{"title":"The Akhiezer iteration","authors":"Cade Ballew, Thomas Trogdon","doi":"arxiv-2312.02384","DOIUrl":"https://doi.org/arxiv-2312.02384","url":null,"abstract":"We develop the Akhiezer iteration, a generalization of the classical\u0000Chebyshev iteration, for the inner product-free, iterative solution of\u0000indefinite linear systems using orthogonal polynomials for measures supported\u0000on multiple, disjoint intervals. The iteration applies to shifted linear solves\u0000and can then be used for efficient matrix function approximation. Using the\u0000asymptotics of orthogonal polynomials, error bounds are provided. A key\u0000component in the efficiency of the method is the ability to compute the first\u0000$k$ orthogonal polynomial recurrence coefficients and the first $k$ weighted\u0000Stieltjes transforms of these orthogonal polynomials in $mathrm{O}(k)$\u0000complexity using a numerical Riemann--Hilbert approach. For a special class of\u0000orthogonal polynomials, the Akhiezer polynomials, the method can be sped up\u0000significantly, with the greatest speedup occurring in the two interval case\u0000where important formulae of Akhiezer are employed and the Riemann--Hilbert\u0000approach is bypassed.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 8","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Mathematical Supplement for the $texttt{gsplat}$ Library","authors":"Vickie Ye, Angjoo Kanazawa","doi":"arxiv-2312.02121","DOIUrl":"https://doi.org/arxiv-2312.02121","url":null,"abstract":"This report provides the mathematical details of the gsplat library, a\u0000modular toolbox for efficient differentiable Gaussian splatting, as proposed by\u0000Kerbl et al. It provides a self-contained reference for the computations\u0000involved in the forward and backward passes of differentiable Gaussian\u0000splatting. To facilitate practical usage and development, we provide a user\u0000friendly Python API that exposes each component of the forward and backward\u0000passes in rasterization at github.com/nerfstudio-project/gsplat .","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 21","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494192","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Lecture Notes on Computerized Tomography","authors":"Matthias Beckmann","doi":"arxiv-2312.02393","DOIUrl":"https://doi.org/arxiv-2312.02393","url":null,"abstract":"These lecture notes give an introduction to the mathematics of computer(ized)\u0000tomography (CT). Treated are the imaging principle of X-ray tomography, the\u0000Radon transform as mathematical model for the measurement process and its\u0000properties, the ill-posedness of the underlying mathematical reconstruction\u0000problem and classical reconstruction techniques. The required background from\u0000Fourier analysis is also briefly summarized.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 7","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ľubomír Baňas, Giorgio Ferrari, Tsiry Avisoa Randrianasolo
{"title":"Numerical approximation of Dynkin games with asymmetric information","authors":"Ľubomír Baňas, Giorgio Ferrari, Tsiry Avisoa Randrianasolo","doi":"arxiv-2312.01847","DOIUrl":"https://doi.org/arxiv-2312.01847","url":null,"abstract":"We propose an implementable, feedforward neural network-based structure\u0000preserving probabilistic numerical approximation for a generalized obstacle\u0000problem describing the value of a zero-sum differential game of optimal\u0000stopping with asymmetric information. The target solution depends on three\u0000variables: the time, the spatial (or state) variable, and a variable from a\u0000standard $(I-1)$-simplex which represents the probabilities with which the $I$\u0000possible configurations of the game are played. The proposed numerical\u0000approximation preserves the convexity of the continuous solution as well as the\u0000lower and upper obstacle bounds. We show convergence of the fully-discrete\u0000scheme to the unique viscosity solution of the continuous problem and present a\u0000range of numerical studies to demonstrate its applicability.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 15","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Using Bayesian Optimization to Design Time Step Size Controllers with Application to Modified Patankar--Runge--Kutta Methods","authors":"Thomas Izgin, Hendrik Ranocha","doi":"arxiv-2312.01796","DOIUrl":"https://doi.org/arxiv-2312.01796","url":null,"abstract":"Modified Patankar--Runge--Kutta (MPRK) methods are linearly implicit time\u0000integration schemes developed to preserve positivity and a linear invariant\u0000such as the total mass in chemical reactions. MPRK methods are naturally\u0000equipped with embedded schemes yielding a local error estimate similar to\u0000Runge--Kutta pairs. To design good time step size controllers using these error\u0000estimates, we propose to use Bayesian optimization. In particular, we design a\u0000novel objective function that captures important properties such as tolerance\u0000convergence and computational stability. We apply our new approach to several\u0000MPRK schemes and controllers based on digital signal processing, extending\u0000classical PI and PID controllers. We demonstrate that the optimization process\u0000yields controllers that are at least as good as the best controllers chosen\u0000from a wide range of suggestions available for classical explicit and implicit\u0000time integration methods.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 14","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Fast Fourier Transform periodic interpolation method for superposition sums in a periodic unit cell","authors":"Fangzhou Ai, Vitaliy Lomakin","doi":"arxiv-2312.02376","DOIUrl":"https://doi.org/arxiv-2312.02376","url":null,"abstract":"We propose a Fast Fourier Transform based Periodic Interpolation Method\u0000(FFT-PIM), a flexible and computationally efficient approach for computing the\u0000scalar potential given by a superposition sum in a unit cell of an infinitely\u0000periodic array. Under the same umbrella, FFT-PIM allows computing the potential\u0000for 1D, 2D, and 3D periodicities for dynamic and static problems, including\u0000problems with and without a periodic phase shift. The computational complexity\u0000of the FFT-PIM is of $O(N log N)$ for $N$ spatially coinciding sources and\u0000observer points. The FFT-PIM uses rapidly converging series representations of\u0000the Green's function serving as a kernel in the superposition sum. Based on\u0000these representations, the FFT-PIM splits the potential into its near-zone\u0000component, which includes a small number of images surrounding the unit cell of\u0000interest, and far-zone component, which includes the rest of an infinite number\u0000of images. The far-zone component is evaluated by projecting the non-uniform\u0000sources onto a sparse uniform grid, performing superposition sums on this\u0000sparse grid, and interpolating the potential from the uniform grid to the\u0000non-uniform observation points. The near-zone component is evaluated using an\u0000FFT-based method, which is adapted to efficiently handle non-uniform\u0000source-observer distributions within the periodic unit cell. The FFT-PIM can be\u0000used for a broad range of applications, such as periodic problems involving\u0000integral equations in computational electromagnetic and acoustic, micromagnetic\u0000solvers, and density functional theory solvers.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 9","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A $W$-weighted generalization of ${1,2,3,1^{k}}$-inverse for rectangular matrices","authors":"Geeta Chowdhry, Falguni Roy","doi":"arxiv-2312.01370","DOIUrl":"https://doi.org/arxiv-2312.01370","url":null,"abstract":"This paper presents a novel extension of the ${1,2,3,1^{k}}$-inverse\u0000concept to complex rectangular matrices, denoted as a $W$-weighted\u0000${1,2,3,1^{k}}$-inverse (or ${1',2',3',{1^{k}}'}$-inverse), where the\u0000weight $W in mathbb{C}^{n times m}$. The study begins by introducing a\u0000weighted ${1,2,3}$-inverse (or ${1',2',3'}$-inverse) along with its\u0000representations and characterizations. The paper establishes criteria for the\u0000existence of ${1',2',3'}$-inverses and extends the criteria to\u0000${1'}$-inverses. It is further demonstrated that $Ain mathbb{C}^{m times\u0000n}$ admits a ${1',2',3',{1^{k}}'}$-inverse if and only if $r(WAW)=r(A)$,\u0000where $r(cdot)$ is the rank of a matrix. The work additionally establishes\u0000various representations for the set $A{ 1',2',3',{1^{k}}'}$, including\u0000canonical representations derived through singular value and core-nilpotent\u0000decompositions. This, in turn, yields distinctive canonical representations for\u0000the set $A{ 1,2,3,{1^{k}}}$. ${ 1',2',3',{1^{k}}'}$-inverse is shown to be\u0000unique if and only if it has index $0$ or $1$, reducing it to the weighted core\u0000inverse. Moreover, the paper investigates properties and characterizations of\u0000${1',2',3',{1^{k}}'}$-inverses, which then results in new insights into the\u0000characterizations of the set $A{ 1,2,3,{1^{k}}}$.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494402","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
J. R. Yusupov, M. Ehrhardt, Kh. Sh. Matyokubov, D. U. Matrasulov
{"title":"Driven transparent quantum graphs","authors":"J. R. Yusupov, M. Ehrhardt, Kh. Sh. Matyokubov, D. U. Matrasulov","doi":"arxiv-2312.01448","DOIUrl":"https://doi.org/arxiv-2312.01448","url":null,"abstract":"In this paper, we discuss the concept of quantum graphs with transparent\u0000vertices by considering the case where the graph interacts with an external\u0000time-independent field. In particular, we address the problem of transparent\u0000boundary conditions for quantum graphs, building on previous work on\u0000transparent boundary conditions for the stationary Schrodinger equation on a\u0000line. Physically relevant constraints making the vertex transparent under\u0000boundary conditions in the form of (weight) continuity and Kirchhoff rules are\u0000derived using two methods, the scattering approach and transparent boundary\u0000conditions for the time-independent Schrodinger equation. The latter is derived\u0000by extending the transparent boundary condition concept to the time-independent\u0000Schrodinger equation on driven quantum graphs. We also discuss how the\u0000eigenvalues and eigenfunctions of a quantum graph are influenced not only by\u0000its topology, but also by the shape(type) of a potential when an external field\u0000is involved.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 23","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494190","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A New Upper Bound For the Growth Factor in Gaussian Elimination with Complete Pivoting","authors":"Ankit Bisain, Alan Edelman, John Urschel","doi":"arxiv-2312.00994","DOIUrl":"https://doi.org/arxiv-2312.00994","url":null,"abstract":"The growth factor in Gaussian elimination measures how large the entries of\u0000an LU factorization can be relative to the entries of the original matrix. It\u0000is a key parameter in error estimates, and one of the most fundamental topics\u0000in numerical analysis. We produce an upper bound of $n^{0.2079 ln n +0.91}$\u0000for the growth factor in Gaussian elimination with complete pivoting -- the\u0000first improvement upon Wilkinson's original 1961 bound of $2 , n ^{0.25ln n\u0000+0.5}$.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 20","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494193","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"The time dimensional reduction method to determine the initial conditions for nonlinear and nonlocal hyperbolic equations","authors":"Trong D. Dang, Loc H. Nguyen, Huong T. T. Vu","doi":"arxiv-2312.01179","DOIUrl":"https://doi.org/arxiv-2312.01179","url":null,"abstract":"The objective of this paper is to compute initial conditions for quasi-linear\u0000hyperbolic equations. Our proposed approach involves approximating the solution\u0000of the hyperbolic equation by truncating its Fourier expansion in the time\u0000domain using the polynomial-exponential basis. This truncation enables the\u0000elimination of the time variable, resulting in a system of quasi-linear\u0000elliptic equations. Thus, we refer to our approach as the \"time dimensional\u0000reduction method.\" To solve this system globally without requesting a good\u0000initial guess, we employ the Carleman contraction principle. To demonstrate the\u0000effectiveness of our method, we provide several numerical examples. The time\u0000dimensional reduction method not only provides accurate solutions but also\u0000exhibits exceptional computational speed.","PeriodicalId":501061,"journal":{"name":"arXiv - CS - Numerical Analysis","volume":" 18","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138494233","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}