Fast Algorithms for the Computation of Oscillatory Integrals
|
|
- Gilbert Wheeler
- 5 years ago
- Views:
Transcription
1 Fast Algorithms for the Computation of Oscillatory Integrals Emmanuel Candès California Institute of Technology EPSRC Symposium Capstone Conference Warwick Mathematics Institute, July 2009
2 Collaborators Lexing Ying Laurent Demanet
3 Our problem Evaluate numerically a Fourier integral operator (FIO) (T f)(x) = a(x, k)e 2πiΦ(x,k) ˆf(k) dk R d at points x given on a Cartesian grid k R d : frequency variable ( ˆf(k) = e 2πix k f(x) dx) R d a(x, k): (smooth) amplitude Φ(x, λk): homogeneous (smooth) phase function as large as k Φ(x, λk) = λ Φ(x, k), λ > 0 e.g. Φ(x, k) = g(x) k
4 A motivating example: wave propagation 2 u t 2 (x, t) = c2 u(x, t), u(x, 0) = u 0 (x) u/ t(x, 0) = 0
5 A motivating example: wave propagation 2 u t 2 (x, t) = c2 u(x, t), u(x, 0) = u 0 (x) u/ t(x, 0) = 0 Solution operator is u(x, t) = 1 ( ) e 2πi(x k+c k t) û 0 (k) dk + e 2πi(x k c k t) û 0 (k) dk 2 R 2 R 2 Two FIOs with phase functions Φ ± (x, k) = x k ± c k t Inhomogeneous medium c(x) solution operator = sum of two FIOs (small times)
6 Importance of FIOs Arise in many (inverse) problems Applying FIOs is often the computational bottleneck
7 Example in seismics Marine survey
8 Kirchhoff migration Wave measurements f s (t, x r ) parametrized by time t receiver location x r source coordinate x s Forward map: F δc = δp Imaging operator is F : FIO under general assumptions Approximations by generalized Radon transform (GRT): integration of f s over fixed set of curves parametrized by travel times g s (x) = δ(t τ(x, x r ) τ(x, x s )) f s (t, x r ) dt dx r Followed by stack operation over the s index
9 Other examples Transmission electron microscopy Radar imaging Ultrasound imaging
10 Discrete computational problem Discrete grids X = {(i 1 /N, i 2 /N) : 0 i 1, i 2 < N} [0, 1] 2 Ω = {(k 1 /N, k 2 /N) : N/2 k 1, k 2 < N} [ 1/2, 1/2] 2 Given input {f(k)} k Ω, evaluate (T f)(x) := 1 a(x, k) e 2πiNΦ(x,k) f(k), N at all x X k Ω with Φ smooth and homogeneous in k
11 Peek at the results (T f)(x) := 1 a(x, k) e 2πiNΦ(x,k) f(k), N k Ω Kernel is not analytic and is highly oscillatory x X Naive evaluation O(N 4 ) (O(N 2 ) inputs/outputs) Algorithm for fast summation O(N 2.5 log N) (C., Demanet and Ying, 06)
12 Peek at the results (T f)(x) := 1 a(x, k) e 2πiNΦ(x,k) f(k), N k Ω Kernel is not analytic and is highly oscillatory x X Naive evaluation O(N 4 ) (O(N 2 ) inputs/outputs) Algorithm for fast summation O(N 2.5 log N) (C., Demanet and Ying, 06) Today: Novel algorithm with optimal complexity for accurate summation O(N 2 log N) flops O(N 2 ) storage
13 Agenda The butterfly structure Fast butterfly algorithm for the evaluation of FIOs
14 The Butterfly Structure
15 Butterfly algorithm General algorithmic structure for evaluating certain types of integrals u i = j K(x i, p j )f j Introduced by Michielssen and Boag ( 96) Generalized by O Neil and Rokhlin ( 07)
16 Butterfly algorithm General algorithmic structure for evaluating certain types of integrals u i = j K(x i, p j )f j Introduced by Michielssen and Boag ( 96) Generalized by O Neil and Rokhlin ( 07) Example: K(x, p) = e 2πiNxp {x i }: N points in [0, 1] {p j }: N points in [0, 1] {f j }: sources at {p j } Applications FFT nonuniform FFTs many others Kernel is dense and oscillatory
17 Low-rank approximation (K(x, p) = exp(2πin xp)) A interval in x B interval in p obeying length(a) length(b) 1/N The submatrix {K(x i, p j ) : x i A, p j B} has approximately low rank: K(x, p) with r = O(log(1/ɛ)) r a t (x)b t (p) ɛ t=1
18 Example DFT Top left block 10 0 Top singular values
19 Interpolative decompositions O Neil and Rokhlin suggest using interpolative decompositions a t (x) := K(x, p t ), p t B Rank-revealing QR decomposition: Gu and Eisenstat ( 96) Interpolative decomp.: Cheng, Gimbutas, Martinsson and Rokhlin ( 05) Interpolative representation r K(x, p) K(x, p t )b t (x) Cost for an m n matrix is O(mn 2 ) t=1
20 Multiscale decompositions Compute low-rank approximations of all submatrices obeying Use two-scale relations for efficiency length(a) length(b) = 1/N
21 Definition of partial sums and equivalent sources Partial sums u B (x) = K(x, p j )f j p j B Approximation for x A and length(a) length(b) 1/N r u B (x) K(x, p AB (p j )f j, x A t=1 t ) p j B b AB t
22 Definition of partial sums and equivalent sources Partial sums u B (x) = K(x, p j )f j p j B Approximation for x A and length(a) length(b) 1/N r u B (x) K(x, p AB (p j )f j, x A t=1 t ) p j B Equivalent sources for (A, B): {ft AB } 1 t r = f AB t p j B b AB t b AB t (p j )f j
23 Definition of partial sums and equivalent sources Partial sums u B (x) = K(x, p j )f j p j B Approximation for x A and length(a) length(b) 1/N r u B (x) K(x, p AB (p j )f j, x A t=1 t ) p j B Equivalent sources for (A, B): {ft AB } 1 t r = f AB t p j B b AB t b AB t (p j )f j Compact representation of K AB : B A r u B (x) = K(x, p AB t )ft AB t=1
24 Definition of partial sums and equivalent sources Partial sums u B (x) = K(x, p j )f j p j B Approximation for x A and length(a) length(b) 1/N r u B (x) K(x, p AB (p j )f j, x A t=1 t ) p j B Equivalent sources for (A, B): {ft AB } 1 t r = f AB t p j B b AB t b AB t (p j )f j Compact representation of K AB : B A r u B (x) = K(x, p AB t )ft AB Butterfly structure Recursive computation of {p AB t }, {ft AB } for length(a) length(b) = 1/N t=1
25 Initialization For all (A, B), l(a) = 1 & l(b) = 1/N, construct {p AB t = f AB t p j B b AB t (p j )f j } 1 t r and {ft AB } 1 t r cost of constructing {p AB t } is O(r 2 N)/pair cost of constructing {ft AB } is O(r 2 )/pair tot. cost is O(r 2 N 2 ) Interpretation: {p AB t } selected columns, {ft AB } column weights
26 Recursion: merge, split, compress For all (A, B), l(a) = 1/2 & l(b) = 2/N, get {p AB t } 1 t r and {ft AB } 1 t r u B (x) 2 r i=1 t=1 K(x, p ApBi t )f ApBi t, x A A p
27 Recursion: merge, split, compress For all (A, B), l(a) = 1/2 & l(b) = 2/N, get {p AB t } 1 t r and {ft AB } 1 t r u B (x) 2 r i=1 t=1 Can reduce the rank of K(x, p ApBi t Treat {f ApBi t K(x, p ApBi t )f ApBi t, x A A p } t,i as sources at {p ApBi t } t,i B f AB t = 2 i=1 s=1 ) : x A, {p ApBi t } t,i K(x, p AB t ) r b AB t (p ApBi s )fs ApBi cost of constructing {p AB t } is O(r 2 N)/pair cost of constructing {ft AB } is O(r 2 )/pair tot. cost is O(r 2 N 2 )
28 Schematic representation
29 Schematic representation u Bi (x) = r t=1 K(x, p ApBi t )f ApBi t u B (x) = r t=1 K(x, p AB t )ft AB
30 Next step... until l(b) = 1 and l(a) = 1/N
31 Termination In the end, l(a) = 1/N and l(b) = 1, and u B (x) r t=1 K(x, p AB t )ft AB cost of evaluating u B (x) is O(r)/pair tot. cost is O(rN)
32 Multiscale recursion
33 Summary: complexity analysis If low-rank expansions available If not O(r 2 N log N) evaluation time O(r 2 N 2 ) evaluation time O(r 2 N log N) storage O(r 2 N log N) storage
34 Summary: complexity analysis If low-rank expansions available If not O(r 2 N log N) evaluation time O(r 2 N 2 ) evaluation time O(r 2 N log N) storage O(r 2 N log N) storage Powerful architecture but Precomputation time may be prohibitive Storage may be prohibitive
35 Summary: complexity analysis If low-rank expansions available If not O(r 2 N log N) evaluation time O(r 2 N 2 ) evaluation time O(r 2 N log N) storage O(r 2 N log N) storage Powerful architecture but Precomputation time may be prohibitive Storage may be prohibitive Our contribution O(r 2 N log N) evaluation time and storage complexity is O(r 2 N)
36 Fast Evaluation of FIOs
37 Recall oscillatory integral u(x) = k Ω e 2πiNΦ(x,k) f k x X k Ω
38 Polar coordinates for frequency variable k Phase may be singular at k = 0 because of homogeneity (e.g. Φ(x, k) = k )
39 Polar coordinates for frequency variable k Phase may be singular at k = 0 because of homogeneity (e.g. Φ(x, k) = k ) Polar coordinates p = (p 1, p 2 ) k 1 = p 1 cos(2πp 2 ) k 2 = p 2 sin(2πp 2 ) Cartesian Polar Partitioning in k Slight abuse of notations u(x) = p Ω e 2πiNΨ(x,p) f p, smooth phase Ψ
40 Hierarchical structure l(a) l(b) = 1/N
41 Low-rank interactions If l(a)l(b) 1/N, kernel e 2πiNΨ(x,p) : x A, p B has approx. low rank Assume wlog A and B centered at 0
42 Low-rank interactions If l(a)l(b) 1/N, kernel e 2πiNΨ(x,p) : x A, p B has approx. low rank Assume wlog A and B centered at 0 Residual phase in 1D (for simplicity) R AB (x, p) = Ψ(x, p) Ψ(0, p) Ψ(x, 0) + Ψ(0, 0) = x p Ψ(x, p ) xp = O(1/N) Same calculation in higher dimensions
43 Low-rank interactions If l(a)l(b) 1/N, kernel e 2πiNΨ(x,p) : x A, p B has approx. low rank Assume wlog A and B centered at 0 Residual phase in 1D (for simplicity) R AB (x, p) = Ψ(x, p) Ψ(0, p) Ψ(x, 0) + Ψ(0, 0) = x p Ψ(x, p ) xp = O(1/N) Same calculation in higher dimensions Shows that e 2πiNRAB (x,p) approx. low rank
44 Low-rank interactions If l(a)l(b) 1/N, kernel e 2πiNΨ(x,p) : x A, p B has approx. low rank Assume wlog A and B centered at 0 Residual phase in 1D (for simplicity) R AB (x, p) = Ψ(x, p) Ψ(0, p) Ψ(x, 0) + Ψ(0, 0) = x p Ψ(x, p ) xp = O(1/N) Same calculation in higher dimensions Shows that e 2πiNRAB (x,p) approx. low rank Factorization [ e 2πiNΨ(x,p) = e 2πiNΨ(0,0) e 2πiNΨ(0,x) e 2πiNRAB (x,p) e 2πiNΨ(0,p)]
45 Oscillatory Chebyshev interpolation Chebyshev interpolation of e 2πiNRAB (x,p) in x when l(a) 1/ N p when l(b) 1/ N E.g. interpolation in p {p t } B : tensor-chebyshev grid in B L B t (p) : Lagrange interpolant with inputs at {p B t }, L B t (p B t ) = δ(t = t )
46 Oscillatory Chebyshev interpolation Chebyshev interpolation of e 2πiNRAB (x,p) in x when l(a) 1/ N p when l(b) 1/ N E.g. interpolation in p {p t } B : tensor-chebyshev grid in B L B t (p) : Lagrange interpolant with inputs at {p B t }, L B t (p B t ) = δ(t = t ) With grid of logarithmic size in 1/ɛ (x,p) e e2πinrab 2πiNRAB (x,p B t ) L B t (p) ɛ t on A B When interpolation in x, low-rank approximation is P t e2πinrab (x A t,p) L A t (x)
47 Demodulation/Interpolation/Remodulation From R AB (x, p) = Ψ(x, p) Ψ(0, p) Ψ(x, 0) + Ψ(0, 0) e 2πiNΨ(x,p) e 2πiNΨ(x,pB t }{{} ) e i2πnψ(0,pb t ) L B t (p) e i2πnψ(0,p) ɛ }{{} t a t(x) b t(x) Similar structure (with different interpretation) when interpolating kernel in x
48 Overall structure and two-scale relation 1 Initialize equivalent sources 2 Propagate equivalent sources (interpolation in p) until mid-level 3 Switch representation at mid-level 4 Propagate equivalent sources (interpolation in x) 5 Terminate by evaluating output Step 2: propagation of equivalent sources f AB t = c,t b B t (p Bc t ApBc )ft = e 2πiNΨ(x0(A),pB t ) c t L B t (p Bc t ) e2πinψ(x0(a),pbc t ) f ApBc t Step 4: similar
49 Finer points and summary {p B t } and polynomials L t (p) are computed all at once Only need to store equivalent sources at pairs of consecutive scales Separation rank higher than that of the interpolative decomposition
50 Finer points and summary {p B t } and polynomials L t (p) are computed all at once Only need to store equivalent sources at pairs of consecutive scales Separation rank higher than that of the interpolative decomposition Complexity is O(polylog(1/ɛ) N 2 log N) Storage is O(polylog(1/ɛ) N 2 for ɛ-accurate computation Easy extensions to varying amplitudes a(x, k)e i2πnφ(x,k)
51 Numerical results Generalized Radon transform integrating f along ellipses centered at x and with axes of length c 1 (x) and c 2 (x) is the sum of two FIOs with phases Φ ± (x, k) = x k ± c 2 1 (x)k2 1 + c2 2 (x)k2 2
52 Numerical results Generalized Radon transform integrating f along ellipses centered at x and with axes of length c 1 (x) and c 2 (x) is the sum of two FIOs with phases Φ ± (x, k) = x k ± c 2 1 (x)k2 1 + c2 2 (x)k2 2 First example (constant amplitude) u(x) = k Ω e 2πiNΦ+(x,k) ˆf(k) with c 1 (x) = 1 3 (2 + sin(2πx 1) sin(2πx 2 )), c 2 (x) = 1 3 (2 + cos(2πx 1) cos(2πx 2 ))
53 (N, Chby. grid) Alg. Time Dir. Time Speedup Error (1024, 5) 1.48e e e e-2 (2048, 5) 6.57e e e e-2 (4096, 5) 3.13e e e e-2 (1024, 7) 2.76e e e e-4 (2048, 7) 1.23e e e e-4 (4096, 7) 5.80e e e e-4 (1024, 9) 4.95e e e e-5 (2048, 9) 2.21e e e e-5 (4096, 9) 1.02e e e e-5 (1024, 11) 8.33e e e e-7 (2048, 11) 3.48e e e e-7 Conclusion: scales like O(log(1/ɛ)) O(N 2 log N)
54 Numerical results II Second example (variable amplitude): exact integration u(x) = a + (x, k)e 2πiNΦ+(x,k) ˆf(k) + a (x, k)e 2πiNΦ (x,k) ˆf(k) k Ω k Ω a ± (x, k) = (J 0 (2πc(x) k ) ± iy 0 (2πc(x) k ) e 2πic(x) k Φ ± (x, k) = x k + c(x) k J 0 and Y 0 are Bessel functions and spheres radii c(x) = 1 4 (3 + sin(2πx 1) sin(2πx 2 ))
55 (N, Chby. grid) Alg. Time Dir. Time Speedup Error (256, 5) 1.39e e e e-2 (512, 5) 7.25e e e e-2 (1024, 5) 3.45e e e e-2 (256, 7) 2.69e e e e-4 (512, 7) 1.38e e e e-4 (1024, 7) 6.43e e e e-4 (256, 9) 5.23e e e e-5 (512, 9) 2.49e e e e-5 (1024, 9) 1.15e e e e-5 (256, 11) 1.04e e e e-7 (512, 11) 4.10e e e e-7 (1024, 11) 1.84e e e e-7 Conclusion: scales like O(log(1/ɛ)) O(N 2 log N)
56 Summary Accurate and near-optimal numerical evaluation of FIOs Operating characteristics Butterfly structure Residue phase Oscillatory Chebyshev interpolation Many applications/extensions Other types of oscillatory kernels K(x, p) Other types of problems: e.g. sparse Fourier transform (Ying, 08) Reference: E. J. Candès, L. Demanet and L. Ying (2008). A Fast Butterfly Algorithm for the Computation of Fourier Integral Operators, to appear in Mult. Model. Sim
Sparse Fourier Transform via Butterfly Algorithm
Sparse Fourier Transform via Butterfly Algorithm Lexing Ying Department of Mathematics, University of Texas, Austin, TX 78712 December 2007, revised June 2008 Abstract This paper introduces a fast algorithm
More informationA Unified Framework for Oscillatory Integral Transform: When to use NUFFT or Butterfly Factorization?
A Unified Framework for Oscillatory Integral Transform: When to use NUFFT or Butterfly Factorization? Haizhao Yang Department of Mathematics, National University of Singapore March 12, 2018 Abstract This
More informationA butterfly algorithm. for the geometric nonuniform FFT. John Strain Mathematics Department UC Berkeley February 2013
A butterfly algorithm for the geometric nonuniform FFT John Strain Mathematics Department UC Berkeley February 2013 1 FAST FOURIER TRANSFORMS Standard pointwise FFT evaluates f(k) = n j=1 e 2πikj/n f j
More informationResearch Statement. James Bremer Department of Mathematics, University of California, Davis
Research Statement James Bremer Department of Mathematics, University of California, Davis Email: bremer@math.ucdavis.edu Webpage: https.math.ucdavis.edu/ bremer I work in the field of numerical analysis,
More informationThe spectral transform method
The spectral transform method by Nils Wedi European Centre for Medium-Range Weather Forecasts wedi@ecmwf.int Advanced Numerical Methods for Earth-System Modelling Slide 1 Advanced Numerical Methods for
More informationWavelets in Scattering Calculations
Wavelets in Scattering Calculations W. P., Brian M. Kessler, Gerald L. Payne polyzou@uiowa.edu The University of Iowa Wavelets in Scattering Calculations p.1/43 What are Wavelets? Orthonormal basis functions.
More informationSparse and Fast Fourier Transforms in Biomedical Imaging
Sparse and Fast Fourier Transforms in Biomedical Imaging Stefan Kunis 1 1 joint work with Ines Melzer, TU Chemnitz and Helmholtz Center Munich, supported by DFG and Helmholtz association Outline 1 Trigonometric
More informationarxiv: v1 [cs.lg] 26 Jul 2017
Updating Singular Value Decomposition for Rank One Matrix Perturbation Ratnik Gandhi, Amoli Rajgor School of Engineering & Applied Science, Ahmedabad University, Ahmedabad-380009, India arxiv:70708369v
More informationHierarchical Matrices. Jon Cockayne April 18, 2017
Hierarchical Matrices Jon Cockayne April 18, 2017 1 Sources Introduction to Hierarchical Matrices with Applications [Börm et al., 2003] 2 Sources Introduction to Hierarchical Matrices with Applications
More informationON INTERPOLATION AND INTEGRATION IN FINITE-DIMENSIONAL SPACES OF BOUNDED FUNCTIONS
COMM. APP. MATH. AND COMP. SCI. Vol., No., ON INTERPOLATION AND INTEGRATION IN FINITE-DIMENSIONAL SPACES OF BOUNDED FUNCTIONS PER-GUNNAR MARTINSSON, VLADIMIR ROKHLIN AND MARK TYGERT We observe that, under
More informationA proof for the full Fourier series on [ π, π] is given here.
niform convergence of Fourier series A smooth function on an interval [a, b] may be represented by a full, sine, or cosine Fourier series, and pointwise convergence can be achieved, except possibly at
More informationNonlinear seismic imaging via reduced order model backprojection
Nonlinear seismic imaging via reduced order model backprojection Alexander V. Mamonov, Vladimir Druskin 2 and Mikhail Zaslavsky 2 University of Houston, 2 Schlumberger-Doll Research Center Mamonov, Druskin,
More informationPart 2 Introduction to Microlocal Analysis
Part 2 Introduction to Microlocal Analysis Birsen Yazıcı & Venky Krishnan Rensselaer Polytechnic Institute Electrical, Computer and Systems Engineering August 2 nd, 2010 Outline PART II Pseudodifferential
More informationMcGill University Department of Mathematics and Statistics. Ph.D. preliminary examination, PART A. PURE AND APPLIED MATHEMATICS Paper BETA
McGill University Department of Mathematics and Statistics Ph.D. preliminary examination, PART A PURE AND APPLIED MATHEMATICS Paper BETA 17 August, 2018 1:00 p.m. - 5:00 p.m. INSTRUCTIONS: (i) This paper
More informationLecture 1: Introduction
CBMS Conference on Fast Direct Solvers Dartmouth College June June 7, 4 Lecture : Introduction Gunnar Martinsson The University of Colorado at Boulder Research support by: Many thanks to everyone who made
More informationInexact Newton Methods and Nonlinear Constrained Optimization
Inexact Newton Methods and Nonlinear Constrained Optimization Frank E. Curtis EPSRC Symposium Capstone Conference Warwick Mathematics Institute July 2, 2009 Outline PDE-Constrained Optimization Newton
More informationMath 671: Tensor Train decomposition methods
Math 671: Eduardo Corona 1 1 University of Michigan at Ann Arbor December 8, 2016 Table of Contents 1 Preliminaries and goal 2 Unfolding matrices for tensorized arrays The Tensor Train decomposition 3
More informationICES REPORT A Multilevel-WENO Technique for Solving Nonlinear Conservation Laws
ICES REPORT 7- August 7 A Multilevel-WENO Technique for Solving Nonlinear Conservation Laws by Todd Arbogast, Chieh-Sen Huang, and Xikai Zhao The Institute for Computational Engineering and Sciences The
More informationFast Structured Spectral Methods
Spectral methods HSS structures Fast algorithms Conclusion Fast Structured Spectral Methods Yingwei Wang Department of Mathematics, Purdue University Joint work with Prof Jie Shen and Prof Jianlin Xia
More informationFast numerical methods for solving linear PDEs
Fast numerical methods for solving linear PDEs P.G. Martinsson, The University of Colorado at Boulder Acknowledgements: Some of the work presented is joint work with Vladimir Rokhlin and Mark Tygert at
More informationPart 2 Introduction to Microlocal Analysis
Part 2 Introduction to Microlocal Analysis Birsen Yazıcı& Venky Krishnan Rensselaer Polytechnic Institute Electrical, Computer and Systems Engineering March 15 th, 2010 Outline PART II Pseudodifferential(ψDOs)
More information8 Singular Integral Operators and L p -Regularity Theory
8 Singular Integral Operators and L p -Regularity Theory 8. Motivation See hand-written notes! 8.2 Mikhlin Multiplier Theorem Recall that the Fourier transformation F and the inverse Fourier transformation
More informationc 2009 International Press
COMMUN. MATH. SCI. Vol. 7, No. 2, pp. 327 345 c 2009 International Press A FAST DIRECTIONAL ALGORITHM FOR HIGH FREQUENCY ACOUSTIC SCATTERING IN TWO DIMENSIONS BJÖRN ENGQUIST AND LEXING YING Abstract. This
More information1 Distributions (due January 22, 2009)
Distributions (due January 22, 29). The distribution derivative of the locally integrable function ln( x ) is the principal value distribution /x. We know that, φ = lim φ(x) dx. x ɛ x Show that x, φ =
More informationPART IV Spectral Methods
PART IV Spectral Methods Additional References: R. Peyret, Spectral methods for incompressible viscous flow, Springer (2002), B. Mercier, An introduction to the numerical analysis of spectral methods,
More informationLucio Demeio Dipartimento di Ingegneria Industriale e delle Scienze Matematiche
Scuola di Dottorato THE WAVE EQUATION Lucio Demeio Dipartimento di Ingegneria Industriale e delle Scienze Matematiche Lucio Demeio - DIISM wave equation 1 / 44 1 The Vibrating String Equation 2 Second
More informationA direct solver for elliptic PDEs in three dimensions based on hierarchical merging of Poincaré-Steklov operators
(1) A direct solver for elliptic PDEs in three dimensions based on hierarchical merging of Poincaré-Steklov operators S. Hao 1, P.G. Martinsson 2 Abstract: A numerical method for variable coefficient elliptic
More informationExploiting off-diagonal rank structures in the solution of linear matrix equations
Stefano Massei Exploiting off-diagonal rank structures in the solution of linear matrix equations Based on joint works with D. Kressner (EPFL), M. Mazza (IPP of Munich), D. Palitta (IDCTS of Magdeburg)
More informationChapter 2 Algorithms for Periodic Functions
Chapter 2 Algorithms for Periodic Functions In this chapter we show how to compute the Discrete Fourier Transform using a Fast Fourier Transform (FFT) algorithm, including not-so special case situations
More informationConstructing Approximation Kernels for Non-Harmonic Fourier Data
Constructing Approximation Kernels for Non-Harmonic Fourier Data Aditya Viswanathan aditya.v@caltech.edu California Institute of Technology SIAM Annual Meeting 2013 July 10 2013 0 / 19 Joint work with
More informationRapid evaluation of electrostatic interactions in multi-phase media
Rapid evaluation of electrostatic interactions in multi-phase media P.G. Martinsson, The University of Colorado at Boulder Acknowledgements: Some of the work presented is joint work with Mark Tygert and
More informationA fast randomized algorithm for computing a Hierarchically Semi-Separable representation of a matrix
A fast randomized algorithm for computing a Hierarchically Semi-Separable representation of a matrix P.G. Martinsson, Department of Applied Mathematics, University of Colorado at Boulder Abstract: Randomized
More informationPositive Definite Kernels: Opportunities and Challenges
Positive Definite Kernels: Opportunities and Challenges Michael McCourt Department of Mathematical and Statistical Sciences University of Colorado, Denver CUNY Mathematics Seminar CUNY Graduate College
More informationProblem descriptions: Non-uniform fast Fourier transform
Problem descriptions: Non-uniform fast Fourier transform Lukas Exl 1 1 University of Vienna, Institut für Mathematik, Oskar-Morgenstern-Platz 1, A-1090 Wien, Austria December 11, 2015 Contents 0.1 Software
More informationFast algorithms based on asymptotic expansions of special functions
Fast algorithms based on asymptotic expansions of special functions Alex Townsend MIT Cornell University, 23rd January 2015, CAM Colloquium Introduction Two trends in the computing era Tabulations Software
More informationMathematics of Physics and Engineering II: Homework problems
Mathematics of Physics and Engineering II: Homework problems Homework. Problem. Consider four points in R 3 : P (,, ), Q(,, 2), R(,, ), S( + a,, 2a), where a is a real number. () Compute the coordinates
More informationA direct solver for variable coefficient elliptic PDEs discretized via a composite spectral collocation method Abstract: Problem formulation.
A direct solver for variable coefficient elliptic PDEs discretized via a composite spectral collocation method P.G. Martinsson, Department of Applied Mathematics, University of Colorado at Boulder Abstract:
More informationA fast method for solving the Heat equation by Layer Potentials
A fast method for solving the Heat equation by Layer Potentials Johannes Tausch Abstract Boundary integral formulations of the heat equation involve time convolutions in addition to surface potentials.
More informationAn Inexact Newton Method for Optimization
New York University Brown Applied Mathematics Seminar, February 10, 2009 Brief biography New York State College of William and Mary (B.S.) Northwestern University (M.S. & Ph.D.) Courant Institute (Postdoc)
More informationMultiscale Geometric Analysis: Theory, Applications, and Opportunities
1 Multiscale Geometric Analysis: Theory, Applications, and Opportunities Emmanuel Candès, California Institute of Technology Wavelet And Multifractal Analysis 2004, Cargese, July 2004 Report Documentation
More informationNonlinear acoustic imaging via reduced order model backprojection
Nonlinear acoustic imaging via reduced order model backprojection Alexander V. Mamonov 1, Vladimir Druskin 2, Andrew Thaler 3 and Mikhail Zaslavsky 2 1 University of Houston, 2 Schlumberger-Doll Research
More informationFirst-order overdetermined systems. for elliptic problems. John Strain Mathematics Department UC Berkeley July 2012
First-order overdetermined systems for elliptic problems John Strain Mathematics Department UC Berkeley July 2012 1 OVERVIEW Convert elliptic problems to first-order overdetermined form Control error via
More informationPH.D. PRELIMINARY EXAMINATION MATHEMATICS
UNIVERSITY OF CALIFORNIA, BERKELEY SPRING SEMESTER 207 Dept. of Civil and Environmental Engineering Structural Engineering, Mechanics and Materials NAME PH.D. PRELIMINARY EXAMINATION MATHEMATICS Problem
More informationIntroduction to the FFT
Introduction to the FFT 1 Introduction Assume that we have a signal fx where x denotes time. We would like to represent fx as a linear combination of functions e πiax or, equivalently, sinπax and cosπax
More informationn 1 f n 1 c 1 n+1 = c 1 n $ c 1 n 1. After taking logs, this becomes
Root finding: 1 a The points {x n+1, }, {x n, f n }, {x n 1, f n 1 } should be co-linear Say they lie on the line x + y = This gives the relations x n+1 + = x n +f n = x n 1 +f n 1 = Eliminating α and
More informationAn algorithm for the rapid evalution of special function transforms
. An algorithm for the rapid evalution of special function transforms Michael O Neil, Franco Woolfe, and Vladimir Rokhlin Technical Report YALEU/DCS/TR-1399 April 15, 2008 Report Documentation Page Form
More informationAn Inexact Newton Method for Nonlinear Constrained Optimization
An Inexact Newton Method for Nonlinear Constrained Optimization Frank E. Curtis Numerical Analysis Seminar, January 23, 2009 Outline Motivation and background Algorithm development and theoretical results
More informationAdaptively Compressed Polarizability Operator
Adaptively Compressed Polarizability Operator For Accelerating Large Scale ab initio Phonon Calculations Ze Xu 1 Lin Lin 2 Lexing Ying 3 1,2 UC Berkeley 3 ICME, Stanford University BASCD, December 3rd
More informationImage Analysis and Approximation via Generalized Polyharmonic Local Trigonometric Transforms
Image Analysis and Approximation via Generalized Polyharmonic Local Trigonometric Transforms Naoki Saito with Jean-François Remy, Noel Smith, Ernest Woei, Katsu Yamatani, & Jim Zhao saito@math.ucdavis.edu
More informationIntroduction to Seismic Imaging
Introduction to Seismic Imaging Alison Malcolm Department of Earth, Atmospheric and Planetary Sciences MIT August 20, 2010 Outline Introduction Why we image the Earth How data are collected Imaging vs
More informationSparse Tensor Galerkin Discretizations for First Order Transport Problems
Sparse Tensor Galerkin Discretizations for First Order Transport Problems Ch. Schwab R. Hiptmair, E. Fonn, K. Grella, G. Widmer ETH Zürich, Seminar for Applied Mathematics IMA WS Novel Discretization Methods
More informationPDE-Constrained and Nonsmooth Optimization
Frank E. Curtis October 1, 2009 Outline PDE-Constrained Optimization Introduction Newton s method Inexactness Results Summary and future work Nonsmooth Optimization Sequential quadratic programming (SQP)
More informationCan There Be a General Theory of Fourier Integral Operators?
Can There Be a General Theory of Fourier Integral Operators? Allan Greenleaf University of Rochester Conference on Inverse Problems in Honor of Gunther Uhlmann UC, Irvine June 21, 2012 How I started working
More informationExam. Matrikelnummer: Points. Question Bonus. Total. Grade. Information Theory and Signal Reconstruction Summer term 2013
Exam Name: Matrikelnummer: Question 1 2 3 4 5 Bonus Points Total Grade 1/6 Question 1 You are traveling to the beautiful country of Markovia. Your travel guide tells you that the weather w i in Markovia
More information(f(x) P 3 (x)) dx. (a) The Lagrange formula for the error is given by
1. QUESTION (a) Given a nth degree Taylor polynomial P n (x) of a function f(x), expanded about x = x 0, write down the Lagrange formula for the truncation error, carefully defining all its elements. How
More informationA Tutorial on Wavelets and their Applications. Martin J. Mohlenkamp
A Tutorial on Wavelets and their Applications Martin J. Mohlenkamp University of Colorado at Boulder Department of Applied Mathematics mjm@colorado.edu This tutorial is designed for people with little
More informationFast algorithms for Jacobi expansions via nonoscillatory phase functions
Fast algorithms for Jacobi expansions via nonoscillatory phase functions James Bremer a, Haizhao Yang b a Department of Mathematics, University of California, Davis b Department of Mathematics, National
More informationTowards a Mathematical Theory of Super-resolution
Towards a Mathematical Theory of Super-resolution Carlos Fernandez-Granda www.stanford.edu/~cfgranda/ Information Theory Forum, Information Systems Laboratory, Stanford 10/18/2013 Acknowledgements This
More informationNumerical Approximation Methods for Non-Uniform Fourier Data
Numerical Approximation Methods for Non-Uniform Fourier Data Aditya Viswanathan aditya@math.msu.edu 2014 Joint Mathematics Meetings January 18 2014 0 / 16 Joint work with Anne Gelb (Arizona State) Guohui
More informationCLASS NOTES Computational Methods for Engineering Applications I Spring 2015
CLASS NOTES Computational Methods for Engineering Applications I Spring 2015 Petros Koumoutsakos Gerardo Tauriello (Last update: July 27, 2015) IMPORTANT DISCLAIMERS 1. REFERENCES: Much of the material
More information1 Assignment 1: Nonlinear dynamics (due September
Assignment : Nonlinear dynamics (due September 4, 28). Consider the ordinary differential equation du/dt = cos(u). Sketch the equilibria and indicate by arrows the increase or decrease of the solutions.
More informationFast Huygens sweeping methods for Helmholtz equations in inhomogeneous media in the high frequency regime
Fast Huygens sweeping methods for Helmholtz equations in inhomogeneous media in the high frequency regime Songting Luo Jianliang Qian Robert Burridge December 2, 23 Abstract In some applications, it is
More informationAppendix C: Recapitulation of Numerical schemes
Appendix C: Recapitulation of Numerical schemes August 31, 2009) SUMMARY: Certain numerical schemes of general use are regrouped here in order to facilitate implementations of simple models C1 The tridiagonal
More informationWhen is the single-scattering approximation valid? Allan Greenleaf
When is the single-scattering approximation valid? Allan Greenleaf University of Rochester, USA Mathematical and Computational Aspects of Radar Imaging ICERM October 17, 2017 Partially supported by DMS-1362271,
More informationMultiscale Geometric Analysis: Thoughts and Applications (a summary)
Multiscale Geometric Analysis: Thoughts and Applications (a summary) Anestis Antoniadis, University Joseph Fourier Assimage 2005,Chamrousse, February 2005 Classical Multiscale Analysis Wavelets: Enormous
More informationChap. 1. Some Differential Geometric Tools
Chap. 1. Some Differential Geometric Tools 1. Manifold, Diffeomorphism 1.1. The Implicit Function Theorem ϕ : U R n R n p (0 p < n), of class C k (k 1) x 0 U such that ϕ(x 0 ) = 0 rank Dϕ(x) = n p x U
More informationFast algorithms for integral formulations of steady-state radiative transfer equation
Fast algorithms for integral formulations of steady-state radiative transfer equation Yuwei Fan, Jing An, Lexing Ying February 11, 218 Abstract We investigate integral formulations and fast algorithms
More informationContinuous Frames and Sampling
NuHAG University of Vienna, Faculty for Mathematics Marie Curie Fellow within the European network HASSIP HPRN-CT-2002-285 SampTA05, Samsun July 2005 Joint work with Massimo Fornasier Overview 1 Continuous
More informationarxiv: v1 [math.na] 1 Apr 2015
Nonlinear seismic imaging via reduced order model backprojection Alexander V. Mamonov, University of Houston; Vladimir Druskin and Mikhail Zaslavsky, Schlumberger arxiv:1504.00094v1 [math.na] 1 Apr 2015
More informationPartial Differential Equations
Part II Partial Differential Equations Year 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2015 Paper 4, Section II 29E Partial Differential Equations 72 (a) Show that the Cauchy problem for u(x,
More informationMath 671: Tensor Train decomposition methods II
Math 671: Tensor Train decomposition methods II Eduardo Corona 1 1 University of Michigan at Ann Arbor December 13, 2016 Table of Contents 1 What we ve talked about so far: 2 The Tensor Train decomposition
More informationNumerical Methods for geodesic X-ray transforms and applications to open theoretical questions
Numerical Methods for geodesic X-ray transforms and applications to open theoretical questions François Monard Department of Mathematics, University of Washington. Nov. 13, 2014 UW Numerical Analysis Research
More informationEdexcel GCE Further Pure Mathematics (FP1) Required Knowledge Information Sheet. Daniel Hammocks
Edexcel GCE Further Pure Mathematics (FP1) Required Knowledge Information Sheet FP1 Formulae Given in Mathematical Formulae and Statistical Tables Booklet Summations o =1 2 = 1 + 12 + 1 6 o =1 3 = 1 64
More informationRecent Results for Moving Least Squares Approximation
Recent Results for Moving Least Squares Approximation Gregory E. Fasshauer and Jack G. Zhang Abstract. We describe two experiments recently conducted with the approximate moving least squares (MLS) approximation
More informationTHE GAMMA FUNCTION AND THE ZETA FUNCTION
THE GAMMA FUNCTION AND THE ZETA FUNCTION PAUL DUNCAN Abstract. The Gamma Function and the Riemann Zeta Function are two special functions that are critical to the study of many different fields of mathematics.
More informationSolving Boundary Value Problems (with Gaussians)
What is a boundary value problem? Solving Boundary Value Problems (with Gaussians) Definition A differential equation with constraints on the boundary Michael McCourt Division Argonne National Laboratory
More informationA fast and well-conditioned spectral method: The US method
A fast and well-conditioned spectral method: The US method Alex Townsend University of Oxford Leslie Fox Prize, 24th of June 23 Partially based on: S. Olver & T., A fast and well-conditioned spectral method,
More informationMATH 220: MIDTERM OCTOBER 29, 2015
MATH 22: MIDTERM OCTOBER 29, 25 This is a closed book, closed notes, no electronic devices exam. There are 5 problems. Solve Problems -3 and one of Problems 4 and 5. Write your solutions to problems and
More information(0, 0), (1, ), (2, ), (3, ), (4, ), (5, ), (6, ).
1 Interpolation: The method of constructing new data points within the range of a finite set of known data points That is if (x i, y i ), i = 1, N are known, with y i the dependent variable and x i [x
More informationH 2 -matrices with adaptive bases
1 H 2 -matrices with adaptive bases Steffen Börm MPI für Mathematik in den Naturwissenschaften Inselstraße 22 26, 04103 Leipzig http://www.mis.mpg.de/ Problem 2 Goal: Treat certain large dense matrices
More informationConformal maps. Lent 2019 COMPLEX METHODS G. Taylor. A star means optional and not necessarily harder.
Lent 29 COMPLEX METHODS G. Taylor A star means optional and not necessarily harder. Conformal maps. (i) Let f(z) = az + b, with ad bc. Where in C is f conformal? cz + d (ii) Let f(z) = z +. What are the
More informationReport Documentation Page
Report Documentation Page Form Approved OMB No. 0704-0188 Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions,
More informationNumerical Methods I: Interpolation (cont ed)
1/20 Numerical Methods I: Interpolation (cont ed) Georg Stadler Courant Institute, NYU stadler@cims.nyu.edu November 30, 2017 Interpolation Things you should know 2/20 I Lagrange vs. Hermite interpolation
More informationarxiv: v3 [math.na] 30 Jun 2018
An adaptive high order direct solution technique for elliptic boundary value problems P. Geldermans and A. Gillman Department of Computational and Applied Mathematics, Rice University arxiv:70.08787v3
More informationFinite Difference Methods for Boundary Value Problems
Finite Difference Methods for Boundary Value Problems October 2, 2013 () Finite Differences October 2, 2013 1 / 52 Goals Learn steps to approximate BVPs using the Finite Difference Method Start with two-point
More informationSolving the 3D Laplace Equation by Meshless Collocation via Harmonic Kernels
Solving the 3D Laplace Equation by Meshless Collocation via Harmonic Kernels Y.C. Hon and R. Schaback April 9, Abstract This paper solves the Laplace equation u = on domains Ω R 3 by meshless collocation
More informationMatrix assembly by low rank tensor approximation
Matrix assembly by low rank tensor approximation Felix Scholz 13.02.2017 References Angelos Mantzaflaris, Bert Juettler, Boris Khoromskij, and Ulrich Langer. Matrix generation in isogeometric analysis
More informationFast and accurate methods for the discretization of singular integral operators given on surfaces
Fast and accurate methods for the discretization of singular integral operators given on surfaces James Bremer University of California, Davis March 15, 2018 This is joint work with Zydrunas Gimbutas (NIST
More informationJim Lambers MAT 419/519 Summer Session Lecture 13 Notes
Jim Lambers MAT 419/519 Summer Session 2011-12 Lecture 13 Notes These notes correspond to Section 4.1 in the text. Least Squares Fit One of the most fundamental problems in science and engineering is data
More informationQ1 Q2 Q3 Q4 Tot Letr Xtra
Mathematics 54.1 Final Exam, 12 May 2011 180 minutes, 90 points NAME: ID: GSI: INSTRUCTIONS: You must justify your answers, except when told otherwise. All the work for a question should be on the respective
More informationA Fast N-Body Solver for the Poisson(-Boltzmann) Equation
A Fast N-Body Solver for the Poisson(-Boltzmann) Equation Robert D. Skeel Departments of Computer Science (and Mathematics) Purdue University http://bionum.cs.purdue.edu/2008december.pdf 1 Thesis Poisson(-Boltzmann)
More informationDifferential equations, comprehensive exam topics and sample questions
Differential equations, comprehensive exam topics and sample questions Topics covered ODE s: Chapters -5, 7, from Elementary Differential Equations by Edwards and Penney, 6th edition.. Exact solutions
More informationAM 205: lecture 14. Last time: Boundary value problems Today: Numerical solution of PDEs
AM 205: lecture 14 Last time: Boundary value problems Today: Numerical solution of PDEs ODE BVPs A more general approach is to formulate a coupled system of equations for the BVP based on a finite difference
More informationReview and problem list for Applied Math I
Review and problem list for Applied Math I (This is a first version of a serious review sheet; it may contain errors and it certainly omits a number of topic which were covered in the course. Let me know
More informationCollocation approximation of the monodromy operator of periodic, linear DDEs
p. Collocation approximation of the monodromy operator of periodic, linear DDEs Ed Bueler 1, Victoria Averina 2, and Eric Butcher 3 13 July, 2004 SIAM Annual Meeting 2004, Portland 1=Dept. of Math. Sci.,
More informationComputational Linear Algebra
Computational Linear Algebra PD Dr. rer. nat. habil. Ralf-Peter Mundani Computation in Engineering / BGU Scientific Computing in Computer Science / INF Winter Term 2018/19 Part 4: Iterative Methods PD
More informationMATH 425, FINAL EXAM SOLUTIONS
MATH 425, FINAL EXAM SOLUTIONS Each exercise is worth 50 points. Exercise. a The operator L is defined on smooth functions of (x, y by: Is the operator L linear? Prove your answer. L (u := arctan(xy u
More informationInteger multiplication with generalized Fermat primes
Integer multiplication with generalized Fermat primes CARAMEL Team, LORIA, University of Lorraine Supervised by: Emmanuel Thomé and Jérémie Detrey Journées nationales du Calcul Formel 2015 (Cluny) November
More informationNonlinear Integral Equation Formulation of Orthogonal Polynomials
Nonlinear Integral Equation Formulation of Orthogonal Polynomials Eli Ben-Naim Theory Division, Los Alamos National Laboratory with: Carl Bender (Washington University, St. Louis) C.M. Bender and E. Ben-Naim,
More informationAn Empirical Chaos Expansion Method for Uncertainty Quantification
An Empirical Chaos Expansion Method for Uncertainty Quantification Melvin Leok and Gautam Wilkins Abstract. Uncertainty quantification seeks to provide a quantitative means to understand complex systems
More information