Searcharxiv⌕ Search

arXiv subjects

L. Ridgway Scott

Publications and source records attributed to L. Ridgway Scott.

27 records · Page 2Linked to original sources
arXiv2021

Extrapolating the Arnoldi Algorithm To Improve Eigenvector Convergence

We consider extrapolation of the Arnoldi algorithm to accelerate computation of the dominant eigenvalue/eigenvector pair. The basic algorithm uses sequences of Krylov vectors to form a small eigenproblem which is solved exactly. The two dominant eigenvectors output from consecutive Arnoldi steps are then recombined to form an extrapolated iterate, and this accelerated iterate is used to restart the next Arnoldi process. We present numerical results testing the algorithm on a variety of cases and find on most examples it substantially improves the performance of restarted Arnoldi. The extrapolation is a simple post-processing step which has minimal computational cost.

math.NA↗
arXiv2020

Van der Waals interactions between two hydrogen atoms: The next orders

We extend a method (E. Cancès and L.R. Scott, SIAM J. Math. Anal., 50, 2018, 381--410) to compute more terms in the asymptotic expansion of the van der Waals attraction between two hydrogen atoms. These terms are obtained by solving a set of modified Slater-Kirkwood partial differential equations. The accuracy of the method is demonstrated by numerical simulations and comparison with other methods from the literature. It is also shown that the scattering states of the hydrogen atom, that are the states associated with the continuous spectrum of the Hamiltonian, have a major contribution to the C$_6$ coefficient of the van der Waals expansion.

math-ph↗
arXiv2019

Point Movement in a DSL for Higher-Order FEM Visualization

Scientific visualization tools tend to be flexible in some ways (e.g., for exploring isovalues) while restricted in other ways, such as working only on regular grids, or only on unstructured meshes (as used in the finite element method, FEM). Our work seeks to expose the common structure of visualization methods, apart from the specifics of how the fields being visualized are formed. Recognizing that previous approaches to FEM visualization depend on efficiently updating computed positions within a mesh, we took an existing visualization domain-specific language, and added a mesh position type and associated arithmetic operators. These are orthogonal to the visualization method itself, so existing programs for visualizing regular grid data work, with minimal changes, on higher-order FEM data. We reproduce the efficiency gains of an earlier guided search method of mesh position update for computing streamlines, and we demonstrate a novel ability to uniformly sample ridge surfaces of higher-order FEM solutions defined on curved meshes.

cs.GR↗
arXiv2016

A Nonlocal Poisson-Fermi Model for Ionic Solvent

We propose a nonlocal Poisson-Fermi model for ionic solvent that includes ion size effects and polarization correlations among water molecules in the calculation of electrostatic potential. It includes the previous Poisson-Fermi models as special cases, and its solution is the convolution of a solution of the corresponding nonlocal Poisson dielectric model with a Yukawa-type kernel function. Moreover, the Fermi distribution is shown to be a set of optimal ionic concentration functions in the sense of minimizing an electrostatic potential free energy. Finally, numerical results are reported to show the difference between a Poisson-Fermi solution and a corresponding Poisson solution.

physics.chem-ph↗
arXiv2016

A Stochastic Performance Model for Pipelined Krylov Methods

Pipelined Krylov methods seek to ameliorate the latency due to inner products necessary for projection by overlapping it with the computation associated with sparse matrix-vector multiplication. We clarify a folk theorem that this can only result in a speedup of $2\times$ over the naive implementation. Examining many repeated runs, we show that stochastic noise also contributes to the latency, and we model this using an analytical probability distribution. Our analysis shows that speedups greater than $2\times$ are possible with these algorithms.

cs.DC↗
arXiv2012

Optimizing the Evaluation of Finite Element Matrices

Assembling stiffness matrices represents a significant cost in many finite element computations. We address the question of optimizing the evaluation of these matrices. By finding redundant computations, we are able to significantly reduce the cost of building local stiffness matrices for the Laplace operator and for the trilinear form for Navier-Stokes. For the Laplace operator in two space dimensions, we have developed a heuristic graph algorithm that searches for such redundancies and generates code for computing the local stiffness matrices. Up to cubics, we are able to build the stiffness matrix on any triangle in less than one multiply-add pair per entry. Up to sixth degree, we can do it in less than about two. Preliminary low-degree results for Poisson and Navier-Stokes operators in three dimensions are also promising.

math.NA↗
arXiv2012

Topological Optimization of the Evaluation of Finite Element Matrices

We present a topological framework for finding low-flop algorithms for evaluating element stiffness matrices associated with multilinear forms for finite element methods posed over straight-sided affine domains. This framework relies on phrasing the computation on each element as the contraction of each collection of reference element tensors with an element-specific geometric tensor. We then present a new concept of complexity-reducing relations that serve as distance relations between these reference element tensors. This notion sets up a graph-theoretic context in which we may find an optimized algorithm by computing a minimum spanning tree. We present experimental results for some common multilinear forms showing significant reductions in operation count and also discuss some efficient algorithms for building the graph we use for the optimization.

math.NA↗
arXiv2011

Unstructured Geometric Multigrid in Two and Three Dimensions on Complex and Graded Meshes

The use of multigrid and related preconditioners with the finite element method is often limited by the difficulty of applying the algorithm effectively to a problem, especially when the domain has a complex shape or adaptive refinement. We introduce a simplification of a general topologically-motivated mesh coarsening algorithm for use in creating hierarchies of meshes for geometric unstructured multigrid methods. The connections between the guarantees of this technique and the quality criteria necessary for multigrid methods for non-quasi-uniform problems are noted. The implementation details, in particular those related to coarsening, remeshing, and interpolation, are discussed. Computational tests on pathological test cases from adaptive finite element methods show the performance of the technique.

math.NA↗