Search NASA⌕ Search

SEARCH · Search NASA

Results for “least squares problem”

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 37 records · Page 2

On damping mechanisms in beams

A partial differential equation model of a cantilevered beam with a tip mass at its free end is used to study damping in a composite. Four separate damping mechanisms consisting of air damping, strain rate damping, spatial hysteresis and time hysteresis are considered experimentally. Dynamic tests were performed to produce time histories. The time history data is then used along with an approximate model to form a sequence of least squares problems. The solution of the least squares problem yields the estimated damping coefficients. The resulting experimentally determined analytical model is compared with the time histories via numerical simulation of the dynamic response. The procedure suggested here is compared with a standard modal damping ratio model commonly used in experimental modal analysis.

Banks, H. T.↗

On damping mechanisms in beams

A partial differential equation model of a cantilevered beam with a tip mass at its free end is used to study damping in a composite. Four separate damping mechanisms consisting of air damping, strain rate damping, spatial hysteresis and time hysteresis are considered experimentally. Dynamic tests were performed to produce time histories. The time history data is then used along with an approximate model to form a sequence of least squares problems. The solution of the least squares problem yields the estimated damping coefficients. The resulting experimentally determined analytical model is compared with the time histories via numerical simulation of the dynamic response. The procedure suggested here is compared with a standard modal damping ratio model commonly used in experimental modal analysis.

Banks, H. T.↗

Randomized Algorithms for Symmetric Nonnegative Matrix Factorization

Symmetric Nonnegative Matrix Factorization (SymNMF) is a technique in data analysis and machine learning that approximates a matrix with a product of a nonnegative, low-rank matrix and it transpose. To design faster and more scalable algorithms for SymNMF we develop two randomized algorithms for its computation. The first method uses randomized matrix sketching to compute an initial low-rank approximation to the input matrix and proceeds to uses this as a low-rank input to rapidly compute a SymNMF. The second methods uses randomized leverage score sampling to approximately solve constrained least squares problems. Many successful methods for SymNMF rely on (approximately) solving sequences of constrained least squares problems. Here, we prove theoretically that leverage score sampling can approximately solve constrained least squares problems to e-accuracy. Finally we demonstrate both methods work in practice by applying them to graph clustering tasks on large real world data sets. These experiments show that our methods approximately maintain solution quality and achieve significant speed ups for both large dense and large sparse problems.

97 MATHEMATICS AND COMPUTING↗

Multivariable frequency domain identification via 2-norm minimization

The author develops a computational approach to multivariable frequency domain identification, based on 2-norm minimization. In particular, a Gauss-Newton (GN) iteration is developed to minimize the 2-norm of the error between frequency domain data and a matrix fraction transfer function estimate. To improve the global performance of the optimization algorithm, the GN iteration is initialized using the solution to a particular sequentially reweighted least squares problem, denoted as the SK iteration. The least squares problems which arise from both the SK and GN iterations are shown to involve sparse matrices with identical block structure. A sparse matrix QR factorization method is developed to exploit the special block structure, and to efficiently compute the least squares solution. A numerical example involving the identification of a multiple-input multiple-output (MIMO) plant having 286 unknown parameters is given to illustrate the effectiveness of the algorithm.

Bayard, David S.↗

Fast Combinatorial Nonnegative Least Squares v 1.0

Software that employs an algorithm that rapidly and rigorously solves nonnegative least squares problems with many right hand sides. This is the latest internal version of MATLAB pseudocode that was described in a published in M. H. Van Benthem and M. R. Keenan, "Fast algorithm for the solution of largescale non-negativity-constrained least squares problems," J. Chemom. 18(10), 441-450 (2004). Example code of this algorithm was included in the publication. This algorithm was also patented in M. H. Van Benthem and M. R. Keenan, "Fast combinatorial algorithm for the solution of linearly constrained least squares problems,"USPTO, Ed., Sandia Corporation, USA (2008). SAND2019-10098 M Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-NA0003525.

Van Benthem, Mark↗

Operator inference for non-intrusive model reduction of systems with non-polynomial nonlinear terms

Here in this work we present a non-intrusive model reduction method to learn low-dimensional models of dynamical systems with non-polynomial nonlinear terms that are spatially local and that are given in analytic form. In contrast to state-of-the-art model reduction methods that are intrusive and thus require full knowledge of the governing equations and the operators of a full model of the discretized dynamical system, the proposed approach requires only the non-polynomial terms in analytic form and learns the rest of the dynamics from snapshots computed with a potentially black-box full-model solver. The proposed method learns operators for the linear and polynomially nonlinear dynamics via a least-squares problem, where the given non-polynomial terms are incorporated on the right-hand side. The least-squares problem is linear and thus can be solved efficiently in practice. The proposed method is demonstrated on three problems governed by partial differential equations, namely the diffusion–reaction Chafee–Infante model, a tubular reactor model for reactive flows, and a batch-chromatography model that describes a chemical separation process. The numerical results provide evidence that the proposed approach learns reduced models that achieve comparable accuracy as models constructed with state-of-the-art intrusive model reduction methods that require full knowledge of the governing equations.

42 ENGINEERING↗

Scalable tensor factorizations with missing data.

The problem of missing data is ubiquitous in domains such as biomedical signal processing, network traffic analysis, bibliometrics, social network analysis, chemometrics, computer vision, and communication networks|all domains in which data collection is subject to occasional errors. Moreover, these data sets can be quite large and have more than two axes of variation, e.g., sender, receiver, time. Many applications in those domains aim to capture the underlying latent structure of the data; in other words, they need to factorize data sets with missing entries. If we cannot address the problem of missing data, many important data sets will be discarded or improperly analyzed. Therefore, we need a robust and scalable approach for factorizing multi-way arrays (i.e., tensors) in the presence of missing data. We focus on one of the most well-known tensor factorizations, CANDECOMP/PARAFAC (CP), and formulate the CP model as a weighted least squares problem that models only the known entries. We develop an algorithm called CP-WOPT (CP Weighted OPTimization) using a first-order optimization approach to solve the weighted least squares problem. Based on extensive numerical experiments, our algorithm is shown to successfully factor tensors with noise and up to 70% missing data. Moreover, our approach is significantly faster than the leading alternative and scales to larger problems. To show the real-world usefulness of CP-WOPT, we illustrate its applicability on a novel EEG (electroencephalogram) application where missing data is frequently encountered due to disconnections of electrodes.

99 GENERAL AND MISCELLANEOUS↗

Scalable tensor factorizations with incomplete data.

The problem of incomplete data - i.e., data with missing or unknown values - in multi-way arrays is ubiquitous in biomedical signal processing, network traffic analysis, bibliometrics, social network analysis, chemometrics, computer vision, communication networks, etc. We consider the problem of how to factorize data sets with missing values with the goal of capturing the underlying latent structure of the data and possibly reconstructing missing values (i.e., tensor completion). We focus on one of the most well-known tensor factorizations that captures multi-linear structure, CANDECOMP/PARAFAC (CP). In the presence of missing data, CP can be formulated as a weighted least squares problem that models only the known entries. We develop an algorithm called CP-WOPT (CP Weighted OPTimization) that uses a first-order optimization approach to solve the weighted least squares problem. Based on extensive numerical experiments, our algorithm is shown to successfully factorize tensors with noise and up to 99% missing data. A unique aspect of our approach is that it scales to sparse large-scale data, e.g., 1000 x 1000 x 1000 with five million known entries (0.5% dense). We further demonstrate the usefulness of CP-WOPT on two real-world applications: a novel EEG (electroencephalogram) application where missing data is frequently encountered due to disconnections of electrodes and the problem of modeling computer network traffic where data may be absent due to the expense of the data collection process.

99 GENERAL AND MISCELLANEOUS↗

Preconditioning for Hyper-reduction in Reduced Order Models

Many projection-based reduced order models (pROM) that utilize the governing equation and data to accelerate physical simulations can be applied to nonlinear dynamical systems. To avoid full order model (FOM) scale update for each time step, hyperreduction techniques are developed to sample high dimensional nonlinear terms. Our study aims to investigate if preconditioning the least-squares problem used for the nonlinear approximation can improve the robustness of the condition number of the problem while achieving high accuracy. In our study, we use the row-normalization matrix motivated by the Christoffel function as the preconditioner and solve the corresponding weighted least-squares problem. Numerical results for Lagrangian hydrodynamics examples are analyzed to explore how the preconditioner works compared to existing hyper-reduction techniques.

97 MATHEMATICS AND COMPUTING↗

A balanced submatrix merging algorithm for multiprocessor architectures

In this article, a parallel algorithm which applies Givens rotations to selectively annihilate k(k + 1)/2 nonzero elements from two k x n(k not more than n) upper trapezoidal submatrices is described. The new algorithm is suitable for implementation on either a pair of directly connected local-memory processors or two clusters of multiple tightly-coupled processors. Analyses show that in both cases the proposed algorithms achieve optimal speed-up by balancing the work load distribution and masking interprocessor or intercluster communication by computation if k is much small than n. In the context of solving large scale least squares problems, this submatrix merging step is repetitively needed during the entire computation and, furthermore, there are usually many pairs of such submatrices to be merged with each submatrix stored in the memory of a processor or a cluster of processors. The proposed algorithm can be applied to each pair of submatrices concurrently, and thus parallelizes an important step in solving the least squares problems.

Chu, Eleanor↗

LOCO with Constraints and Improved Fitting Technique

LOCO has been a powerful beam-based diagnostics and optics control method for storage rings and synchrotrons worldwide ever since it was established at NSLS by J. Safranek. This method measures the orbit response matrix and optionally the dispersion function of the machine. The data are then fitted to a lattice model by adjusting parameters such as quadrupole and skew quadrupole strengths in the model, BPM gains and rolls, corrector gains and rolls of the measurement system. Any abnormality of the machine that affects the machine optics can then be identified. The resulting lattice model is equivalent to the real machine lattice as seen by the BPMs. Since there are usually two or more BPMs per betatron period in modern circular accelerators, the model is often a very accurate representation of the real machine. According to the fitting result, one can correct the machine lattice to the design lattice by changing the quadrupole and skew quadrupole strengths. LOCO is so important that it is routinely performed at many electron storage rings to guarantee machine performance, especially after the Matlab-based LOCO code became available. However, for some machines, LOCO is not easy to carry out. In some cases, LOCO fitting converges to an unrealistic solution with large changes to the quadrupole strengths {Delta}K. The quadrupole gradient changes can be so large that the resulting lattice model fails to find a closed orbit and subsequent iterations become impossible. In cases when LOCO converges, the solution can have {Delta}K that is larger than realistic and often along with a spurious zigzag pattern between adjacent quadrupoles. This degeneracy behavior of LOCO is due to the correlation between the fitting parameters - usually between neighboring quadrupoles. The fitting scheme is therefore less restrictive over certain patterns of changes to these quadrupoles with which the correlated quadrupoles fight each other and the net effect is very inefficient {chi}{sup 2} reduction, i.e., small {chi}{sup 2} reduction with large changes of {Delta}K. Under effects of random noise, the fitting solution tends to crawl toward these patterns and ends up with unrealistically large {Delta}K. Such a solution is not very useful in optics correction because after the solution is dialed in, the quadrupoles will not respond as predicted by the lattice model due to magnet hysteresis. We will show that adding constraints to the fitting parameters is an effective way to combat this problem of LOCO. In fact, it improves optics calibration precision even for machines that don't show severe degeneracy behavior. LOCO fitting is essentially to solve a nonlinear least square problem with an iterative approach. The linear least square technique is applied in each iteration to move the solution toward the minimum. This approach is commonly referred to as the Gauss-Newton method. By using singular value decomposition (SVD) to invert the Jacobian matrix, this method has generally been very successful for LOCO. However, this method is based on a linear expansion of the residual vector over the fitting parameters which is valid only when the starting solution is sufficiently close to the real minimum. The fitting algorithm can have difficulties to converge when the initial guess is too far off. For example, it's possible for the {chi}{sup 2} merit function to increase after an iteration instead of decrease. This situation can be improved by using more robust nonlinear least square fitting algorithms, such as the Levenberg-Marquardt method. We will discuss the degeneracy problem in section 2 and then show how the constrained fitting can help in section 3. The application of Levenberg-Marquadt method to LOCO is shown in section 4. A summary is given in section 5.

43 PARTICLE ACCELERATORS↗

Calibration of Gyros with Temperature Dependent Scale Factors

The general problem of gyro calibration can be stated as the estimation of the scale factors, misalignments, and drift-rate biases of the gyro using the on-orbit sensor measurements. These gyro parameters have been traditionally treated as temperature-independent in the operational flight dynamics ground systems at NASA Goddard Space Flight Center (GSFC), a scenario which has been successfully applied in the gyro calibration of a large number of missions. A significant departure from this is the Microwave Anisotropy Probe (MAP) mission where, due to the high thermal variations expected during the mission phase, it is necessary to model the scale factors as functions of temperature. This paper addresses the issue of gyro calibration for the MAP gyro model using a manufacturer-supplied model of the variation of scale factors with temperature. The problem is formulated as a least squares problem and solved using the Levenberg-Marquardt algorithm in the MATLAB(R) library function NLSQ. The algorithm was tested on simulated data with Gaussian noise for the quaternions as well as the gyro rates and was found to consistently converge close to the true values. Significant improvement in accuracy was noticed due to the estimation of the temperature-dependent scale factors as against constant scale factors.

Belur, Sheela V.↗

P- and L-Band Retrieval of Subsurface Soil Moisture and Temperature Profiles as First-Order Polynomial Function

This paper demonstrates the potential use of P and L band passive measurements to determine root zone soil moisture (SM) and soil temperature(ST). SM and ST data have been taken as a function of depth during the NASA GSFC PLEX19 experiment in the summer of 2019 at Beltsville, MD, USA. Using these data, a coherent model has been used to compute H and V brightness temperatures at frequencies of 0.8 and 1.4 GHz with an observation angle of 35 degrees. These synthetic brightness data are then used to estimate the SM and ST profiles which are represented by linear polynomials. The inversion problem is formulated as a least square problem that is solved by a global optimization method known as the Adaptive Simulated Annealing(ASA) method. Four inversion examples having different SM and ST profiles are presented. Selected results show that the standard deviation between the retrieved and measured data is less than 0.077 cm3/cm3 for SM, and 2.245 °C for ST.

Ming Li↗

Optimal model reduction and frequency-weighted extension

In this paper the quadratically optimal model reduction problem for single-input, single-output systems is considered. The reduced order model is determined by minimizing the integral of the magnitude-squared of the transfer function error. It is shown that the numerator coefficients of the optimal approximant satisfy a weighted least squares problem and, on this basis, a two-step iterative algorithm is developed combining a least squares solver with a gradient minimizer. The existence of globally optimal stable solutions to the optimization problem is established, and convergence of the algorithm to stationary values of the cost function is proved. The formulation is extended to handle the frequency-weighted optimal model reduction problem. Three examples demonstrate the optimization algorithm.

Spanos, J. T.↗

Least-squares finite elements for Stokes problem

A least-squares method based on the first-order velocity-pressure-vorticity formulation for the Stokes problem is proposed. This method leads to a minimization problem rather than to a saddle-point problem. The choice of the combinations of elements is thus not subject to the Ladyzhenskaya-Babuska-Brezzi (LBB) condition. Numerical results are given for the optimal rate of convergence for equal-order interpolations.

Jiang, Bo-Nan↗

Time and Frequency Domain Methods for Basis Selection in Random Linear Dynamical Systems

Polynomial chaos methods have been extensively used to analyze systems in uncertainty quantification. Furthermore, several approaches exist to determine a low-dimensional approximation (or sparse approximation) for some quantity of interest in a model, where just a few orthogonal basis polynomials are required. In this work, we consider linear dynamical systems consisting of ordinary differential equations with random variables. The aim of this paper is to explore methods for producing low-dimensional approximations of the quantity of interest further. We investigate two numerical techniques to compute a low-dimensional representation, which both fit the approximation to a set of samples in the time domain. On the one hand, a frequency domain analysis of a stochastic Galerkin system yields the selection of the basis polynomials. It follows a linear least squares problem. On the other hand, a sparse minimization yields the choice of the basis polynomials by information from the time domain only. An orthogonal matching pursuit produces an approximate solution of the minimization problem. Finally, we compare the two approaches using a test example from a mechanical application.

97 MATHEMATICS AND COMPUTING↗

A new algorithm for L2 optimal model reduction

In this paper the quadratically optimal model reduction problem for single-input, single-output systems is considered. The reduced order model is determined by minimizing the integral of the magnitude-squared of the transfer function error. It is shown that the numerator coefficients of the optimal approximant satisfy a weighted least squares problem and, on this basis, a two-step iterative algorithm is developed combining a least squares solver with a gradient minimizer. Convergence of the proposed algorithm to stationary values of the quadratic cost function is proved. The formulation is extended to handle the frequency-weighted optimal model reduction problem. Three examples demonstrate the optimization algorithm.

Spanos, J. T.↗

A line search algorithm for wind field adjustment with incomplete data and RBF approximation

The problem of concern in this work is the construction of free divergence fields given scattered horizontal components. As customary, the problem is formulated as a PDE constrained least squares problem. The novelty of our approach is to construct the so-called adjusted field, as the unique solution along an appropriately chosen descent direction. The latter is obtained by the adjoint equation technique. It is shown that the classical adjusted field of Sasaki’s is a particular case. On choosing descent directions, the underlying mass consistent model leads to the solution of an elliptic problem which is solved by means of a radial basis functions method. Finally, some numerical results for wind field adjustment are presented.

97 MATHEMATICS AND COMPUTING↗