, more sources, better modeling); new API and standards; ontologies (new Mondo unified disease ontology, improvements to ontologies such as HPO and uPheno); user interface (a redesigned website); and community development. Monarch data, algorithms and tools are being used and extended by resources such as GA4GH and NCATS Translator, among others, to aid mechanistic discovery and diagnostics.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2015

The additive volume density law implemented in the Latec workbench to deal with fissile media that contain moderator elements - 14342

The moderation ratio is one of the key parameters for safety assessment in criticality-safety studies. It can be expressed as a moderator to fissile atoms ratio and is related to the concentration of fissile isotopes. The relationship linking these parameters illustrates a physical behaviour of the mixture, which could be very complex depending on the different kinds of media and their associated physical states encountered in the nuclear fuel cycle. To guarantee the safety when modelling a given mixture, criticality-safety practitioners consider either a best estimate approach by fitting experimental data, or, a bounding approach with a fictitious fissile mixture. This theoretical model is generally more penalizing in terms of k{sub eff} and does not involve numeric algorithms for the resolution reducing the overall uncertainty. This is the case for the additive volume density law, which is based on the assumption that the volume of a mixture is the sum of the volumes of all its compounds. This naturally leads to differentiate all the compounds in the mixture and, in particular, the fissile compound and the moderator. A first algorithm has been derived from this approach and implemented in the LATEC criticality workbench developed at IRSN, but a second assumption is required to solve the problem encountered when moderator's elements have also to be used to define the fissile compound. A typical example of this major issue is an UZrH fissile compound moderated by water. The reference moderator element of the water is obviously hydrogen that belongs to the fissile compound. This paper is dedicated to the implementation in the LATEC criticality workbench of an algorithm based on the additive volume density law that is able to deal with such a configuration. The major advantage of this new algorithm is the versatility in terms of input data. The mixture compounds and the moderation can be expressed in multiple ways still using the same algorithm. A validation work has been performed on different UZrH compounds and for different expressions of moderation. The calculated isotopic concentrations have been compared to theoretical calculations using a more straightforward algorithm but that should be adapted for every change in terms of moderation expression. The results obtained are in good agreement with validation results. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2012

MODELING COLLISIONAL CASCADES IN DEBRIS DISKS: THE NUMERICAL METHOD

We develop a new numerical algorithm to model collisional cascades in debris disks. Because of the large dynamical range in particle masses, we solve the integro-differential equations describing erosive and catastrophic collisions in a particle-in-a-box approach, while treating the orbital dynamics of the particles in an approximate fashion. We employ a new scheme for describing erosive (cratering) collisions that yields a continuous set of outcomes as a function of colliding masses. We demonstrate the stability and convergence characteristics of our algorithm and compare it with other treatments. We show that incorporating the effects of erosive collisions results in a decay of the particle distribution that is significantly faster than with purely catastrophic collisions.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2015

Comparison of the kinetics of different Markov models for ligand binding under varying conditions

We recently derived a Markov model for macromolecular ligand binding dynamics from few physical assumptions and showed that its stationary distribution is the grand canonical ensemble [J. W. R. Martini, M. Habeck, and M. Schlather, J. Math. Chem. 52, 665 (2014)]. The transition probabilities of the proposed Markov process define a particular Glauber dynamics and have some similarity to the Metropolis-Hastings algorithm. Here, we illustrate that this model is the stochastic analog of (pseudo) rate equations and the corresponding system of differential equations. Moreover, it can be viewed as a limiting case of general stochastic simulations of chemical kinetics. Thus, the model links stochastic and deterministic approaches as well as kinetics and equilibrium described by the grand canonical ensemble. We demonstrate that the family of transition matrices of our model, parameterized by temperature and ligand activity, generates ligand binding kinetics that respond to changes in these parameters in a qualitatively similar way as experimentally observed kinetics. In contrast, neither the Metropolis-Hastings algorithm nor the Glauber heat bath reflects changes in the external conditions correctly. Both converge rapidly to the stationary distribution, which is advantageous when the major interest is in the equilibrium state, but fail to describe the kinetics of ligand binding realistically. To simulate cellular processes that involve the reversible stochastic binding of multiple factors, our pseudo rate equation model should therefore be preferred to the Metropolis-Hastings algorithm and the Glauber heat bath, if the stationary distribution is not of only interest.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

NASA NTRS2013

NASA Tech Briefs, March 2013

Topics covered include: Remote Data Access with IDL Data Compression Algorithm Architecture for Large Depth-of-Field Particle Image Velocimeters Vectorized Rebinning Algorithm for Fast Data Down-Sampling Display Provides Pilots with Real-Time Sonic-Boom Information Onboard Algorithms for Data Prioritization and Summarization of Aerial Imagery Monitoring and Acquisition Real-time System (MARS) Analog Signal Correlating Using an Analog-Based Signal Conditioning Front End Micro-Textured Black Silicon Wick for Silicon Heat Pipe Array Robust Multivariable Optimization and Performance Simulation for ASIC Design; Castable Amorphous Metal Mirrors and Mirror Assemblies; Sandwich Core Heat-Pipe Radiator for Power and Propulsion Systems; Apparatus for Pumping a Fluid; Cobra Fiber-Optic Positioner Upgrade; Improved Wide Operating Temperature Range of Li-Ion Cells; Non-Toxic, Non-Flammable, -80 C Phase Change Materials; Soft-Bake Purification of SWCNTs Produced by Pulsed Laser Vaporization; Improved Cell Culture Method for Growing Contracting Skeletal Muscle Models; Hand-Based Biometric Analysis; The Next Generation of Cold Immersion Dry Suit Design Evolution for Hypothermia Prevention; Integrated Lunar Information Architecture for Decision Support Version 3.0 (ILIADS 3.0); Relay Forward-Link File Management Services (MaROS Phase 2); Two Mechanisms to Avoid Control Conflicts Resulting from Uncoordinated Intent; XTCE GOVSAT Tool Suite 1.0; Determining Temperature Differential to Prevent Hardware Cross-Contamination in a Vacuum Chamber; SequenceL: Automated Parallel Algorithms Derived from CSP-NT Computational Laws; Remote Data Exploration with the Interactive Data Language (IDL); Mixture-Tuned, Clutter Matched Filter for Remote Detection of Subpixel Spectral Signals; Partitioned-Interval Quantum Optical Communications Receiver; and Practical UAV Optical Sensor Bench with Minimal Adjustability.

Source record↗

DOE OSTI2022

A phase transition for finding needles in nonlinear haystacks with LASSO artificial neural networks

To fit sparse linear associations, a LASSO sparsity inducing penalty with a single hyperparameter provably allows to recover the important features (needles) with high probability in certain regimes even if the sample size is smaller than the dimension of the input vector (haystack). More recently learners known as artificial neural networks (ANN) have shown great successes in many machine learning tasks, in particular fitting nonlinear associations. Small learning rate, stochastic gradient descent algorithm and large training set help to cope with the explosion in the number of parameters present in deep neural networks. Yet few ANN learners have been developed and studied to find needles in nonlinear haystacks. Driven by a single hyperparameter, our ANN learner, like for sparse linear associations, exhibits a phase transition in the probability of retrieving the needles, which we do not observe with other ANN learners. To select our penalty parameter, we generalize the universal threshold of Donoho and Johnstone (Biometrika 81(3):425–455, 1994) which is a better rule than the conservative (too many false detections) and expensive cross-validation. In the spirit of simulated annealing, we propose a warm-start sparsity inducing algorithm to solve the high-dimensional, non-convex and non-differentiable optimization problem. We perform simulated and real data Monte Carlo experiments to quantify the effectiveness of our approach.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Single-stage gradient-based stellarator coil design: Optimization for near-axis quasi-symmetry

Here we present a new coil design paradigm for magnetic confinement in stellarators. Our approach directly optimizes coil shapes and coil currents to produce a vacuum quasi-symmetric magnetic field with a target rotational transform on the magnetic axis. This approach differs from the traditional two-stage approach in which first a magnetic configuration with desirable physics properties is found, and then coils to approximately realize this magnetic configuration are designed. The proposed single-stage approach allows us to find a compromise between confinement and engineering requirements, i.e., find easy-to-build coils with good confinement properties. Using forward and adjoint sensitivities, we derive derivatives of the physical quantities in the objective, which is constrained by a nonlinear periodic differential equation. In two numerical examples, we compare different gradient-based descent algorithms and find that incorporating approximate second-order derivative information through a quasi-Newton method is crucial for convergence. We also explore the optimization landscape in the neighborhood of a minimizer and find many directions in which the objective is mostly flat, indicating ample freedom to find simple and thus easy-to-build coils.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Partial differential equations preconditioner resilient to soft and hard faults

We present a domain-decomposition-based preconditioner for the solution of partial differential equations (PDEs) that is resilient to both soft and hard faults. The algorithm reformulates the PDE as a sampling problem, followed by a solution update through data manipulation that is resilient to both soft and hard faults. This reformulation allows us to recast the problem as a set of independent tasks, and exploit data locality to reduce global communication. We discuss two different parallel implementations: (a) a single program multiple data (SPMD) version based on a one-to-one mapping between subdomain and MPI processes responsible for both state and computation; and (b) an asynchronous server–client implementation where all state information is held by the servers and clients are designed solely as computational units. We present a scalability comparison of both implementations under nominal conditions, showing efficiency within ~80% for up to 12,000 cores. We present a resilience analysis under different fault scenarios based on the server–client implementation. This framework provides resiliency to hard faults such that if a client crashes, it stops asking for work, and the servers simply distribute the work among all of the other clients alive. Erroneous subdomain solves (e.g. due to soft faults) appear as corrupted data, which is either rejected if that causes a task to fail, or is seamlessly filtered out during the regression stage through a suitable noise model. Three different types of faults are modeled: hard faults modeling nodes (or clients) crashing; soft faults occurring during the communication of the tasks between server and clients; and soft faults occurring during task execution. We demonstrate the resiliency of the approach for a 2D elliptic PDE, and explore the effect of the faults at various failure rates.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Regularized inversion of aerosol hygroscopic growth factor probability density function: application to humidity-controlled fast integrated mobility spectrometer measurements

Abstract. Aerosol hygroscopic growth plays an important role in atmospheric particle chemistry and the effects of aerosol on radiation and hence climate. The hygroscopic growth is often characterized by a growth factor probability density function (GF-PDF), where the growth factor is defined as the ratio of the particle size at a specified relative humidity to its dry size. Parametric, least-squares methods are the most widely used algorithms for inverting the GF-PDF from measurements of the humidified tandem differential mobility analyzer (HTDMA) and have been recently applied to the GF-PDF inversion from measurements of the humidity-controlled fast integrated mobility spectrometer (HFIMS). However, these least-squares methods suffer from noise amplification due to the lack of regularization in solving the ill-posed problem, resulting in significant fluctuations in the retrieved GF-PDF and even occasional failures of convergence. In this study, we introduce nonparametric, regularized methods to invert the aerosol GF-PDF and apply them to HFIMS measurements. Based on the HFIMS kernel function, the forward convolution is transformed into a matrix-based form, which facilitates the application of the nonparametric inversion methods with regularizations, including Tikhonov regularization and Twomey's iterative regularization. Inversions of the GF-PDF using the nonparameteric methods with regularization are demonstrated using HFIMS measurements simulated from representative GF-PDFs of ambient aerosols. The characteristics of reconstructed GF-PDFs resulting from different inversion methods, including previously developed least-squares methods, are quantitatively compared. The result shows that Twomey's method generally outperforms other inversion methods. The capabilities of Twomey's method in reconstructing the pre-defined GF-PDFs and recovering the mode parameters are validated.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1974

Changing stepsize in the integration of differential equations using modified divided differences

Multistep methods for solving differential equations based on numerical integration formulas or numerical differentiation formulas (for stiff equations) require special provision for changing the stepsize. New algorithms are given which make the use of modified divided differences an attractive way to carry out the change in stepsize for such methods. Error estimation and some of the important factors in stepsize selection and the selection of integration order are also considered.

Krogh, F. T.↗

NASA NTRS1986

Physical and numerical sources of computational inefficiency in integration of chemical kinetic rate equations: Etiology, treatment and prognosis

The design of a very fast, automatic black-box code for homogeneous, gas-phase chemical kinetics problems requires an understanding of the physical and numerical sources of computational inefficiency. Some major sources reviewed in this report are stiffness of the governing ordinary differential equations (ODE's) and its detection, choice of appropriate method (i.e., integration algorithm plus step-size control strategy), nonphysical initial conditions, and too frequent evaluation of thermochemical and kinetic properties. Specific techniques are recommended (and some advised against) for improving or overcoming the identified problem areas. It is argued that, because reactive species increase exponentially with time during induction, and all species exhibit asymptotic, exponential decay with time during equilibration, exponential-fitted integration algorithms are inherently more accurate for kinetics modeling than classical, polynomial-interpolant methods for the same computational work. But current codes using the exponential-fitted method lack the sophisticated stepsize-control logic of existing black-box ODE solver codes, such as EPISODE and LSODE. The ultimate chemical kinetics code does not exist yet, but the general characteristics of such a code are becoming apparent.

Pratt, D. T.↗

NASA NTRS1987

Optimal cure cycle design of a resin-fiber composite laminate

A unified computed aided design method was studied for the cure cycle design that incorporates an optimal design technique with the analytical model of a composite cure process. The preliminary results of using this proposed method for optimal cure cycle design are reported and discussed. The cure process of interest is the compression molding of a polyester which is described by a diffusion reaction system. The finite element method is employed to convert the initial boundary value problem into a set of first order differential equations which are solved simultaneously by the DE program. The equations for thermal design sensitivities are derived by using the direct differentiation method and are solved by the DE program. A recursive quadratic programming algorithm with an active set strategy called a linearization method is used to optimally design the cure cycle, subjected to the given design performance requirements. The difficulty of casting the cure cycle design process into a proper mathematical form is recognized. Various optimal design problems are formulated to address theses aspects. The optimal solutions of these formulations are compared and discussed.

Hou, Jean W.↗

NASA NTRS1986

Numerical prediction of cold turbulent flow in combustor configurations with different centerbody flame holders

A numerical study of cold turbulent flow in a combustor containing a centerbody flame holder is presented. The axisymmetric Navier-Stokes equations incorporating a k-epsilon turbulence model were solved in a nonorthogonal curvilinear coordinate system. The finite volume method, applied to a staggered grid system, was used to discretize the differential equations. These finite differenced equations were then solved iteratively utilizing the SIMPLE algorithm described by Patankar (1980). Solutions were obtained for two combustor duct geometries with various centerbody flame holders which included a disk, a cone, and a sphere. The extent of mixing due to these bodies was evaluated. A comparison with previously obtained experimental data yielded moderately good agreement.

Yung, C. N.↗

NASA NTRS1991

Video data compression using artificial neural network differential vector quantization

An artificial neural network vector quantizer is developed for use in data compression applications such as Digital Video. Differential Vector Quantization is used to preserve edge features, and a new adaptive algorithm, known as Frequency-Sensitive Competitive Learning, is used to develop the vector quantizer codebook. To develop real time performance, a custom Very Large Scale Integration Application Specific Integrated Circuit (VLSI ASIC) is being developed to realize the associative memory functions needed in the vector quantization algorithm. By using vector quantization, the need for Huffman coding can be eliminated, resulting in superior performance against channel bit errors than methods that use variable length codes.

Krishnamurthy, Ashok K.↗

NASA NTRS1996

Numerical Algorithms Based on Biorthogonal Wavelets

Wavelet bases are used to generate spaces of approximation for the resolution of bidimensional elliptic and parabolic problems. Under some specific hypotheses relating the properties of the wavelets to the order of the involved operators, it is shown that an approximate solution can be built. This approximation is then stable and converges towards the exact solution. It is designed such that fast algorithms involving biorthogonal multi resolution analyses can be used to resolve the corresponding numerical problems. Detailed algorithms are provided as well as the results of numerical tests on partial differential equations defined on the bidimensional torus.

Ponenti, Pj.↗

NASA NTRS1996

Numerical Algorithms Based on Biorthogonal Wavelets

Ponenti, Pj.↗

NASA NTRS2009

Radar Differential Phase Signatures of Ice Orientation for the Prediction of Lightning Initiation and Cessation

The majority of lightning-related casualties typically occur during thunderstorm initiation (e.g., first flash) or dissipation (e.g., last flash). The physics of electrification and lightning production during thunderstorm initiation is fairly well understood. As such, the literature includes a number of studies presenting various radar techniques (using reflectivity and, if available, other dual-polarimetric parameters) for the anticipation of initial electrification and first lightning flash. These radar techniques have shown considerable skill at forecasting first flash. On the other hand, electrical processes and lightning production during thunderstorm dissipation are not nearly as well understood and few, if any, successful techniques have been developed to anticipate the last flash and subsequent cessation of lightning. One promising approach involves the use of dual-polarimetric radar variables to infer the presence of oriented ice crystals in lightning producing storms. In the absence of strong vertical electric fields, ice crystals fall with their largest (semi-major) axis in the horizontal associated with gravitational and aerodynamic forces. In thunderstorms, strong vertical electric fields (100-200 kV m(sup -1)) have been shown to orient small (less than 2 mm) ice crystals such that their semi-major axis is vertical (or nearly vertical). After a lightning flash, the electric field is typically relaxed and prior radar research suggests that ice crystals rapidly resume their preferred horizontal orientation. In active thunderstorms, the vertical electric field quickly recovers and the ice crystals repeat this cycle of orientation for each nearby flash. This change in ice crystal orientation from primarily horizontal to vertical during the development of strong vertical electric fields prior to a lightning flash forms the physical basis for anticipating lightning initiation and, potentially, cessation. Research has shown that radar reflectivity (Z) and other co-polar back-scattering radar measurements like differential reflectivity (Z(sub dr)) typically measured by operational dual-polarimetric radars are not sensitive to these changes in ice crystal orientation. However, prior research has demonstrated that oriented ice crystals cause significant propagation effects that can be routinely measured by most dual-polarimetric radars from X-band (3 cm) to S-band (10 cm) wavelengths using the differential propagation phase shift (often just called differential phase, phi(sub dp)) or its range derivative, the specific differential phase (K(sub dp)). Advantages of the differential phase include independence from absolute or relative power calibration, attenuation, differential attenuation and relative insensitivity to ground clutter and partial beam occultation effects (as long as the signal remains above noise). In research mode, these sorts of techniques have been used to anticipate initial cloud electrification, lightning initiation, and cessation. In this study, we develop a simplified model of ice crystal size, shape, orientation, dielectric, and associated radar scattering and propagation effects in order to simulate various idealized scenarios of ice crystals responding to a hypothetical electric field and their dual-polarimetric radar signatures leading up to lightning initiation and particularly cessation. The sensitivity of the K(sub dp) ice orientation signature to various ice properties and radar wavelength will be explored. Since K(sub dp) is proportional to frequency in the Rayleigh- Gans scattering regime, the ice orientation signatures should be more obvious at higher (lower) frequencies (wavelengths). As a result, simulations at radar wavelengths from 10 cm down to 1 cm (Ka-band) will be conducted. Resonance effects will be considered using the T-matrix method. Since most K(sub dp) Vbased observations have been shown at S-band, we will present ice orientation signatures from C-band (UAH/NASA ARMOR) and X-bd (UAH MAX) dual-polarimetric radars located in Northern Alabama. Issues related to optimal radar scanning for the detection of oriented ice will be discussed. Preliminary suggestions on how these differential phase signatures of oriented ice could contribute to lightning initiation and cessation algorithms will be presented.

Carey, L.D.↗