Search NASA⌕ Search

SEARCH · Search NASA

Results for “tensor 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 325 records · Page 18

Tensor hypercontracted ppRPA: Reducing the cost of the particle-particle random phase approximation from O(r {sup 6}) to O(r {sup 4})

In recent years, interest in the random-phase approximation (RPA) has grown rapidly. At the same time, tensor hypercontraction has emerged as an intriguing method to reduce the computational cost of electronic structure algorithms. In this paper, we combine the particle-particle random phase approximation with tensor hypercontraction to produce the tensor-hypercontracted particle-particle RPA (THC-ppRPA) algorithm. Unlike previous implementations of ppRPA which scale as O(r{sup 6}), the THC-ppRPA algorithm scales asymptotically as only O(r{sup 4}), albeit with a much larger prefactor than the traditional algorithm. We apply THC-ppRPA to several model systems and show that it yields the same results as traditional ppRPA to within mH accuracy. Our method opens the door to the development of post-Kohn Sham functionals based on ppRPA without the excessive asymptotic cost of traditional ppRPA implementations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Denoising atomic resolution 4D scanning transmission electron microscopy data with tensor singular value decomposition

Tensor singular value decomposition (SVD) is a method to find a low-dimensional representation of data with meaningful structure in three or more dimensions. Here, tensor SVD has been applied to denoise atomic-resolution 4D scanning transmission electron microscopy (4D STEM) data. On data simulated from a SrTiO 3 [100] perfect crystal and a Si [110] edge dislocation, tensor SVD achieved an average peak signal-to-noise ratio (PSNR) of ~40 dB, which matches or exceeds the performance of other denoising methods, with processing times at least 100 times shorter. On experimental data from SrTiO 3 [100] and LiZnSb [11 2 ¯ 0]/GaSb [110] samples, tensor SVD denoises multiple GB 4D STEM data sets in ten minutes on a typical personal computer. Denoising with tensor SVD improves both convergent beam electron diffraction patterns and virtual-aperture annular dark field images.

36 MATERIALS SCIENCE↗

A Low-Rank Solver for the Navier--Stokes Equations with Uncertain Viscosity

In this work, we study an iterative low-rank approximation method for the solution of the steady-state stochastic Navier--Stokes equations with uncertain viscosity. The method is based on linearization schemes using Picard and Newton iterations and stochastic finite element discretizations of the linearized problems. For computing the low-rank approximate solution, we adapt the nonlinear iterations to an inexact and low-rank variant, where the solution of the linear system at each nonlinear step is approximated by a quantity of low rank. This is achieved by using a tensor variant of the GMRES method as a solver for the linear systems. We explore the inexact low-rank nonlinear iteration with a set of benchmark problems, using a model of flow over an obstacle, under various configurations characterizing the statistical features of the uncertain viscosity, and we demonstrate its effectiveness by extensive numerical experiments.

97 MATHEMATICS AND COMPUTING↗

MO-C-17A-08: Evaluation of Lung Deformation Using Three Dimensional Strain Maps

Purpose: To develop a systematic approach to generate three dimensional (3D) strain maps of lung using the displacement vector field (DVF) during the respiratory deformation, and to demonstrate its application in evaluating deformable image registration (DIR). Methods: A DVF based strain tensor at each voxel of interest (VOI) was calculated from the relative displacements between the VOI and each of the six nearest neighbors. The maximum and minimum stretches of a VOI can be determined by the principal strains (E{sub 1}, E{sub 2} and E{sub 3}), which are the eigenvalues and the corresponding strain tensors. Two healthy volunteers enrolled in this study under IRB-approved protocol, each was scanned using 3D Hyperpolarized He-3 tagging-MRI and 3D proton-MRI with TrueFISP sequence at the endof- inhalation (EOI) and the end-of-exhalation (EOE) phases. 3D DVFs of tagging- and proton-MRI were obtained by the direct measurements of the tagging grid trajectory and by the DIR method implemented in commercial software. Results: 3D strain maps were successfully generated for all DVFs. The principal strain E1s were calculated as 0.43±0.05 and 0.17±0.25 for tagging-MRI and proton-MRI, respectively. The large values of E{sub 1} indicate the predominant lung motion in the superior-inferior (SI) direction. Given that the DVFs from the tagging images are considered as the ground truth, the discrepancies in the DIR-based strain maps suggest the inaccuracy of the DIR algorithm. In the E{sub 1} maps of tagging-MRI for subject 1, the fissures were distinguishable by the larger values (0.49±0.02) from the adjacent tissues (0.41±0.03) due to the larger relative displacement between the lung lobes. Conclusion: We have successfully developed a methodology to generate DVF-based 3D strain maps of lung. It can potentially enable us to better understand the pulmonary biomechanics and to evaluate and improve the DIR algorithms for the lung deformation. We are currently studying more subjects to evaluate this tool.

60 APPLIED LIFE SCIENCES↗

Differences in Brainstem Fiber Tract Response to Radiation: A Longitudinal Diffusion Tensor Imaging Study

Purpose: To determine whether radiation-induced changes in white matter tracts are uniform across the brainstem. Methods and Materials: We analyzed serial diffusion tensor imaging data, acquired before radiation therapy and over 48 to 72 months of follow-up, from 42 pediatric patients (age 6-20 years) with medulloblastoma. FSL software (FMRIB, Oxford, UK) was used to calculate fractional anisotropy (FA) and axial, radial, and mean diffusivities. For a consistent identification of volumes of interest (VOIs), the parametric maps of each patient were transformed to a standard brain space (MNI152), on which we identified VOIs including corticospinal tract (CST), medial lemniscus (ML), transverse pontine fiber (TPF), and middle cerebellar peduncle (MCP) at the level of pons. Temporal changes of DTI parameters in VOIs were compared using a linear mixed effect model. Results: Radiation-induced white matter injury was marked by a decline in FA after treatment. The decline was often accompanied by decreased axial diffusivity, increased radial diffusivity, or both. This implied axonal damage and demyelination. We observed that the magnitude of the changes was not always uniform across substructures of the brainstem. Specifically, the changes in DTI parameters for TPF were more pronounced than in other regions (P<.001 for FA) despite similarities in the distribution of dose. We did not find a significant difference among CST, ML, and MCP in these patients (P>.093 for all parameters). Conclusions: Changes in the structural integrity of white matter tracts, assessed by DTI, were not uniform across the brainstem after radiation therapy. These results support a role for tract-based assessment in radiation treatment planning and determination of brainstem tolerance.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Analytical effective tensor for flow-through composites

A machine, method and computer-usable medium for modeling an average flow of a substance through a composite material. Such a modeling includes an analytical calculation of an effective tensor K.sup.a suitable for use with a variety of media. The analytical calculation corresponds to an approximation to the tensor K, and follows by first computing the diagonal values, and then identifying symmetries of the heterogeneity distribution. Additional calculations include determining the center of mass of the heterogeneous cell and its angle according to a defined Cartesian system, and utilizing this angle into a rotation formula to compute the off-diagonal values and determining its sign.

42 ENGINEERING↗

Towards Compact Neural Networks via End-to-End Training: A Bayesian Tensor Approach with Automatic Rank Determination

Post-training model compression can reduce the inference costs of deep neural networks, but uncompressed training still consumes enormous hardware resources and energy. To enable low-energy training on edge devices, it is highly desirable to directly train a compact neural network from scratch with a low memory cost. Low-rank tensor decomposition is an effective approach to reduce the memory and computing costs of large neural networks. However, directly training low-rank tensorized neural networks is a very challenging task because it is hard to determine a proper tensor rank a priori, and the tensor rank controls both model complexity and accuracy. Here, this paper presents a novel end-to-end framework for low-rank tensorized training. We first develop a Bayesian model that supports various low-rank tensor formats (e.g., CANDECOMP/PARAFAC, Tucker, tensor-train, and tensor-train matrix) and reduces neural network parameters with automatic rank determination during training. Then we develop a customized Bayesian solver to train large-scale tensorized neural networks. Our training methods shows orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the experiments. On a very large deep learning recommendation system with over 4.2 ×10 9 model parameters, our method can reduce the parameter number to 1.6 ×10 5 automatically in the training process (i.e., by 2.6 ×10 4 times) while achieving almost the same accuracy. Code is available at https://github.com/colehawkins/bayesian-tensor-rank-determination.

compact neural networks↗

Inclusive reactions from finite Minkowski spacetime correlation functions

The need to determine scattering amplitudes of few-hadron systems for arbitrary kinematics expands a broad set of subfields of modern-day nuclear and hadronic physics. In this work, we expand upon previous explorations on the use of real-time methods, like quantum computing or tensor networks, to determine few-body scattering amplitudes. Such calculations must be performed in a finite Minkowski spacetime, where scattering amplitudes are not well defined. Our previous work presented a conjecture of a systematically improvable estimator for scattering amplitudes constructed from finite-volume correlation functions. Here we provide further evidence that the prescription works for larger kinematic regions than previously explored as well as a broader class of scattering amplitudes. Finally, we devise a new method for estimating the order of magnitude of the error associated with finite time separations needed for such calculations. In units of the lightest mass of the theory, we find that to constrain amplitudes using real-time methods within O ( 10 % ) , the spacetime volumes must satisfy m L ∼ O ( 10 – 10 2 ) ) and m T ∼ O ( 10 2 – 10 4 ) . Published by the American Physical Society 2024

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Mean-field entanglement transitions in random tree tensor networks

Entanglement phase transitions in quantum chaotic systems subject to projective measurements and in random tensor networks have emerged as a new class of critical points separating phases with different entanglement scaling. We propose a mean-field theory of such transitions by studying the entanglement properties of random tree tensor networks. As a function of bond dimension, we find a phase transition separating area-law from logarithmic scaling of the entanglement entropy. Using a mapping onto a replica statistical mechanics model defined on a Cayley tree and the cavity method, we analyze the scaling properties of such transitions. Our approach provides a tractable, mean-field-like example of an entanglement transition. Furthermore, we verify our predictions numerically by computing directly the entanglement of random tree tensor network states.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Sparse tensor spherical harmonics approximation in radiative transfer

The stationary monochromatic radiative transfer equation is a partial differential transport equation stated on a five-dimensional phase space. To obtain a well-posed problem, boundary conditions have to be prescribed on the inflow part of the domain boundary. We solve the equation with a multi-level Galerkin FEM in physical space and a spectral discretization with harmonics in solid angle and show that the benefits of the concept of sparse tensor products, known from the context of sparse grids, can also be leveraged in combination with a spectral discretization. Our method allows us to include high spectral orders without incurring the 'curse of dimension' of a five-dimensional computational domain. Neglecting boundary conditions, we find analytically that for smooth solutions, the convergence rate of the full tensor product method is retained in our method up to a logarithmic factor, while the number of degrees of freedom grows essentially only as fast as for the purely spatial problem. For the case with boundary conditions, we propose a splitting of the physical function space and a conforming tensorization. Numerical experiments in two physical and one angular dimension show evidence for the theoretical convergence rates to hold in the latter case as well.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Experimental evaluation of electrical conductivity imaging of anisotropic brain tissues using a combination of diffusion tensor imaging and magnetic resonance electrical impedance tomography

Anisotropy of biological tissues is a low-frequency phenomenon that is associated with the function and structure of cell membranes. Imaging of anisotropic conductivity has potential for the analysis of interactions between electromagnetic fields and biological systems, such as the prediction of current pathways in electrical stimulation therapy. To improve application to the clinical environment, precise approaches are required to understand the exact responses inside the human body subjected to the stimulated currents. In this study, we experimentally evaluate the anisotropic conductivity tensor distribution of canine brain tissues, using a recently developed diffusion tensor-magnetic resonance electrical impedance tomography method. At low frequency, electrical conductivity of the biological tissues can be expressed as a product of the mobility and concentration of ions in the extracellular space. From diffusion tensor images of the brain, we can obtain directional information on diffusive movements of water molecules, which correspond to the mobility of ions. The position dependent scale factor, which provides information on ion concentration, was successfully calculated from the magnetic flux density, to obtain the equivalent conductivity tensor. By combining the information from both techniques, we can finally reconstruct the anisotropic conductivity tensor images of brain tissues. The reconstructed conductivity images better demonstrate the enhanced signal intensity in strongly anisotropic brain regions, compared with those resulting from previous methods using a global scale factor.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Rank-reduced coupled-cluster. III. Tensor hypercontraction of the doubles amplitudes

In this work, we develop a quartic-scaling implementation of coupled-cluster singles and doubles (CCSD) based on low-rank tensor hypercontraction (THC) factorizations of both the electron repulsion integrals (ERIs) and the doubles amplitudes. This extends our rank-reduced (RR) coupled-cluster method to incorporate higher-order tensor factorizations. The THC factorization of the doubles amplitudes accounts for most of the gain in computational efficiency as it is sufficient, in conjunction with a Cholesky decomposition of the ERIs, to reduce the computational complexity of most contributions to the CCSD amplitude equations. Further THC factorization of the ERIs reduces the complexity of certain terms arising from nested commutators between the doubles excitation operator and the two-electron operator. We implement this new algorithm using graphical processing units and demonstrate that it enables CCSD calculations for molecules with 250 atoms and 2500 basis functions using a single computer node. Furthermore, we show that the new method computes correlation energies with comparable accuracy to the underlying RR-CCSD method.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Signal-to-noise assessment for diffusion tensor imaging with single data set and validation using a difference image method with data from a multicenter study

Purpose: To describe a quantitative method for determination of SNR that extracts the local noise level using a single diffusion data set. Methods: Brain data sets came from a multicenter study (eight sites; three MR vendors). Data acquisition protocol required b = 0, 700 s/mm{sup 2}, fov = 256 × 256 mm{sup 2}, acquisition matrix size 128 × 128, reconstruction matrix size 256 × 256; 30 gradient encoding directions and voxel size 2 × 2 × 2 mm{sup 3}. Regions-of-interest (ROI) were placed manually on the b = 0 image volume on transverse slices, and signal was recorded as the mean value of the ROI. The noise level from the ROI was evaluated using Fourier Transform based Butterworth high-pass filtering. Patients were divided into two groups, one for filter parameter optimization (N = 17) and one for validation (N = 10). Six white matter areas (the genu and splenium of corpus callosum, right and left centrum semiovale, right and left anterior corona radiata) were analyzed. The Bland–Altman method was used to compare the resulting SNR with that from the difference image method. The filter parameters were optimized for each brain area, and a set of “global” parameters was also obtained, which represent an average of all regions. Results: The Bland–Altman analysis on the validation group using “global” filter parameters revealed that the 95% limits of agreement of percent bias between the SNR obtained with the new and the reference methods were −15.5% (median of the lower limit, range [−24.1%, −8.9%]) and 14.5% (median of the higher limits, range [12.7%, 18.0%]) for the 6 brain areas. Conclusions: An FT-based high-pass filtering method can be used for local area SNR assessment using only one DTI data set. This method could be used to evaluate SNR for patient studies in a multicenter setting.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A minimal set of invariants as a systematic approach to higher order gravity models

Higher-order gravity models have been recently the subject of much attention in the context of cosmic acceleration. These models are derived by adding various curvature invariants to the Einstein-Hilbert action. Several studies showed that these models can have late-time self-acceleration and could, in some cases, fit various observational constraints. In view of the infinite spectrum of invariants that could be built from curvature tensors, we propose here a method based on minimal sets of independent invariants as a possible route for a systematic approach to these models. We explore a connection made between theorems on bases in invariants theory in relativity and higher-order cosmological models. A dynamical system analysis is performed and some models with accelerating attractors are discussed. The asymptotic behavior of the models is also studied using analytical calculations.

79 ASTRONOMY AND ASTROPHYSICS↗

Highly Efficient and Scalable Compound Decomposition of Two-Electron Integral Tensor and Its Application in Coupled Cluster Calculations

The representation and storage of two-electron integral tensors are vital in large- scale applications of accurate electronic structure methods. Low-rank representation and efficient storage strategy of integral tensors can significantly reduce the numerical overhead and consequently time-to-solution of these methods. In this paper, by combining pivoted incomplete Cholesky decomposition (CD) with a follow-up truncated singular vector decomposition (SVD), we develop a decomposition strategy to approximately represent the two-electron integral tensor in terms of low-rank vectors. A systematic benchmark test on a series of 1-D, 2-D, and 3-D carbon-hydrogen systems demonstrates high efficiency and scalability of the compound two-step decomposition of the two-electron integral tensor in our implementation. For the size of atomic basis set N_b ranging from ~ 100 up to ~ 2, 000, the observed numerical scaling of our implementation shows O(N_b^{2.5~3}) versus O(N_b^{3~4}) of single CD in most of other implementations. More importantly, this decomposition strategy can significantly reduce the storage requirement of the atomic-orbital (AO) two-electron integral tensor from O(N_b^4) to O(N_b^2 log_{10}(N_b)) with moderate decomposition thresholds. The accuracy tests have been performed using ground- and excited-state formulations of coupled- cluster formalism employing single and double excitations (CCSD) on several bench- mark systems including the C_{60} molecule described by nearly 1,400 basis functions. The results show that the decomposition thresholds can be generally set to 10^{-4} to 10^{-3} to give acceptable compromise between efficiency and accuracy.

54 ENVIRONMENTAL SCIENCES↗

Full-sky formulae for weak lensing power spectra from total angular momentum method

We systematically derive full-sky formulae for the weak lensing power spectra generated by scalar, vector and tensor perturbations from the total angular momentum (TAM) method. Based on both the geodesic and geodesic deviation equations, we first give the gauge-invariant expressions for the deflection angle and Jacobi map as observables of the CMB lensing and cosmic shear experiments. We then apply the TAM method, originally developed in the theoretical studies of CMB, to a systematic derivation of the angular power spectra. The TAM representation, which characterizes the total angular dependence of the spatial modes projected along a line-of-sight, can carry all the information of the lensing modes generated by scalar, vector, and tensor metric perturbations. This greatly simplifies the calculation, and we present a complete set of the full-sky formulae for angular power spectra in both the E-/B-mode cosmic shear and gradient-/curl-mode lensing potential of deflection angle. Based on the formulae, we give illustrative examples of non-vanishing B-mode cosmic shear and curl-mode of deflection angle in the presence of the vector and tensor perturbations, and explicitly compute the power spectra.

79 ASTRONOMY AND ASTROPHYSICS↗

Frontal Slice Approaches for Tensor Linear Systems

Inspired by the row and column action methods for solving large-scale linear systems, in this work, we explore the use of frontal slices for solving tensor linear systems. In particular, this paper presents a novel approach for using frontal slices of a tensor $\mathcal{A}$ to solve tensor linear systems $\mathcal{A} ∗\mathcal{X} = \mathcal{B}$ where ∗ denotes the $t$-product. In addition, we consider variations of this method, including cyclic, block, and randomized approaches, each designed to optimize performance in different operational contexts. Our primary contribution lies in the development and convergence analysis of these methods. Experimental results on synthetically generated and real-world data, including applications such as image and video deblurring, demonstrate the efficacy of our proposed approaches and validate our theoretical findings.

Luo, Hengrui↗

TTDFT: A GPU accelerated Tucker tensor DFT code for large-scale Kohn-Sham DFT calculations

We present the Tucker tensor DFT (TTDFT) code which uses a tensor-structured algorithm with graphic processing unit (GPU) acceleration for conducting ground-state DFT calculations on large-scale systems. The Tucker tensor DFT algorithm uses a localized Tucker tensor basis computed from an additive separable approximation to the Kohn-Sham Hamiltonian. The discrete Kohn-Sham problem is solved using Chebyshev filtered subspace iteration method that relies on matrix-matrix multiplications of a sparse symmetric Hamiltonian matrix and a dense wavefunction matrix, expressed in the localized Tucker tensor basis. These matrix-matrix multiplication operations, which constitute the most computationally intensive step of the solution procedure, are GPU accelerated providing ~8-fold GPU-CPU speedup for these operations on the largest systems studied. In conclusion, the computational performance of the TTDFT code is presented using benchmark studies on aluminum nano-particles and silicon quantum dots with system sizes ranging up to ~7,000 atoms.

97 MATHEMATICS AND COMPUTING↗