Search NASA⌕ Search

SEARCH · Search NASA

Results for “gradient methods”

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 667 records · Page 37
All sourcesNASA NTRSDOE OSTI
NASA NTRS1987

Solving Finite-Element Problems on a Concurrent Processor

By use of "conjugate gradients" technique, concurrent efficiency greater than 90 percent. Algorithm applies method of conjugate gradients to iterative solution of finite-element problems on concurrent processor. With algorithm, iteration rates nearly proportional to number of processors. For sufficiently large problems, fraction of proportional speedup achieved, called concurrent efficiency, exceeds 90 percent. Results indicate future application of this and related algorithms to large finite-element problems depend primarily upon applicability of iteratiave techniques, not upon issues of concurrency or efficiency.

Lyzenga, G. A.↗
DOE OSTI2014

Accurate gradient approximation for complex interface problems in 3D by an improved coupling interface method

Most elliptic interface solvers become complicated for complex interface problems at those “exceptional points” where there are not enough neighboring interior points for high order interpolation. Such complication increases especially in three dimensions. Usually, the solvers are thus reduced to low order accuracy. In this paper, we classify these exceptional points and propose two recipes to maintain order of accuracy there, aiming at improving the previous coupling interface method [26]. Yet the idea is also applicable to other interface solvers. The main idea is to have at least first order approximations for second order derivatives at those exceptional points. Recipe 1 is to use the finite difference approximation for the second order derivatives at a nearby interior grid point, whenever this is possible. Recipe 2 is to flip domain signatures and introduce a ghost state so that a second-order method can be applied. This ghost state is a smooth extension of the solution at the exceptional point from the other side of the interface. The original state is recovered by a post-processing using nearby states and jump conditions. The choice of recipes is determined by a classification scheme of the exceptional points. The method renders the solution and its gradient uniformly second-order accurate in the entire computed domain. Numerical examples are provided to illustrate the second order accuracy of the presently proposed method in approximating the gradients of the original states for some complex interfaces which we had tested previous in two and three dimensions, and a real molecule ( (1D63)) which is double-helix shape and composed of hundreds of atoms.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Compositionally gradient nickel-rich cathode materials and methods for the manufacture thereof

The present invention relates to lithium rechargeable battery cathode materials. More specifically, the cathode materials are compositionally gradient nickel-rich cathode materials produced using single-source composite precursor materials containing inorganic and/or metalorganic salts of lithium, nickel, manganese, and cobalt. Methods and systems for manufacturing the cathode materials by a combined spray pyrolysis/fluidized bed process are also disclosed.

Kodenkandath, Thomas A.↗
DOE OSTI2015

SU-E-T-780: Use Robustness Optimization (RO) Method to Improve the Planning Efficiency for Pencil Beam Scanning Cranial Spinal Irradiation

Purpose: We evaluate the feasibility of using robustness optimization (RO) function to improve the planning efficiency of pencil beam scanning (PBS) craniospinal irradiation (CSI) with gradient matching technique. Methods: A CSI patient was planned with 2 lateral brain fields and 4 posterior fields to cover the entire spine to maximal field of 24 cm × 20 cm on a compact PBS gantry, ProteusONE. CSI plans were generated using traditional volumetric gradient dose optimization (VGDO) and robustness optimization (RO) method respectively. In traditional VGDO, besides the sectioned spine target volumes, gradient volume (GV) were generated as 4 equally spaced structures e.g. 80%, 60%, 40%, and 20% of prescription dose. In RO method, only sectioned spine target volumes with an overlap of 4cm were created. In the robustness optimization settings, 5mm uncertainty in superior and inferior direction was defined for auto gradient optimization. Dosimetric metrics of conformity number (CN), homogeneity index (HI), and maximal cord doses were compared in Raystation version 4.6.100.6. Results: In VGDO method, total 16 GV structures and five 100% dose level target structures were contoured compared to total 5 target structures in RO method which saves 30 min in contour. With the same PTV coverage (95% volume receive 30.6Gy prescription dose), maximum cord dose is 32.64Gy in VGDO and 31.94Gy in RO. HI is 1.03 and 1.04 for VGDO and RO respectively. CN is 0.93 and 0.94 for VGDO and RO respectively. Conclusions: The dosimetric comparison demonstrated both methods are equivalent in terms of plan quality. With robust optimization for CSI gradient matching, it efficiently reduces the amount of planning target contour structure by factor of 4 and thus improves the planning efficiency especially for 4 or more gradient junction area.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Multi-variance replica exchange SGMCMC for inverse and forward problems via Bayesian PINN

Physics-informed neural network (PINN) has been successfully applied in solving a variety of nonlinear non-convex forward and inverse problems. However, the training is challenging because of the non-convex loss functions and the multiple optima in the Bayesian inverse problem. In this work, we propose a multi-variance replica exchange stochastic gradient Langevin dynamics method to tackle the challenge of the multiple local optima in the optimization and the challenge of the multiple modal posterior distribution in the inverse problem. Replica exchange methods are capable of escaping from the local traps and accelerating the convergence; two chains with different temperatures are designed where the low temperature chain aims for the local convergence, and the target of the high temperature chain is to travel globally and explore the whole loss function entropy landscape. However, it may not be efficient to solve mathematical inversion problems by using the vanilla replica method directly since the method doubles the computational cost in evaluating the forward solvers (likelihood functions) in the two chains. To address this issue, we propose to make different assumptions on the energy function estimation and this facilities one to use solvers of different fidelities in the likelihood function evaluation. More precisely, one can use a solver with low fidelity in the high temperature chain while using a solver with high fidelity in the low temperature chain. Our proposed method significantly lowers the computational cost in the high temperature chain, meanwhile preserving the accuracy and converging very fast. Here we give an unbiased estimate of the swapping rate and give an estimation of the discretization error of the scheme. To verify our idea, we design and solve four inverse problems which have multiple modes. The proposed method is also employed to train the Bayesian PINN to solve the forward and inverse problems; faster and more accurate convergence has been observed when compared to the stochastic gradient Langevin dynamics (SGLD) method and vanilla replica exchange methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2011

Application of the Differential Evolution Method to Solving Inverse Transport Problems

The differential evolution method, a powerful stochastic optimization algorithm that mimics the process of evolution in nature, is applied to inverse transport problems with several unknown parameters of mixed types, including interface location identification, source composition identification, and material mass density identification, in spherical and cylindrical radioactive source/shield systems. In spherical systems, measurements of leakages of discrete gamma-ray lines are assumed, while in cylindrical systems, measurements of scalar fluxes of discrete lines at points outside the system are assumed. The performance of the differential evolution algorithm is compared to the Levenberg-Marquardt method, a standard gradient-based technique, and the covariance matrix adaptation evolution strategy, another stochastic technique, on a variety of numerical test problems with several (i.e., three or more) unknown parameters. Numerical results indicate that differential evolution is the most adept method for finding the global optimum for these problems. In spherical geometry, differential evolution implemented serially is run-time competitive with gradient-based methods, while a parallel version of differential evolution would be run-time competitive with gradient-based techniques in cylindrical geometry. A hybrid differential evolution/Levenberg-Marquardt method is also introduced, and numerical results indicate that it can be a fast and robust optimizer for inverse transport problems.

99 GENERAL AND MISCELLANEOUS↗
NASA NTRS1976

A turbulence model for nonequilibrium adverse pressure gradient flows

This paper presents a method for calculating compressible adverse pressure gradient boundary layers by using an algebraic eddy viscosity turbulence model that has been modified for variable pressure gradient and turbulence memory effects. The pressure gradient corrections are based on previous incompressible data correlations. Several methods for including the effects of turbulence memory are evaluated. A new lag model, which gives good agreement with available experimental data, is developed. Finally, a correlation is developed for the lag length parameter employed in the model as a function of the known experimental flow variables.

Horstman, C. C.↗
DOE OSTI2021

Laplacian Smoothing Stochastic Gradient Markov Chain Monte Carlo

As an important Markov chain Monte Carlo (MCMC) method, the stochastic gradient Langevin dynamics (SGLD) algorithm has achieved great success in Bayesian learning and posterior sampling. Furthermore, SGLD typically suffers from a slow convergence rate due to its large variance caused by the stochastic gradient. In order to alleviate these drawbacks, we leverage the recently developed Laplacian smoothing technique and propose a Laplacian smoothing stochastic gradient Langevin dynamics (LS-SGLD) algorithm. We prove that for sampling from both log-concave and non-log-concave densities, LS-SGLD achieves strictly smaller discretization error in 2-Wasserstein distance, although its mixing rate can be slightly slower. Experiments on both synthetic and real datasets verify our theoretical results and demonstrate the superior performance of LS-SGLD on different machine learning tasks including posterior sampling, Bayesian logistic regression, and training Bayesian convolutional neural networks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

A learned closure method applied to phase mixing in a turbulent gradient-driven gyrokinetic system in simple geometry

We present a new method for formulating closures that learn from kinetic simulation data. We apply this method to phase mixing in a simple gyrokinetic turbulent system – temperature-gradient-driven turbulence in an unsheared slab. The closure, called the learned multi-mode (LMM) closure, is constructed by, first, extracting an optimal basis from a nonlinear kinetic simulation using singular value decomposition. Subsequent nonlinear fluid simulations are projected onto this basis and the results are used to formulate the closure. We compare the closure with other closures schemes over a broad range of the relevant two-dimensional parameter space (collisionality and gradient drive). We find that the turbulent kinetic system produces phase-mixing rates much lower than the linear expectations, which the LMM closure is capable of capturing. We also compare radial heat fluxes. A Hammett–Perkins closure, generalized to include collisional effects, is quite successful throughout the parameter space, producing ${\sim }14\,\%$ root-mean-square (r.m.s.) error. The LMM closure is also very effective: when trained at three (two) points (in a 35 point parameter grid), the LMM closure produces $8\,\%$ ( $12\,\%$ ) r.m.s. errors. The LMM procedure can be readily generalized to other closure problems.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

A deterministic gradient-based approach to avoid saddle points

Abstract Loss functions with a large number of saddle points are one of the major obstacles for training modern machine learning (ML) models efficiently. First-order methods such as gradient descent (GD) are usually the methods of choice for training ML models. However, these methods converge to saddle points for certain choices of initial guesses. In this paper, we propose a modification of the recently proposed Laplacian smoothing gradient descent (LSGD) [Osher et al., arXiv:1806.06317 ], called modified LSGD (mLSGD), and demonstrate its potential to avoid saddle points without sacrificing the convergence rate. Our analysis is based on the attraction region, formed by all starting points for which the considered numerical scheme converges to a saddle point. We investigate the attraction region’s dimension both analytically and numerically. For a canonical class of quadratic functions, we show that the dimension of the attraction region for mLSGD is $\lfloor (n-1)/2\rfloor$ , and hence it is significantly smaller than that of GD whose dimension is $n-1$ .

Mathematics↗
DOE OSTI2023

Adaptive sampling quasi-Newton methods for zeroth-order stochastic optimization

Here, we consider unconstrained stochastic optimization problems with no available gradient information. Such problems arise in settings from derivative-free simulation optimization to reinforcement learning. We propose an adaptive sampling quasi-Newton method where we estimate the gradients using finite differences of stochastic function evaluations within a common random number framework. We develop modified versions of a norm test and an inner product quasi-Newton test to control the sample sizes used in the stochastic approximations and provide global convergence results to the neighborhood of a locally optimal solution. We present numerical experiments on simulation optimization problems to illustrate the performance of the proposed algorithm. When compared with classical zeroth-order stochastic gradient methods, we observe that our strategies of adapting the sample sizes significantly improve performance in terms of the number of stochastic function evaluations required.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2014

Demonstrating the Temperature Gradient Impact on Grain Growth in UO2 Using the Phase Field Method

Grain boundaries (GBs) are driven to migrate up a temperature gradient. In this work, we use a phase field (PF) model to investigate the impact of temperature gradients on normal grain growth. GB motion in 2D UO2 polycrystals is predicted under increasing temperature gradients. We find that the temperature gradient does not significantly impact the average grain growth behavior, because the curvature driving force is dominant. However, it does cause significant local migration of the individual grains. In addition, the change in the GB mobility due to the temperature gradient results in larger grains in the hot portion of the polycrystal.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2020

Finding wombling boundaries in LHC data with Voronoi and Delaunay tessellations

We address the problem of finding a wombling boundary in point data generated by a general Poisson point process, a specific example of which is an LHC event sample distributed in the phase space of a final state signature, with the wombling boundary created by some new physics. We discuss the use of Voronoi and Delaunay tessellations of the point data for estimating the local gradients and investigate methods for sharpening the boundaries by reducing the statistical noise. The outcome from traditional wombling algorithms is a set of boundary cell candidates with relatively large gradients, whose spatial properties must then be scrutinized in order to construct the boundary and evaluate its significance. Here we propose an alternative approach where we simultaneously form and evaluate the significance of all possible boundaries in terms of the total gradient flux. We illustrate our method with several toy examples of both straight and curved boundaries with varying amounts of signal present in the data.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Influence of Wetting on Viscous Fingering Via 2D Lattice Boltzmann Simulations

We present simulations of two-phase flow using the Rothman and Keller colour gradient Lattice Boltzmann method to study viscous fingering when a “red fluid” invades a porous model initially filled with a “blue” fluid with different viscosity. We conducted eleven suites of 81 numerical experiments totalling 891 simulations, where each suite had a different random realization of the porous model and spanned viscosity ratios in the range $$M\in [0.01,100]$$ M ϵ [ 0.01 , 100 ] and wetting angles in the range $$\theta _w\in [180^\circ ,0^\circ ]$$ θ w ϵ [ 180 ° , 0 ° ] to allow us to study the effect of these parameters on the fluid-displacement morphology and saturation at breakthrough (sweep). Although sweep often increased with wettability, this was not always so and the sweep phase space landscape, defined as the difference in saturation at a given wetting angle relative to saturation for the non-wetting case, had hills, ridges and valleys. At low viscosity ratios, flow at breakthrough is localized through narrow fingers that span the model. After breakthrough, the flow field continues to evolve and the saturation continues to increase albeit at a reduced rate, and eventually exceeds 90% for both non-wetting and wetting cases. The existence of a complicated sweep phase space at breakthrough, and continued post-breakthrough evolution suggests the hydrodynamics and sweep is a complicated function of wetting angle, viscosity ratio and time, which has major potential implications to Enhanced Oil Recovery by water flooding, and hence, on estimates of global oil reserves. Validation of these results via experiments is required to ensure they translate to field studies.

Engineering↗
NASA NTRS1984

Separation of cells from the rat anterior pituitary gland

Data concerned with analyzing the cellular organization of the rat anterior pituitary gland are examined. The preparation of the cell suspensions and the methods used to separate pituitary cell types are described. Particular emphasis is given to velocity sedimentation at unit gravity, density gradient centrifugation, affinity methods, fluorescence activated cell sorting, and density gradient and continuous-flow electrophoresis. The difficulties encountered when attempting to compare data from different pituitary cell separation studies are discussed, and results from various experiments are presented. The functional capabilities of the separated cell populations can be tested in various culture systems.

Hymer, W. C.↗
DOE OSTI2026

Classical-Quantum Algorithm for Solving Stochastic Programs

Stochastic programming provides a rigorous mathematical framework for making decisions under uncertainty in a risk-aware manner. Two-stage stochastic programming is, perhaps, the simplest form of this framework. Here the first-stage variables represent decisions that must be made "here and now" in the face of uncertainty, while the second-stage variables are decisions made after uncertain events. However, the broad adoption of stochastic programming has been hindered by computational challenges caused by the two-stage stochastic programming formulation which requires solving an ensemble of optimization problems. Using quantum amplitude estimation (QAE), quantum computers have shown the theoretic ability to compute expectations with Monte-Carlo methods with quadratically fewer samples than classical methods. In this work, we present a quantum algorithm for computing the expectation term using QAE for given first-stage decisions. Further, we detail methods of computing gradient information from the quantum calculation enabling the application of classical gradient-based optimization techniques. The result is a classical-quantum hybrid method of solving two-stage stochastic programs. These techniques are demonstrated with computational experiments based an engineering optimization problem.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1990

Statistical analysis of static shape control in space structures

The article addresses the problem of efficient analysis of the statistics of initial and corrected shape distortions in space structures. Two approaches for improving efficiency are considered. One is an adjoint technique for calculating distortion shapes: the second is a modal expansion of distortion shapes in terms of pseudo-vibration modes. The two techniques are applied to the problem of optimizing actuator locations on a 55 m radiometer antenna. The adjoint analysis technique is used with a discrete-variable optimization method. The modal approximation technique is coupled with a standard conjugate-gradient continuous optimization method. The agreement between the two sets of results is good, validating both the approximate analysis and optimality of the results.

Burdisso, Ricardo A.↗
NASA NTRS1997

Methods for Sensitivity Analysis in Gradient-Based Shape Optimization: A Review

CFD is now being widely used in the design of mechanical and aerospace systems and their components. Improving the accuracy and efficiency of this process can reduce its cost and increase its effectiveness. Other motivating factors that push and pull CFD into design optimization are; (1) the need for design methodologies (single- and multi-disciplinary) to reduce cost and time in developing new fluid and aero systems for fast response to market changes; (2) new design requirements outside existing database; (3) new CFD algorithms and growth in computing power; (4) need for increased product quality and reliability; (5) lack of test facilities; (6) limitations and cost of testing techniques; (7) limitations of existing design practices; and (8) prohibitive cost of and conflicting designs derived from isolated component analyses and design, which are repetitive in nature and omit mutual interference.

Baysal, Oktay↗