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.

486 records · Page 27

A novel cosmic filament catalogue from SDSS data

Here, in this work, we present a new catalogue of cosmic filaments obtained from the latest Sloan Digital Sky Survey (SDSS) public data. In order to detect filaments, we implement a version of the Subspace-Constrained Mean-Shift algorithm that is boosted by machine learning techniques. This allows us to detect cosmic filaments as one-dimensional maxima in the galaxy density distribution. Our filament catalogue uses the cosmological sample of SDSS, including Data Release 16, and therefore inherits its sky footprint (aside from small border effects) and redshift coverage. In particular, this means that, taking advantage of the quasar sample, our filament reconstruction covers redshifts up to z = 2.2, making it one of the deepest filament reconstructions to our knowledge. We follow a tomographic approach and slice the galaxy data in 269 shells at different redshift. The reconstruction algorithm is applied to 2D spherical maps. The catalogue provides the position and uncertainty of each detection for each redshift slice. The quality of our detections, which we assess with several metrics, show improvement with respect to previous public catalogues obtained with similar methods. We also detect a highly significant correlation between our filament catalogue and galaxy cluster catalogues built from microwave observations of the Planck Satellite and the Atacama Cosmology Telescope.

79 ASTRONOMY AND ASTROPHYSICS↗

Three-Dimensional Full Core Power Calculations for Pressurized Water Reactors

We have implemented a new multilevel parallel decomposition in the Denovo discrete ordinates radiation transport code. In concert with Krylov subspace iterative solvers, the multilevel decomposition allows concurrency over energy in addition to space-angle. The original space-angle partitioning in Denovo placed an eective limit on the scalability of the transport solver that was highly dependent on the problem size. The added phase-space concurrency combined with the high-performance Krylov solvers has enabled weak scaling to 100K cores on the Jaguar XT5 supercomputer. Furthermore, the multilevel decomposition provides enough concurrency to scale to exascale computing and beyond.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Optimizing Kernel Machines Using Deep Learning

Building highly non-linear and non-parametric models is central to several state-of-the-art machine learning systems. Kernel methods form an important class of techniques that induce a reproducing kernel Hilbert space (RKHS) for inferring non-linear models through the construction of similarity functions from data. These methods are particularly preferred in cases where the training data sizes are limited and when prior knowledge of the data similarities is available. Despite their usefulness, they are limited by the computational complexity and their inability to support end-to-end learning with a task specific objective. On the other hand, deep neural networks have become the de facto solution for end-to-end inference in several learning paradigms. In this article, we explore the idea of using deep architectures to perform kernel machine optimization, for both computational efficiency and end-to-end inferencing. To this end, we develop the DKMO (Deep Kernel Machine Optimization) framework, that creates an ensemble of dense embeddings using Nystrõm kernel approximations and utilizes deep learning to generate task-specific representations through the fusion of the embeddings. Intuitively, the filters of the network are trained to fuse information from an ensemble of linear subspaces in the RKHS. Furthermore, we introduce the kernel dropout regularization to enable improved training convergence. Finally, we extend this framework to the multiple kernel case, by coupling a global fusion layer with pre-trained deep kernel machines for each of the constituent kernels. Using case studies with limited training data, and lack of explicit feature sources, we demonstrate the effectiveness of our framework over conventional model inferencing techniques.

97 MATHEMATICS AND COMPUTING↗

Non-intrusive reduced order modeling of natural convection in porous media using convolutional autoencoders: Comparison with linear subspace techniques

Natural convection in porous media is a highly nonlinear multiphysical problem relevant to many engineering applications (e.g., the process of CO 2 sequestration). Here, we extend and present a non-intrusive reduced order model of natural convection in porous media employing deep convolutional autoencoders for the compression and reconstruction and either radial basis function (RBF) interpolation or artificial neural networks (ANNs) for mapping parameters of partial differential equations (PDEs) on the corresponding nonlinear manifolds. To benchmark our approach, we also describe linear compression and reconstruction processes relying on proper orthogonal decomposition (POD) and ANNs. Further, we present comprehensive comparisons among different models through three benchmark problems. The reduced order models, linear and nonlinear approaches, are much faster than the finite element model, obtaining a maximum speed-up of 7 × 10 6 because our framework is not bound by the Courant–Friedrichs–Lewy condition; hence, it could deliver quantities of interest at any given time contrary to the finite element model. Our model’s accuracy still lies within a relative error of 7% in the worst-case scenario. We illustrate that, in specific settings, the nonlinear approach outperforms its linear counterpart and vice versa. We hypothesize that a visual comparison between principal component analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE) could indicate which method will perform better prior to employing any specific compression strategy.

97 MATHEMATICS AND COMPUTING↗

Multigrid in energy preconditioner for Krylov solvers

We have added a new multigrid in energy (MGE) preconditioner to the Denovo discrete-ordinates radiation transport code. This preconditioner takes advantage of a new multilevel parallel decomposition. A multigroup Krylov subspace iterative solver that is decomposed in energy as well as space-angle forms the backbone of the transport solves in Denovo. The space-angle-energy decomposition facilitates scaling to hundreds of thousands of cores. The multigrid in energy preconditioner scales well in the energy dimension and significantly reduces the number of Krylov iterations required for convergence. This preconditioner is well-suited for use with advanced eigenvalue solvers such as Rayleigh Quotient Iteration and Arnoldi.

97 MATHEMATICS AND COMPUTING↗

Lyapunov control on quantum open systems in decoherence-free subspaces

A scheme to drive and manipulate a finite-dimensional quantum system in decoherence-free subspaces (DFSs) by Lyapunov control is proposed. Control fields are established by the Lyapunov function. This proposal can drive the open quantum system into DFSs and manipulate it to any desired eigenstate of the free Hamiltonian. An example which consists of a four-level system with three long-lived states driven by two lasers is presented to demonstrate the scheme. We have performed numerical simulations of the dynamics of the four-level system, which show that the scheme works well.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Atomic-position independent descriptor for machine learning of material properties

The high-throughput screening of periodic inorganic solids using machine learning methods requires atomic positions to encode structural and compositional details into appropriate material descriptors. These atomic positions are not available a priori for new materials, which severely limits exploration of novel materials. In this work, we overcome this limitation by using only crystallographic symmetry information in the structural description of materials. We show that for materials with identical structural symmetry, machine learning is trivial, and accuracies similar to that of density functional theory calculations can be achieved by using only atomic numbers in the material description. For machine learning of formation energies of bulk crystalline solids, this simple material descriptor is able to achieve prediction mean absolute errors of only 0.07 eV/at on a test dataset consisting of more than 85 000 diverse materials. Finally, this atomic-position independent material descriptor presents a new route of materials discovery wherein millions of materials can be screened by training a machine learning model over a drastically reduced subspace of materials.

36 MATERIALS SCIENCE↗

Easy-plane anisotropic-exchange magnets on a honeycomb lattice: Quantum effects and dealing with them

We provide analytical and numerical insights into the phase diagram and other properties of the extended Kitaev-Heisenberg model on the honeycomb lattice in the easy-plane limit, in which interactions are only between spin components that belong to the plane of magnetic ions. This parameter subspace allows for a much-needed systematic quantitative investigation of spin excitations in the ordered phases and of their generic features. Specifically, we demonstrate that in this limit one can consistently take into account magnon interactions in both zero-field zigzag and field-polarized phases. For the nominally polarized phase, we propose a regularization of the unphysical divergences that occur at the critical field and are plaguing the 1/S approximation in this class of models. For the explored parameter subspace, all symmetry-allowed terms of the standard parametrization of the extended Kitaev-Heisenberg model, such as K, J, and Γ, are significant, making the offered consideration relevant to a much wider parameter space. Furthermore, the dynamical structure factor near the paramagnetic critical point illustrates this relevance by showing features that are reminiscent of the ones observed in α–RuCl 3 , underscoring that they are not unique and should be common to a wide range of parameters of the model and, by extension, to other materials.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Equivalence between Fermion-to-Qubit Mappings in two Spatial Dimensions

We argue that all locality-preserving mappings between fermionic observables and Pauli matrices on a two-dimensional lattice can be generated from the exact bosonization in Chen et al., whose gauge constraints project onto the subspace of the toric code with emergent fermions. Starting from the exact bosonization and applying Clifford finite-depth generalized local unitary transformation, we can achieve all possible fermion-to-qubit mappings (up to the re-pairing of Majorana fermions). In particular, we discover a new supercompact encoding using 1.25 qubits per fermion on the square lattice. We prove the existence of finite-depth quantum circuits to obtain fermion-to-qubit mappings with qubit-fermion ratios r = 1 + 1/2k for positive integers k, utilizing the trivialness of quantum cellular automata in two spatial dimensions. Also, we provide direct constructions of fermion-to-qubit mappings with ratios arbitrarily close to 1. When the ratio reaches 1, the fermion-to-qubit mapping reduces to the one-dimensional Jordan-Wigner transformation along a certain path in the two-dimensional lattice. Finally, we explicitly demonstrate that the Bravyi-Kitaev superfast simulation, the Verstraete-Cirac auxiliary method, Kitaev’s exactly solved model, the Majorana loop stabilizer codes, and the compact fermion-to-qubit mapping can all be obtained from the exact bosonization.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Machine learning strategy for accelerated design of polymer dielectrics

The ability to efficiently design new and advanced dielectric polymers is hampered by the lack of sufficient, reliable data on wide polymer chemical spaces, and the difficulty of generating such data given time and computational/experimental constraints. Here, we address the issue of accelerating polymer dielectrics design by extracting learning models from data generated by accurate state-of-the-art first principles computations for polymers occupying an important part of the chemical subspace. The polymers are ‘fingerprinted’ as simple, easily attainable numerical representations, which are mapped to the properties of interest using a machine learning algorithm to develop an on-demand property prediction model. Further, a genetic algorithm is utilised to optimise polymer constituent blocks in an evolutionary manner, thus directly leading to the design of polymers with given target properties. Furthermore, while this philosophy of learning to make instant predictions and design is demonstrated here for the example of polymer dielectrics, it is equally applicable to other classes of materials as well.

36 MATERIALS SCIENCE↗

Spinodal enhancement of fluctuations in nucleus-nucleus collisions

Subensemble Acceptance Method (SAM) [1, 2] is an essential link between measured event-by-event fluctuations and their grand canonical theoretical predictions such as lattice QCD. The method allows quantifying the global conservation law effects in fluctuations. In its basic formulation, SAM requires a sufficiently large system such as created in central nucleus-nucleus collisions and sufficient space-momentum correlations. Directly in the spinodal region of the First Order Phase Transition (FOPT) different approximations should be used that account for finite size effects. Thus, we present the generalization of SAM applicable in both the pure phases, metastable and unstable regions of the phase diagram [3]. Obtained analytic formulas indicate the enhancement of fluctuations due to crossing the spinodal region of FOPT and are tested using molecular dynamics simulations. A rather good agreement is observed. Using transport model calculations with interaction potential we show that the spinodal enhancement of fluctuations survives till the later stages of collision via the memory effect [4]. However, at low collision energies the space-momentum correlation is not strong enough for this signal to be transferred to second and third order cumulants measured in momentum subspace. This result agrees well with recent HADES data on proton number fluctuations at $\sqrt{S_{NN}}$ = 2.4 GeV which are found to be consistent with the binomial momentum space acceptance [5].

Poberezhnyuk, Roman↗

Polarization contributions to intermolecular interactions revisited with fragment electric-field response functions

The polarization energy in intermolecular interactions treated by self-consistent field electronic structure theory is often evaluated using a constraint that the atomic orbital (AO) to molecular orbital transformation is blocked by fragments. This approach is tied to AO basis sets, overestimates polarization energies in the overlapping regime, particularly in large AO basis sets, and lacks a useful complete basis set limit. These problems are addressed by the construction of polarization subspaces based on the responses of isolated fragments to weak electric fields. These subspaces are spanned by fragment electric-field response functions, which can capture effects up to the dipole (D), or quadrupole (DQ) level, or beyond. Schemes are presented for the creation of both non-orthogonal and orthogonal fragment subspaces, and the basis set convergence of the polarization energies computed using these spaces is assessed. Numerical calculations for the water dimer, water–Na{sup +}, water–Mg{sup 2+}, water–F{sup −}, and water–Cl{sup −} show that the non-orthogonal DQ model is very satisfactory, with small differences relative to the orthogonalized model. Additionally, we prove a fundamental difference between the polarization degrees of freedom in the fragment-blocked approaches and in constrained density schemes. Only the former are capable of properly prohibiting charge delocalization during polarization.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Efficient core-excited state orbital perspective on calculating x-ray absorption transitions in determinant framework

X-ray absorption spectroscopy (XAS) is an explicit probe of the unoccupied electronic structure of materials and an invaluable tool for fingerprinting various electronic properties and phenomena. Computational methods capable of simulating and analyzing such spectra are therefore in high demand for complementing experimental results and for extracting valuable insights therefrom. In particular, a recently proposed first-principles approach titled many-body XAS (MBXAS), which approximates the final (initial) state as a Slater determinant constructed from Kohn-Sham (KS) orbitals optimized in the absence (presence) of the relevant core electron has shown promising prospects in evaluating the transition amplitudes. In this article, we show that the MBXAS approach can be rederived using a transition operator expressed entirely in the basis of core-excited state KS orbitals and that this reformulation offers substantial practical and conceptual advantages. In addition to circumventing previous issues of convergence with respect to the number of unoccupied ground-state orbitals, the aforementioned representation reduces the computational expense by rendering the calculation of such orbitals unnecessary altogether. Furthermore, the reformulated approach also provides a direct pathway for comparing the many-body approximation with the so-called single-particle treatment and indicates the relative importance in observed XAS intensity of the relaxation of the valence occupied subspace induced by the core excitation. Finally, using the core-excited state basis, we define auxiliary orbitals for x-ray absorption and demonstrate their utility in explaining the spectral intensity by contrasting them with single-particle approximations to the excited state.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Canonical-ensemble state-averaged complete active space self-consistent field (SA-CASSCF) strategy for problems with more diabatic than adiabatic states: Charge-bond resonance in monomethine cyanines

This paper reviews basic results from a theory of the a priori classical probabilities (weights) in state-averaged complete active space self-consistent field (SA-CASSCF) models. It addresses how the classical probabilities limit the invariance of the self-consistency condition to transformations of the complete active space configuration interaction (CAS-CI) problem. Such transformations are of interest for choosing representations of the SA-CASSCF solution that are diabatic with respect to some interaction. I achieve the known result that a SA-CASSCF can be self-consistently transformed only within degenerate subspaces of the CAS-CI ensemble density matrix. For uniformly distributed (“microcanonical”) SA-CASSCF ensembles, self-consistency is invariant to any unitary CAS-CI transformation that acts locally on the ensemble support. Most SA-CASSCF applications in current literature are microcanonical. A problem with microcanonical SA-CASSCF models for problems with “more diabatic than adiabatic” states is described. The problem is that not all diabatic energies and couplings are self-consistently resolvable. A canonical-ensemble SA-CASSCF strategy is proposed to solve the problem. For canonical-ensemble SA-CASSCF, the equilibrated ensemble is a Boltzmann density matrix parametrized by its own CAS-CI Hamiltonian and a Lagrange multiplier acting as an inverse “temperature,” unrelated to the physical temperature. Like the convergence criterion for microcanonical-ensemble SA-CASSCF, the equilibration condition for canonical-ensemble SA-CASSCF is invariant to transformations that act locally on the ensemble CAS-CI density matrix. The advantage of a canonical-ensemble description is that more adiabatic states can be included in the support of the ensemble without running into convergence problems. The constraint on the dimensionality of the problem is relieved by the introduction of an energy constraint. The method is illustrated with a complete active space valence-bond (CASVB) analysis of the charge/bond resonance electronic structure of a monomethine cyanine: Michler’s hydrol blue. The diabatic CASVB representation is shown to vary weakly for “temperatures” corresponding to visible photon energies. Canonical-ensemble SA-CASSCF enables the resolution of energies and couplings for all covalent and ionic CASVB structures contributing to the SA-CASSCF ensemble. The CASVB solution describes resonance of charge- and bond-localized electronic structures interacting via bridge resonance superexchange. The resonance couplings can be separated into channels associated with either covalent charge delocalization or chemical bonding interactions, with the latter significantly stronger than the former.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A fast and accurate domain decomposition nonlinear manifold reduced order model

Here, this paper integrates nonlinear-manifold reduced order models (NM-ROMs) with domain decomposition (DD). NM ROMs approximate the full order model (FOM) state in a nonlinear-manifold by training a shallow, sparse autoencoder using FOM snapshot data. These NM-ROMs can be advantageous over linear-subspace ROMs (LS-ROMs) for problems with slowly decaying Kolmogorov n-width. However, the number of NM-ROM parameters that need to be trained scales with the size of the FOM. Moreover, for “extreme-scale” problems, the storage of high-dimensional FOM snapshots alone can make ROM training expensive. To alleviate the training cost, this paper applies DD to the FOM, computes NM-ROMs on each subdomain, and couples them to obtain a global NM-ROM. This approach has several advantages: Subdomain NM-ROMs can be trained in parallel, involve fewer parameters to be trained than global NM-ROMs, require smaller subdomain FOM dimensional training data, and can be tailored to subdomain specific features of the FOM. The shallow, sparse architecture of the autoencoder used in each subdomain NM-ROM allows application of hyper-reduction (HR), reducing the complexity caused by nonlinearity and yielding computational speedup of the NM-ROM. This paper provides the first application of NM-ROM (with HR) to a DD problem. In particular, this paper details an algebraic DD reformulation of the FOM, training a NM-ROM with HR for each sub domain, and a sequential quadratic programming (SQP) solver to evaluate the coupled global NM-ROM. Theoretical convergence results for the SQP method and a priori and a posteriori error estimates for the DD NM-ROM with HR are provided. The proposed DD NM-ROM with HR approach is numerically compared to a DD LS-ROM with HR on the 2D steady-state Burgers’ equation, showing an order of magnitude improvement in accuracy of the proposed DD NM-ROM over the DD LS-ROM.

97 MATHEMATICS AND COMPUTING↗

A projected preconditioned conjugate gradient algorithm for computing many extreme eigenpairs of a Hermitian matrix

We present an iterative algorithm for computing an invariant subspace associated with the algebraically smallest eigenvalues of a large sparse or structured Hermitian matrix A. We are interested in the case in which the dimension of the invariant subspace is large (e.g., over several hundreds or thousands) even though it may still be small relative to the dimension of A. These problems arise from, for example, density functional theory (DFT) based electronic structure calculations for complex materials. The key feature of our algorithm is that it performs fewer Rayleigh–Ritz calculations compared to existing algorithms such as the locally optimal block preconditioned conjugate gradient or the Davidson algorithm. It is a block algorithm, and hence can take advantage of efficient BLAS3 operations and be implemented with multiple levels of concurrency. We discuss a number of practical issues that must be addressed in order to implement the algorithm efficiently on a high performance computer.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Methanol synthesis on ZnO(0001{sup ¯}). IV. Reaction mechanisms and electronic structure

Methanol synthesis from CO and H{sub 2} over ZnO, which requires high temperatures and high pressures giving rise to a complex interplay of physical and chemical processes over this heterogeneous catalyst surface, is investigated using ab initio simulations. The redox properties of the surrounding gas phase are known to directly impact on the catalyst properties and thus, set the overall catalytic reactivity of this easily reducible oxide material. In Paper III of our series [J. Kiss, J. Frenzel, N. N. Nair, B. Meyer, and D. Marx, J. Chem. Phys. 134, 064710 (2011)] we have qualitatively shown that for the partially hydroxylated and defective ZnO(0001{sup ¯}) surface there exists an intricate network of surface chemical reactions. In the present study, we employ advanced molecular dynamics techniques to resolve in detail this reaction network in terms of elementary steps on the defective surface, which is in stepwise equilibrium with the gas phase. The two individual reduction steps were investigated by ab initio metadynamics sampling of free energy landscapes in three-dimensional reaction subspaces. By also sampling adsorption and desorption processes and thus molecular species that are in the gas phase but close to the surface, our approach successfully generated several alternative pathways of methanol synthesis. The obtained results suggest an Eley-Rideal mechanism for both reduction steps, thus involving “near-surface” molecules from the gas phase, to give methanol preferentially over a strongly reduced catalyst surface, while important side reactions are of Langmuir-Hinshelwood type. Catalyst re-reduction by H{sub 2} stemming from the gas phase is a crucial process after each reduction step in order to maintain the catalyst's activity toward methanol formation and to close the catalytic cycle in some reaction channels. Furthermore, the role of oxygen vacancies, side reactions, and spectator species is investigated and mechanistic details are discussed based on extensive electronic structure analysis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Gaussian processes with built-in dimensionality reduction: Applications to high-dimensional uncertainty propagation

Uncertainty quantification (UQ) tasks, such as model calibration, uncertainty propagation, and optimization under uncertainty, typically require several thousand evaluations of the underlying computer codes. To cope with the cost of simulations, one replaces the real response surface with a cheap surrogate based, e.g., on polynomial chaos expansions, neural networks, support vector machines, or Gaussian processes (GP). However, the number of simulations required to learn a generic multivariate response grows exponentially as the input dimension increases. This curse of dimensionality can only be addressed, if the response exhibits some special structure that can be discovered and exploited. A wide range of physical responses exhibit a special structure known as an active subspace (AS). An AS is a linear manifold of the stochastic space characterized by maximal response variation. The idea is that one should first identify this low dimensional manifold, project the high-dimensional input onto it, and then link the projection to the output. If the dimensionality of the AS is low enough, then learning the link function is a much easier problem than the original problem of learning a high-dimensional function. The classic approach to discovering the AS requires gradient information, a fact that severely limits its applicability. Furthermore, and partly because of its reliance to gradients, it is not able to handle noisy observations. The latter is an essential trait if one wants to be able to propagate uncertainty through stochastic simulators, e.g., through molecular dynamics codes. In this work, we develop a probabilistic version of AS which is gradient-free and robust to observational noise. Our approach relies on a novel Gaussian process regression with built-in dimensionality reduction. In particular, the AS is represented as an orthogonal projection matrix that serves as yet another covariance function hyper-parameter to be estimated from the data. To train the model, we design a two-step maximum likelihood optimization procedure that ensures the orthogonality of the projection matrix by exploiting recent results on the Stiefel manifold, i.e., the manifold of matrices with orthogonal columns. The additional benefit of our probabilistic formulation, is that it allows us to select the dimensionality of the AS via the Bayesian information criterion. We validate our approach by showing that it can discover the right AS in synthetic examples without gradient information using both noiseless and noisy observations. We demonstrate that our method is able to discover the same AS as the classical approach in a challenging one-hundred-dimensional problem involving an elliptic stochastic partial differential equation with random conductivity. Finally, we use our approach to study the effect of geometric and material uncertainties in the propagation of solitary waves in a one dimensional granular system.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗