1
Suppose we have equispaced samples of a smooth function f defined on [1,1]. If f is periodic, then the obvious way to interpolate the data is by a trigonometric polynomial. The integral of the interpolant gives the trapezoidal rule quadrature formula. What if f is not periodic? The trigonometric interpolant and the trapezoidal rule are not much good, because of the Gibbs phenomenon. However, we can improve the trapezoidal rule to any order by endpoint corrections. The EulerMaclaurin formula uses corrections involving derivatives, and NewtonGregory quadrature uses onesided finite differences. These ideas go back 300 years. What about using endpoint corrections to improve the interpolant, not just the quadrature formula? Amazingly, so far as we can tell, in 300 years this question has never been considered. In this talk we present such corrections, which we call EulerMaclaurin and NewtonGregory interpolants. Each takes the form of an algebraic polynomial plus a trigonometric polynomial. They approximate f to any order, and their integrals are exactly the results of the EulerMaclaurin and NewtonGregory quadrature formulas. EulerMaclaurin and NewtonGregory Interpolants Tuesday, April 21, 2015 4: 00 PM 5:00 PM 127 HayesHealy Center Colloquium Tea 3:30 PM to 4:00 PM 154 Hurley Hall Nick Trefethen Mathematical Institute Oxford University Department of Applied and Computational Mathematics and Statistics Colloquium Nick Trefethen is Professor of Numerical Analysis and head of the Numerical Analysis Group in the Mathematical Institute at Oxford University. He was educated at Harvard and Stanford and held professorial positions at NYU, MIT, and Cornell before moving to Oxford in 1997. He is a Fellow of the Royal Society and a member of the US National Academy of Engineering.

4'21'15 Abstract Nick Trefethen - University of Notre Dame€¦ ·  · 2016-06-02Nick%Trefethen! Mathematical*Institute ... Microsoft Word - 4'21'15 Abstract Nick Trefethen.docx

Embed Size (px)

Citation preview

Suppose   we   have   equispaced   samples   of   a   smooth   function   f   defined   on   [-­‐1,1].     If   f   is   periodic,   then   the  obvious  way  to  interpolate  the  data  is  by  a  trigonometric  polynomial.  The  integral  of  the  interpolant  gives  the  trapezoidal  rule  quadrature  formula.    

What  if  f  is  not  periodic?    The  trigonometric  interpolant  and  the  trapezoidal  rule  are  not  much  good,  because  of   the   Gibbs   phenomenon.     However,   we   can   improve   the   trapezoidal   rule   to   any   order   by   endpoint  corrections.     The   Euler-­‐Maclaurin   formula   uses   corrections   involving   derivatives,   and   Newton-­‐Gregory  quadrature  uses  one-­‐sided  finite  differences.    These  ideas  go  back  300  years.    

What   about   using   endpoint   corrections   to   improve   the   interpolant,   not   just   the   quadrature   formula?    Amazingly,  so  far  as  we  can  tell,  in  300  years  this  question  has  never  been  considered.    In  this  talk  we  present  such  corrections,  which  we  call  Euler-­‐Maclaurin  and  Newton-­‐Gregory  interpolants.  Each  takes  the  form  of  an  algebraic  polynomial  plus  a  trigonometric  polynomial.    They  approximate  f  to  any  order,  and  their  integrals  are  exactly  the  results  of  the  Euler-­‐Maclaurin  and  Newton-­‐Gregory  quadrature  formulas.

Euler-­‐Maclaurin  and  Newton-­‐Gregory  Interpolants  

                                   Tuesday,  April  21,  2015                                            4:00  PM  –  5:00  PM    

127  Hayes-­‐Healy  Center   Colloquium Tea 3:30 PM to 4:00 PM 154 Hurley Hall

Nick  Trefethen  Mathematical  Institute  

Oxford  University    

Department  of  Applied  and  Computational    Mathematics  and  Statistics  Colloquium  

Nick  Trefethen  is  Professor  of  Numerical  Analysis  and  head  of  the  Numerical  Analysis  Group  in  the  Mathematical  Institute  at  Oxford  University.  He  was  educated  at  Harvard  and  Stanford  and  held  professorial  positions  at  NYU,  MIT,  and  Cornell  before  moving  to  Oxford  in  1997.  He  is  a  Fellow  of  the  Royal  Society  and  a  member  of  the  US  National  Academy  of  Engineering.