Search NASA⌕ Search

SEARCH · Search NASA

Results for “subspace method”

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 199 records · Page 11

Chebyshev polynomial filtered subspace iteration in the discontinuous Galerkin method for large-scale electronic structure calculations

The Discontinuous Galerkin (DG) electronic structure method employs an adaptive local basis (ALB) set to solve the Kohn-Sham equations of density functional theory in a discontinuous Galerkin framework. The adaptive local basis is generated on-the-fly to capture the local material physics and can systematically attain chemical accuracy with only a few tens of degrees of freedom per atom. A central issue for large-scale calculations, however, is the computation of the electron density (and subsequently, ground state properties) from the discretized Hamiltonian in an efficient and scalable manner. We show in this work how Chebyshev polynomial filtered subspace iteration (CheFSI) can be used to address this issue and push the envelope in large-scale materials simulations in a discontinuous Galerkin framework. We describe how the subspace filtering steps can be performed in an efficient and scalable manner using a two-dimensional parallelization scheme, thanks to the orthogonality of the DG basis set and block-sparse structure of the DG Hamiltonian matrix. The on-the-fly nature of the ALB functions requires additional care in carrying out the subspace iterations. We demonstrate the parallel scalability of the DG-CheFSI approach in calculations of large-scale twodimensional graphene sheets and bulk three-dimensional lithium-ion electrolyte systems. In conclusion, employing 55 296 computational cores, the time per self-consistent field iteration for a sample of the bulk 3D electrolyte containing 8586 atoms is 90 s, and the time for a graphene sheet containing 11 520 atoms is 75 s.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Chemistry on Quantum Computers with Virtual Quantum Subspace Expansion

Several novel methods for performing calculations relevant to quantum chemistry on quantum computers have been proposed but not yet explored experimentally. Virtual quantum subspace expansion is one such algorithm developed for modeling complex molecules using their full orbital space and without the need for additional quantum resources. Here, we implement this method on the IBM Q platform and calculate the potential energy curves of the hydrogen and lithium dimers using only two qubits and simple classical post-processing. A comparable level of accuracy would require twenty qubits with previous approaches. We also develop an approach to minimize the impact of experimental noise on the stability of a generalized eigenvalue problem that is a crucial component of the algorithm. Our results demonstrate that virtual quantum subspace expansion works well in practice.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Efficient High-Order Solver for Diffusion Equations with Strong Anisotropy on Non-Anisotropy-Aligned Meshes

This paper concerns numerical solution of the diffusion equation with strong anisotropy on meshes not aligned with the anisotropic vector field. In order to resolve the numerical pollution for simulations on a non-anisotropy-aligned mesh and reduce the associated high computational cost we propose an effective preconditioner, extending our previous work. Similar to the anisotropy-aligned mesh case, we apply the auxiliary space preconditioning framework to design a preconditioner where a continuous finite element space is used as the auxiliary space for the discontinuous finite element space. The key component is an effective line smoother that can mitigate the high-frequency errors perpendicular to the magnetic field. We design a graph-based approach to find such a line smoother that is approximately perpendicular to the vector fields when the mesh does not align with the anisotropy. Finally, numerical experiments for several benchmark problems are presented, demonstrating the effectiveness and robustness of the proposed preconditioner when applied to Krylov iterative methods.

97 MATHEMATICS AND COMPUTING↗

Hybrid eigensolvers for nuclear configuration interaction calculations

We examine and compare several iterative methods for solving large-scale eigenvalue problems arising from nuclear structure calculations. In particular, we discuss the possibility of using block Lanczos method, a Chebyshev filtering based subspace iterations and the residual minimization method accelerated by direct inversion of iterative subspace (RMM-DIIS) and describe how these algorithms compare with the standard Lanczos algorithm and the locally optimal block preconditioned conjugate gradient (LOBPCG) algorithm. Although the RMM-DIIS method does not exhibit rapid convergence when the initial approximations to the desired eigenvectors are not sufficiently accurate, it can be effectively combined with either the block Lanczos or the LOBPCG method to yield a hybrid eigensolver that has several desirable properties. We will describe a few practical issues that need to be addressed to make the hybrid solver efficient and robust.

97 MATHEMATICS AND COMPUTING↗

Inverse Subspace Iteration for Spectral Stochastic Finite Element Methods

We study random eigenvalue problems in the context of spectral stochastic finite elements. In particular, given a parameter-dependent, symmetric positive-definite matrix operator, we explore the performance of algorithms for computing its eigenvalues and eigenvectors represented using polynomial chaos expansions. We formulate a version of stochastic inverse subspace iteration, which is based on the stochastic Galerkin finite element method, and we compare its accuracy with that of Monte Carlo and stochastic collocation methods. The coefficients of the eigenvalue expansions are computed from a stochastic Rayleigh quotient. Our approach allows the computation of interior eigenvalues by deflation methods, and we can also compute the coefficients of multiple eigenvectors using a stochastic variant of the modified Gram-Schmidt process. The effectiveness of the methods is illustrated by numerical experiments on benchmark problems arising from vibration analysis.

97 MATHEMATICS AND COMPUTING↗

Acceleration of GPU-based Krylov solvers via data transfer reduction

Krylov subspace iterative solvers are often the method of choice when solving large sparse linear systems. At the same time, hardware accelerators such as graphics processing units continue to offer significant floating point performance gains for matrix and vector computations through easy-to-use libraries of computational kernels. However, as these libraries are usually composed of a well optimized but limited set of linear algebra operations, applications that use them often fail to reduce certain data communications, and hence fail to leverage the full potential of the accelerator. In this study, we target the acceleration of Krylov subspace iterative methods for graphics processing units, and in particular the Biconjugate Gradient Stabilized solver that significant improvement can be achieved by reformulating the method to reduce data-communications through application-specific kernels instead of using the generic BLAS kernels, e.g. as provided by NVIDIA’s cuBLAS library, and by designing a graphics processing unit specific sparse matrix-vector product kernel that is able to more efficiently use the graphics processing unit’s computing power. Furthermore, we derive a model estimating the performance improvement, and use experimental data to validate the expected runtime savings. Finally, considering that the derived implementation achieves significantly higher performance, we assert that similar optimizations addressing algorithm structure, as well as sparse matrix-vector, are crucial for the subsequent development of high-performance graphics processing units accelerated Krylov subspace iterative methods.

97 MATHEMATICS AND COMPUTING↗

Model Order Reduction of Aeroservoelastic Model of Flexible Aircraft

This paper presents a holistic model order reduction (MOR) methodology and framework that integrates key technological elements of sequential model reduction, consistent model representation, and model interpolation for constructing high-quality linear parameter-varying (LPV) aeroservoelastic (ASE) reduced order models (ROMs) of flexible aircraft. The sequential MOR encapsulates a suite of reduction techniques, such as truncation and residualization, modal reduction, and balanced realization and truncation to achieve optimal ROMs at grid points across the flight envelope. The consistence in state representation among local ROMs is obtained by the novel method of common subspace reprojection. Model interpolation is then exploited to stitch ROMs at grid points to build a global LPV ASE ROM feasible to arbitrary flight condition. The MOR method is applied to the X-56A MUTT vehicle with flexible wing being tested at NASA/AFRC for flutter suppression and gust load alleviation. Our studies demonstrated that relative to the fullorder model, our X-56A ROM can accurately and reliably capture vehicles dynamics at various flight conditions in the target frequency regime while the number of states in ROM can be reduced by 10X (from 180 to 19), and hence, holds great promise for robust ASE controller synthesis and novel vehicle design.

aeroelasticity↗

Accurate quantum chemical calculations

An important goal of quantum chemical calculations is to provide an understanding of chemical bonding and molecular electronic structure. A second goal, the prediction of energy differences to chemical accuracy, has been much harder to attain. First, the computational resources required to achieve such accuracy are very large, and second, it is not straightforward to demonstrate that an apparently accurate result, in terms of agreement with experiment, does not result from a cancellation of errors. Recent advances in electronic structure methodology, coupled with the power of vector supercomputers, have made it possible to solve a number of electronic structure problems exactly using the full configuration interaction (FCI) method within a subspace of the complete Hilbert space. These exact results can be used to benchmark approximate techniques that are applicable to a wider range of chemical and physical problems. The methodology of many-electron quantum chemistry is reviewed. Methods are considered in detail for performing FCI calculations. The application of FCI methods to several three-electron problems in molecular physics are discussed. A number of benchmark applications of FCI wave functions are described. Atomic basis sets and the development of improved methods for handling very large basis sets are discussed: these are then applied to a number of chemical and spectroscopic problems; to transition metals; and to problems involving potential energy surfaces. Although the experiences described give considerable grounds for optimism about the general ability to perform accurate calculations, there are several problems that have proved less tractable, at least with current computer resources, and these and possible solutions are discussed.

Bauschlicher, Charles W., Jr.↗

Data-driven surrogates for high dimensional models using Gaussian process regression on the Grassmann manifold

This paper introduces a surrogate modeling scheme based on Grassmannian manifold learning to be used for cost-efficient predictions of high-dimensional stochastic systems. The method exploits subspace-structured features of each solution by projecting it onto a Grassmann manifold. This point-wise linear dimensionality reduction harnesses the structural information to assess the similarity between solutions at different points in the input parameter space. The method utilizes a solution clustering approach in order to identify regions of the parameter space over which solutions are sufficiently similarly such that they can be interpolated on the Grassmannian. In this clustering, the reduced-order solutions are partitioned into disjoint clusters on the Grassmann manifold using the eigen-structure of properly defined Grassmannian kernels and, the Karcher mean of each cluster is estimated. Then, the points in each cluster are projected onto the tangent space with origin at the corresponding Karcher mean using the exponential mapping. For each cluster, a Gaussian process regression model is trained that maps the input parameters of the system to the reduced solution points of the corresponding cluster projected onto the tangent space. Using this Gaussian process model, the full-field solution can be efficiently predicted at any new point in the parameter space. In certain cases, the solution clusters will span disjoint regions of the parameter space. In such cases, for each of the solution clusters we utilize a second, density-based spatial clustering to group their corresponding input parameter points in the Euclidean space. The proposed method is applied to two numerical examples. Here, the first is a nonlinear stochastic ordinary differential equation with uncertain initial conditions where the surrogate is used to predict the time history solution. The second involves modeling of plastic deformation in a model amorphous solid using the Shear Transformation Zone theory of plasticity, where the proposed surrogate is used to predict the full strain field of a material specimen under large shear strains.

42 ENGINEERING↗

Feedback Controllability Components Analysis (FCCA) v1.0

FCCA is a linear dimensionality reduction method that find subspaces of high-dimensional time-series data that are most feedback controllable. The key innovation is to formulate an objective function that quantifies the joint cost of state reconstruction and state regulation that can be evaluated from purely observational data. To do this, it leverages the duality between controllability and observability. We provide analytic results demonstrating the validity of the cost function. We evaluated this method in both synthetic and real neural data (from multiple organisms and brain areas).

Kumar, Ankit↗

What is the gradient of a scalar function defined on a subspace of square matrices?

We illustrate a technique to calculate the gradient of scalar functions that are defined on any arbitrary matrix subspace. It generalizes our earlier work titled “What is the gradient of a scalar function of a symmetric matrix ?”(Indian Journal of Pure and Applied Mathematics (2022), doi:10.1007/s13226-022-00313-x), in which we considered the special case of the subspace of symmetric matrices. Extant methods to calculate the gradient in such cases have an inherent flaw which leads to spurious results that populate several publications, as well as respected textbooks and handbooks on matrix calculus. We examine these sources and results in a rigorous and concrete mathematical setting of a finite-dimensional inner-product space and discover the inherent flaw and also a remedy. We demonstrate two ways to calculate the derivative/gradient and second derivative for scalar functions of matrices defined over an arbitrary matrix subspace; the first method is by considering any (differentiable) extension to the space of square matrices and projection of its gradient onto the given subspace. The second method utilizes an ordered basis and computes each component of the gradient through evaluation of the directional derivative. All the ideas presented are illustrated by non-trivial examples, namely, considering the subspace of 3 × 3 circulant and Toeplitz matrices and presenting the results of gradient-descent with both the spurious and correct gradients. Moreover, our bibliography makes it clear that a rigorous approach to matrix calculus is not common in practice, and our presentation of matrix calculus in the language of inner-product spaces will be significant and meaningful for applied mathematicians, engineers and researchers working in inter-disciplinary fields to avoid the conceptual pitfalls that exist.

97 MATHEMATICS AND COMPUTING↗

What is the gradient of a scalar function defined on a subspace of square matrices?

We illustrate a technique to calculate the gradient of scalar functions that are defined on any arbitrary matrix subspace. It generalizes our earlier work titled “What is the gradient of a scalar function of a symmetric matrix ?”(Indian Journal of Pure and Applied Mathematics (2022), https://doi.org/10.1007/s13226-022-00313-x), in which we considered the special case of the subspace of symmetric matrices. Extant methods to calculate the gradient in such cases have an inherent flaw which leads to spurious results that populate several publications, as well as respected textbooks and handbooks on matrix calculus. Here, we examine these sources and results in a rigorous and concrete mathematical setting of a finite-dimensional inner-product space and discover the inherent flaw and also a remedy. We demonstrate two ways to calculate the derivative/gradient and second derivative for scalar functions of matrices defined over an arbitrary matrix subspace; the first method is by considering any (differentiable) extension to the space of square matrices and projection of its gradient onto the given subspace. The second method utilizes an ordered basis and computes each component of the gradient through evaluation of the directional derivative. All the ideas presented are illustrated by non-trivial examples, namely, considering the subspace of 3 x 3 circulant and Toeplitz matrices and presenting the results of gradient-descent with both the spurious and correct gradients. Moreover, our bibliography makes it clear that a rigorous approach to matrix calculus is not common in practice, and our presentation of matrix calculus in the language of inner-product spaces will be significant and meaningful for applied mathematicians, engineers and researchers working in inter-disciplinary fields to avoid the conceptual pitfalls that exist.

97 MATHEMATICS AND COMPUTING↗

Adaptive Waveform Correlation Detectors for Arrays: Algorithms for Autonomous Calibration

Waveform correlation detectors compare a signal template with successive windows of a continuous data stream and report a detection when the correlation coefficient, or some comparable detection statistic, exceeds a specified threshold. Since correlation detectors exploit the fine structure of the full waveform, they are exquisitely sensitive when compared to power (STA/LTA) detectors. The drawback of correlation detectors is that they require complete knowledge of the signal to be detected, which limits such methods to instances of seismicity in which a very similar signal has already been observed by every station used. Such instances include earthquake swarms, aftershock sequences, repeating industrial seismicity, and many other forms of controlled explosions. The reduction in the detection threshold is even greater when the techniques are applied to arrays since stacking can be performed on the individual channel correlation traces to achieve significant array gain. In previous years we have characterized the decrease in detection threshold afforded by correlation detection across an array or network when observations of a previous event provide an adequate template for signals from subsequent events located near the calibration event. Last year we examined two related issues: (1) the size of the source region calibration footprint afforded by a master event, and (2) the use of temporally incoherent detectors designed to detect the gross envelope structure of the signal to extend the footprint. In Case 1, results from the PETROBAR-1 marine refraction profile indicated that array correlation gain was usable at inter-source separations out to one or two wavelengths. In Case 2, we found that incoherent detectors developed from a magnitude 6 event near Svalbard were successful at detecting aftershocks where correlation detectors derived from individual aftershocks were not. Incoherent detectors might provide 'seed' events for correlation detectors that then could extend detection to lower magnitudes. This year we addressed a problem long known to limit the acceptance of correlation detectors in practice: the labor intensive development of templates. For example, existing design methods cannot keep pace with rapidly unfolding aftershock sequences. We successfully built and tested an object-oriented framework (as described in our 2005 proposal) for autonomous calibration of waveform correlation detectors for an array. The framework contains a dynamic list of detectors of several types operating on a continuous array data stream. The list has permanent detectors: beam forming power (STA/LTA) detectors which serve the purpose of detecting signals not yet characterized with a waveform template. The framework also contains an arbitrary number of subspace detectors which are launched automatically using the waveforms from validated power detections as templates. The implementation is very efficient such that the computational cost of adding subspace detectors was low. The framework contains a supervisor that oversees the validation of power detections, and periodically halts the processing to revise the portfolio of detectors. The process of revision consists of collecting the waveforms from all detections, performing cross-correlations pairwise among all waveforms, clustering the detections using correlations as a distance measure, then creating a new subspace detector from each cluster. The collection of new subspace detectors replaces the existing portfolio and processing of the data stream resumes. This elaborate scheme was implemented to prevent proliferation of closely-related subspace detectors. The method performed very well on several simple sequences: 2005 'drumbeat' events observed locally at Mt. St. Helens, and the 2003 Orinda, CA aftershock sequence. Our principal test entailed detection of the aftershocks of the San Simeon earthquake using the NVAR array; in this case, the system automatically detected and categorized approximately 2/3 of the events above magnitude 2.8.

58 GEOSCIENCES↗

Comparison of Fuel Loading Pattern Optimization Results using Exhaustive Search for Fresh Fuels and Local Search for Burned Fuels

Fuel loading pattern optimization is a combinatorial optimization problem from the mathematical point of view. However, exhaustive search is impractical because enormous computational cost is necessary due to huge search space. For this reason, various optimization methods have been developed so far, but no methods are able to guarantee to reach the optimum solution. In the present study, search space is divided into subspaces using key parameters that well represent characteristics of a loading pattern and then promising subspaces are extensively searched. In the present method, balance between division of a search space into subspaces and a search method used in a subspace is essential. For example, when the number of division is set to be one, it is equivalent to the exhaustive search thus is impractical. Similarly, when the division to subspaces is too coarse, an optimum (or near optimum) solution cannot be found through local search in each subspace since design space of a subspace is still too large. Contrary, if division to subspaces is too fine, number of subspaces becomes to large thus promising subspaces cannot be found. In the present study, search space is divided focusing on the location of fresh fuel assemblies with gadolinia since it has significant impact on core characteristics. Number of possible loading patterns of gadolinia bearing fresh fuel assemblies is not very large thus exhaustive search is possible. Local fine search is executed by considering location of burned fuels for each loading pattern of gadolinia bearing fresh fuel assembly. If the same solution is obtained by multiple local searches starting from different initial burnt fuel locations (but the same loading patterns for gadolinia bearing fresh fuel assemblies), the obtained solution is highly likely the optimum in the subspace. In more general, the present method is based on the 'divide and conquer' approach, which is familiar in the optimization field. Namely, after division of search space into subspaces and then promising subspace is searched. Once the promising subspace is found, the subspace is further subdivided into subspaces and repeat the above procedures. This method would be efficient when appropriate balance between search space division method and performance of local search is achieved. The previous works utilizes the similar approach. In reference [1], exhaustive search in small region in a core is carried out focusing on loading patterns of gadolinia bearing fresh fuel assemblies. In reference [2], rough exhaustive search based on the 'batch-wise' loading pattern is performed with the assist of the linear programming. By contrast, in the present study, exhaustive search is carried out for loading patterns of gadolinia bearing fresh fuel assemblies. The random Multi-start Local Search (MSLS) with the direct search or the exhaustive binary swap is used to find out optimal solutions in a subspace, which consists of loading patterns having the same arrangement of gadolinia bearing fresh fuel assemblies. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Classification of journal surfaces using surface topography parameters and software methods to compensate for stylus geometry

Measurements made with a stylus surface tracer which provides a digitized representation of a surface profile are discussed. Parameters are defined to characterize the height (e.g., RMS roughness, skewness, and kurtosis) and length (e.g., autocorrelation) of the surface topography. These are applied to the characterization of crank shaft journals which were manufactured by different grinding and lopping procedures known to give significant differences in crank shaft bearing life. It was found that three parameters (RMS roughness, skewness, and kurtosis) are necessary to adequately distinguish the character of these surfaces. Every surface specimen has a set of values for these three parameters. They can be regarded as a set coordinate in a space constituted by three characteristics axes. The various journal surfaces can be classified along with the determination of a proper wavelength cutoff (0.25 mm) by using a method of separated subspace. The finite radius of the stylus used for profile tracing gives an inherent measurement error as it passes over the fine structure of the surface. A mathematical model is derived to compensate for this error.

Li, C. J.↗

Solution of the symmetric eigenproblem AX=lambda BX by delayed division

Delayed division is an iterative method for solving the linear eigenvalue problem AX = lambda BX for a limited number of small eigenvalues and their corresponding eigenvectors. The distinctive feature of the method is the reduction of the problem to an approximate triangular form by systematically dropping quadratic terms in the eigenvalue lambda. The report describes the pivoting strategy in the reduction and the method for preserving symmetry in submatrices at each reduction step. Along with the approximate triangular reduction, the report extends some techniques used in the method of inverse subspace iteration. Examples are included for problems of varying complexity.

Thurston, G. A.↗

Stochastic Trust-Region Algorithm in Random Subspaces with Convergence and Expected Complexity Analyses

Here, this work proposes a framework for large-scale stochastic derivative-free optimization (DFO) by introducing STARS, a trust-region method based on iterative minimization in random subspaces. This framework is both an algorithmic and theoretical extension of a random subspace derivative-free optimization (RSDFO) framework, and an algorithm for stochastic optimization with random models (STORM). Moreover, like RSDFO, STARS achieves scalability by minimizing interpolation models that approximate the objective in low-dimensional affine subspaces, thus significantly reducing per-iteration costs in terms of function evaluations and yielding strong performance on largescale stochastic DFO problems. The user-determined dimension of these subspaces, when the latter are defined, for example, by the columns of so-called Johnson-Lindenstrauss transforms, turns out to be independent of the dimension of the problem. For convergence purposes, inspired by the analyses of RSDFO and STORM, both a particular quality of the subspace and the accuracies of random function estimates and models are required to hold with sufficiently high, but fixed, probabilities. Using martingale theory under the latter assumptions, an almost sure global convergence of STARS to a first-order stationary point is shown, and the expected number of iterations required to reach a desired first-order accuracy is proved to be similar to that of STORM and other stochastic DFO algorithms, up to constants.

97 MATHEMATICS AND COMPUTING↗