Search NASA⌕ Search

SEARCH · Search NASA

Results for “Gradient Search Algorithm”

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 73 records · Page 4
All sourcesNASA NTRSDOE OSTI
NASA NTRS2018

Data Mining for Vortices on the Earth's Magnetosphere - Algorithm Application for Detection and Analysis

Unsteady processes in the solar wind– magnetosphere interaction, such as vortices developed at the magnetopause boundary by the Kelvin–Helmholtz instability, may contribute to the process of mass, momentum and energy transfer into the Earth’s magnetosphere. The research described in this paper validates an algorithm to automatically detect and characterize vortices based on velocity data from simulations. The vortex identification algorithm (VIA) systematically searches the 3-D velocity fields to identify critical points where the magnitude of the velocity vector vanishes. The velocity gradient tensor is computed and its invariants are used to assess vortex structure in the flow field. We use the Community Coordinated Modeling Center (CCMC) Runs on Request capability to create a series of model runs initialized from the conditions observed by the Cluster mission in the Hwang et al. (2011) analysis of Kelvin–Helmholtz vortices observed during southward interplanetary magnetic field (IMF) conditions. We analyze further the properties of the vortices found in the runs, including the velocity changes within their motion across the magnetosheath. We also demonstrate the potential of our tool to identify and characterize other transient features (e.g., flux transfer events, FTEs) with vortical internal structures. We find that the vortices are associated with flows on the magnetosheath side of the magnetopause that reach speeds greater than the solar wind speed at the bow shock.

Collado-Vega, Yaireska M.↗
DOE OSTI2022

An adaptive stochastic sequential quadratic programming with differentiable exact augmented lagrangians

In this study, we consider solving nonlinear optimization problems with a stochastic objective and deterministic equality constraints. We assume for the objective that its evaluation, gradient, and Hessian are inaccessible, while one can compute their stochastic estimates by, for example, subsampling. We propose a stochastic algorithm based on sequential quadratic programming (SQP) that uses a differentiable exact augmented Lagrangian as the merit function. To motivate our algorithm design, we first revisit and simplify an old SQP method Lucidi developed for solving deterministic problems, which serves as the skeleton of our stochastic algorithm. Based on the simplified deterministic algorithm, we then propose a non-adaptive SQP for dealing with stochastic objective, where the gradient and Hessian are replaced by stochastic estimates but the stepsizes are deterministic and prespecified. Finally, we incorporate a recent stochastic line search procedure Paquette and Scheinberg into the non-adaptive stochastic SQP to adaptively select the random stepsizes, which leads to an adaptive stochastic SQP. The global "almost sure" convergence for both non-adaptive and adaptive SQP methods is established. Numerical experiments on nonlinear problems in CUTEst test set demonstrate the superiority of the adaptive algorithm.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1993

Trajectory optimization based on differential inclusion

A method for generating finite-dimensional approximations to the solutions of optimal control problems is introduced. By employing a description of the dynamical system in terms of its attainable sets in favor of using differential equations, the controls are completely eliminated from the system model. Besides reducing the dimensionality of the discretized problem compared to state-of-the-art collocation methods, this approach also alleviates the search for initial guesses from where standard gradient search methods are able to converge. The mechanics of the new method are illustrated on a simple double integrator problem. The performance of the new algorithm is demonstrated on a 1-D rocket ascent problem ('Goddard Problem') in presence of a dynamic pressure constraint.

Seywald, Hans↗
NASA NTRS1993

Trajectory optimization based on differential inclusion

A method for generating finite-dimensional approximations to the solutions of optimal control problems is introduced. By employing a description of the dynamical system in terms of its attainable sets in favor of using differential equations, the controls are completely eliminated from the system model. Besides reducing the dimensionality of the discretized problem compared to state-of-the-art collocation methods, this approach also alleviates the search for initial guesses from where standard gradient search methods are able to converge. The mechanics of the new method are illustrated on a simple double integrator problem. The performance of the new algorithm is demonstrated on a I-D rocket ascent problem ('Goddard Problem') in presence of a dynamic pressure constraint.

Seywald, Hans↗
NASA NTRS1993

Trajectory optimization based on differential inclusion

A method for generating finite-dimensional approximations to the solutions of optimal control problems is introduced. By employing a description of the dynamical system in terms of its attainable sets in favor of using differential equations, the controls are completely eliminated from the system model. Besides reducing the dimensionality of the discretized problem compared to state-of-the-art collocation methods, this approach also alleviates the search for initial guesses from where standard gradient search methods are able to converge. The mechanics of the new method are illustrated on a simple double integrator problem. The performance of the new algorithm is demonstrated on a 1-D rocket ascent problem (`Goddard Problem') in presence of a dynamic pressure constraint.

Seywald, Hans↗
NASA NTRSDate not supplied

Analytical Partial Derivatives of the Q-Law Guidance Algorithm

One method of producing low-thrust trajectories is the closed-loop Q-Law guidance algorithm. This work wraps Q-Law inside a nonlinear programming problem to search for locally optimal gain combinations and enforce nonlinear constraints on the initial state. Gradient-based trajectory optimization has been shown to benefit greatly when analytical partial derivatives are supplied to the optimizer. This paper presents the Q-Law optimization problem setup and derives the Q-Law thrust vector partial derivatives for the State Transition Matrix required by standard nonlinear optimization software.

Jackson L Shannon↗
NASA NTRS1992

Sea surface velocities from visible and infrared multispectral atmospheric mapping sensor imagery

High resolution (100 m), sequential Multispectral Atmospheric Mapping Sensor (MAMS) images were used in a study to calculate advective surface velocities using the Maximum Cross Correlation (MCC) technique. Radiance and brightness temperature gradient magnitude images were formed from visible (0.48 microns) and infrared (11.12 microns) image pairs, respectively, of Chandeleur Sound, which is a shallow body of water northeast of the Mississippi delta, at 145546 GMT and 170701 GMT on 30 Mar. 1989. The gradient magnitude images enhanced the surface water feature boundaries, and a lower cutoff on the gradient magnitudes calculated allowed the undesirable sunglare and backscatter gradients in the visible images, and the water vapor absorption gradients in the infrared images, to be reduced in strength. Requiring high (greater than 0.4) maximum cross correlation coefficients and spatial coherence of the vector field aided in the selection of an optimal template size of 10 x 10 pixels (first image) and search limit of 20 pixels (second image) to use in the MCC technique. Use of these optimum input parameters to the MCC algorithm, and high correlation and spatial coherence filtering of the resulting velocity field from the MCC calculation yielded a clustered velocity distribution over the visible and infrared gradient images. The velocity field calculated from the visible gradient image pair agreed well with a subjective analysis of the motion, but the velocity field from the infrared gradient image pair did not. This was attributed to the changing shapes of the gradient features, their nonuniqueness, and large displacements relative to the mean distance between them. These problems implied a lower repeat time for the imagery was needed in order to improve the velocity field derived from gradient imagery. Suggestions are given for optimizing the repeat time of sequential imagery when using the MCC method for motion studies. Applying the MCC method to the infrared brightness temperature imagery yielded a velocity field which did agree with the subjective analysis of the motion and that derived from the visible gradient imagery. Differences between the visible and infrared derived velocities were 14.9 cm/s in speed and 56.7 degrees in direction. Both of these velocity fields also agreed well with the motion expected from considerations of the ocean bottom topography and wind and tidal forcing in the study area during the 2.175 hour time interval.

Pope, P. A.↗
NASA NTRS2017

Development of Multi-Sensor Global Cloud and Radiance Composites for Earth Radiation Budget Monitoring from DSCOVR

The Deep Space Climate Observatory (DSCOVR) enables analysis of the daytime Earth radiation budget via the onboard Earth Polychromatic Imaging Camera (EPIC) and National Institute of Standards and Technology Advanced Radiometer (NISTAR). Radiance observations and cloud property retrievals from low earth orbit and geostationary satellite imagers have to be co-located with EPIC pixels to provide scene identification in order to select anisotropic directional models needed to calculate shortwave and longwave fluxes. A new algorithm is proposed for optimal merging of selected radiances and cloud properties derived from multiple satellite imagers to obtain seamless global hourly composites at 5-kilometer resolution. An aggregated rating is employed to incorporate several factors and to select the best observation at the time nearest to the EPIC measurement. Spatial accuracy is improved using inverse mapping with gradient search during reprojection and bicubic interpolation for pixel resampling. The composite data are subsequently remapped into EPIC-view domain by convolving composite pixels with the EPIC point spread function (PSF) defined with a half-pixel accuracy. PSF-weighted average radiances and cloud properties are computed separately for each cloud phase. The algorithm has demonstrated contiguous global coverage for any requested time of day with a temporal lag of under 2 hours in over 95 percent of the globe.

Khlopenkov, Konstantin↗
NASA NTRS1988

MHOST: An efficient finite element program for inelastic analysis of solids and structures

An efficient finite element program for 3-D inelastic analysis of gas turbine hot section components was constructed and validated. A novel mixed iterative solution strategy is derived from the augmented Hu-Washizu variational principle in order to nodally interpolate coordinates, displacements, deformation, strains, stresses and material properties. A series of increasingly sophisticated material models incorporated in MHOST include elasticity, secant plasticity, infinitesimal and finite deformation plasticity, creep and unified viscoplastic constitutive model proposed by Walker. A library of high performance elements is built into this computer program utilizing the concepts of selective reduced integrations and independent strain interpolations. A family of efficient solution algorithms is implemented in MHOST for linear and nonlinear equation solution including the classical Newton-Raphson, modified, quasi and secant Newton methods with optional line search and the conjugate gradient method.

Nakazawa, S.↗
DOE OSTI2024

Automatic Multiple Experiment Simulation and Fitting (Ames-Fit)

AMES-Fit is a program used to automatically fit the multi-field solid-state NMR spectra of half-integer quadrupolar nuclei. Due to the high dimensional space, gradient algorithms have failed to address the fitting of such data, which is at present done manually. AMES-Fit diverges from these approaches by using an adaptive step size random search algorithm to fit the NMR spectra to consistently find the global best fit parameters.

Perras, Frederic↗
DOE OSTI2020

Sign-OPT: A Query-Efficient Hard-label Adversarial Attack

We study the most practical problem setup for evaluating adversarial robustness of a machine learning system with limited access: the hard-label black-box attack setting for generating adversarial examples, where limited model queries are allowed and only the decision is provided to a queried data input. Several algorithms have been proposed for this problem but they typically require huge amount (>20,000) of queries for attacking one example. Among them, one of the state-of-the-art approaches (Cheng et al., 2019) showed that hard-label attack can be modeled as an optimization problem where the objective function can be evaluated by binary search with additional model queries, thereby a zeroth order optimization algorithm can be applied. In this paper, we adopt the same optimization formulation but propose to directly estimate the sign of gradient at any direction instead of the gradient itself, which enjoys the benefit of single query. Using this single query oracle for retrieving sign of directional derivative, we develop a novel query-efficient Sign-OPT approach for hard-label black-box attack. We provide a convergence analysis of the new algorithm and conduct experiments on several models on MNIST, CIFAR-10 and ImageNet. We find that Sign-OPT attack consistently requires 5X to 10X fewer queries when compared to the current state-of-the-art approaches, and usually converges to an adversarial example with smaller perturbation.

Cheng, Minhao↗
DOE OSTI2019

Quantum control landscape of bipartite systems

The control landscape of a quantum system A interacting with another quantum system B is studied. Only system A is accessible through time dependent controls, while system B is not accessible. The objective is to find controls that implement a desired unitary transformation on A, regardless of the evolution on B, at a sufficiently large final time. The freedom in the evolution on B is used to define an extended control landscape on which the critical points are investigated in terms of kinematic and dynamic gradients. A spectral decomposition of the corresponding extended unitary system simplifies the landscape analysis which provides: (i) a sufficient condition on the rank of the dynamic gradient of the extended landscape that guarantees a trap free search for the final time unitary matrix of system A, and (ii) a detailed decomposition of the components of the overall dynamic gradient matrix. Consequently, if the rank condition is satisfied, a gradient algorithm will find the controls that implements the target unitary on system A. It is shown that even if the dynamic gradient with respect to the controls alone is not full rank, the additional flexibility due to the parameters that define the extended landscape still can allow for the rank condition of the extended landscape to hold. Moreover, satisfaction of the latter rank condition subsumes any assumptions about controllability, reachability and control resources. Here satisfaction of the rank condition is taken as an assumption. The conditions which ensure that it holds remain an open research question. We lend some numerical support with two common examples for which the rank condition holds.

97 MATHEMATICS AND COMPUTING↗
NASA NTRSDate not supplied

Q-Law for Rapid Assessment of Low Thrust Cislunar Trajectories Via Automatic Differentiation

Q-Law is a Lyapunov-based control law used to determine optimal controls for a low thrust trajectory. One major issue with its use is the difficult derivatives re-quired for calculating optimal controls at a given time. In this paper, an implementation of Q-Law with automatic differentiation via a Python package called JAX is applied. With automatic differentiation, the difficult derivatives for Q-Law’s optimal controls are calculated with ease, and derivatives of final states with respect to Q-Law’s weights are found enabling gradient-based optimization of Q-Law for the first time. Different search and optimization methods for finding optimal weights are then compared using the LEO to GEO problem, and it was found that gradient-free methods like design of experiments and genetic algorithm produced the best results, but they took the longest time to get a solution, while the gradient-based method found a locally optimal result in a much faster time. Overall, the run time for a single propagation is manageable and well-suited for a mission designer to use as an initial guess generator for trajectory optimization, or for simple orbit transfer analysis.

Nathan Steffen↗
DOE OSTI2021

Bayesian sparse learning with preconditioned stochastic gradient MCMC and its applications

Deep neural networks have been successfully employed in an extensive variety of research areas, including solving partial differential equations. Despite its significant success, there are some challenges in effectively training DNN, such as avoiding overfitting in over-parameterized DNNs and accelerating the optimization in DNNs with pathological curvature. Here, we propose a Bayesian type sparse deep learning algorithm. The algorithm utilizes a set of spike-and-slab priors for the parameters in the deep neural network. The hierarchical Bayesian mixture will be trained using an adaptive empirical method. That is, one will alternatively sample from the posterior using preconditioned stochastic gradient Langevin Dynamics (PSGLD), and optimize the latent variables via stochastic approximation. The sparsity of the network is achieved while optimizing the hyperparameters with adaptive searching and penalizing. A popular SG-MCMC approach is Stochastic gradient Langevin dynamics (SGLD). However, considering the complex geometry in the model parameter space in nonconvex learning, updating parameters using a universal step size in each component as in SGLD may cause slow mixing. To address this issue, we apply a computationally manageable preconditioner in the updating rule, which provides a step-size parameter to adapt to local geometric properties. Moreover, by smoothly optimizing the hyperparameter in the preconditioning matrix, our proposed algorithm ensures a decreasing bias, which is introduced by ignoring the correction term in the preconditioned SGLD. According to the existing theoretical framework, we show that the proposed algorithm can asymptotically converge to the correct distribution with a controllable bias under mild conditions. Numerical tests are performed on both synthetic regression problems and learning solutions of elliptic PDE, which demonstrate the accuracy and efficiency of the present work.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Accelerated Sparse Recovery via Gradient Descent with Nonlinear Conjugate Gradient Momentum

This paper applies an idea of adaptive momentum for the nonlinear conjugate gradient to accelerate optimization problems in sparse recovery. Specifically, we consider two types of minimization problems: a (single) differentiable function and the sum of a non-smooth function and a differentiable function. In the first case, we adopt a fixed step size to avoid the traditional line search and establish the convergence analysis of the proposed algorithm for a quadratic problem. This acceleration is further incorporated with an operator splitting technique to deal with the non-smooth function in the second case. As a result, we use the convex ι 1 and the nonconvex ι 1 – ι 2 functionals as two case studies to demonstrate the efficiency of the proposed approaches over traditional methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Application and Evaluation of Surrogate Models for Radiation Source Search

Surrogate models are increasingly required for applications in which first-principles simulation models are prohibitively expensive to employ for uncertainty analysis, design, or control. They can also be used to approximate models whose discontinuous derivatives preclude the use of gradient-based optimization or data assimilation algorithms. We consider the problem of inferring the 2D location and intensity of a radiation source in an urban environment using a ray-tracing model based on Boltzmann transport theory. Whereas the code implementing this model is relatively efficient, extension to 3D Monte Carlo transport simulations precludes subsequent Bayesian inference to infer source locations, which typically requires thousands to millions of simulations. Additionally, the resulting likelihood exhibits discontinuous derivatives due to the presence of buildings. To address these issues, we discuss the construction of surrogate models for optimization, Bayesian inference, and uncertainty propagation. Specifically, we consider surrogate models based on Legendre polynomials, multivariate adaptive regression splines, radial basis functions, Gaussian processes, and neural networks. We detail strategies for computing training points and discuss the merits and deficits of each method.

Cook, Jared A.↗
DOE OSTI2020

Parallel-in-Time Simulation of Lindblad's Equation

Constructing fast quantum logic gates is critical to building a scalable quantum computer. We consider a qudit, a quantum version of a bit that can take an arbitrary number of states, coupled with a cavity. In this project, we wish to force the qudit to reach the 0-state, for any possible initial state. The coupled system changes in time according to Lindblad’s equation, an ordinary differential equation on the density matrix of the quantum system. Lindblad’s equation contains some parameters that we can control, so-called control functions. We seek control functions which force the qudit to the 0-state within 2 microseconds, which is much faster than what is currently done in practice. The search method is gradient descent, a numerical optimization method that uses gradient information to iteratively improve the control parameters. My contribution to this project is an attempt to speed up the computation of the gradient. It currently takes about 40 seconds to compute the gradient which involves solving a set of ODEs sequentially. Current supercomputers have thousands of cores, but sequential computations can only make use of 1 core at a time. We wish to divide up the work better, so that we can use many more cores at once. To this end, we have implemented the Multigrid Reduction in Time (MGRIT) algorithm. We perform a systematic parameter search on how to best apply this algorithm. Results indicate a 25 percent speed up for solving Lindblad’s equation and determining how close the final state is the 0-state.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Machine learning and deep learning for mineralogy interpretation and CO 2 saturation estimation in geological carbon Storage: A case study in the Illinois Basin

Carbon capture and storage (CCS) is a promising approach to simultaneously maintaining energy security and reducing carbon dioxide (CO 2 ) emissions under the current energy portfolio that is dominated by fossil fuel energy. Pre-injection formation characterization and post-injection CO 2 monitoring are two critical tasks to guarantee storage efficiency in CCS. The CCS projects in the Illinois Basin, the first large-scale CO 2 injection into saline aquifers in the United States, employed conventional and the latest pulsed neutron logging (PNL) tools for mineralogy interpretation and CO 2 saturation estimation, which provide valuable references for future CCS projects. Because of the inherent fuzziness of petrophysical measurements and complex subsurface heterogeneity, interpreting well-logging data is time-consuming, and its accuracy can be user-biased. In recent years, data-driven methods have been widely used to capture the non-linear patterns between input features and interpretation results. This work applied and evaluated four commonly used machine learning (ML) models, including ridge regression (RR), random forest (RF), gradient boosting regression (GBR), support vector regression (SVR), and one deep learning (DL) model, the artificial neural network (ANN). We optimized the hyperparameters of the four ML models and the DL model using the simulated annealing algorithm and the grid search strategy, respectively. The input features of the mineralogy interpretation models were eleven conventional well-logging parameters, and the label data (i.e., ground truth) were the porosity and volumetric fractions of six minerals, including quartz, feldspar, dolomite, calcite, clay, and iron minerals. The results demonstrated that the GBR and RF models were superior in predicting volumetric fractions of minerals and porosity; label data with low coefficient of variation (CV) values tended to yield better performance. For CO 2 saturation estimation, the RF was the best-performing model, followed by SVR, ANN, GBR, and RR. Furthermore, we conducted feature importance ranking using the permutation importance algorithm and found that the formation sigma and well pressure were the most important features in this study. In conclusion, the study of CCS projects in the Illinois Basin bridges the gap between the limited knowledge and understanding of geological carbon storage and the increasing demand for reliable, cost-effective, and sustainable energy solutions.

58 GEOSCIENCES↗