Top Banner
Radial Basis Functions for Scientific Computing Grady B. Wright Boise State University * This work is supported by NSF grants DMS 0934581 2014 Montestigliano Workshop
98

Radial Basis Functions for Scientific Computing

Feb 13, 2017

Download

Documents

vuthien
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Radial Basis Functions for Scientific Computing

Radial Basis Functions for Scientific Computing

Grady B. Wright Boise State University

*This work is supported by NSF grants DMS 0934581

2014 Montestigliano Workshop

Page 2: Radial Basis Functions for Scientific Computing

Part I: Introduction Supplementary lecture slides

Grady B. Wright Boise State University

*This work is supported by NSF grants DMS 0934581

2014 Montestigliano Workshop

Page 3: Radial Basis Functions for Scientific Computing

Supplementary material

●  Scattered data interpolation in ●  Positive definite radial kernels: radial basis functions (RBF) ●  Some theory

●  Scattered data interpolation on the sphere ●  Positive definite (PD) zonal kernels ●  Brief review of spherical harmonics ●  Characterization of PD zonal kernels ●  Conditionally positive definite zonal kernels ●  Examples

●  Error estimates: ●  Reproducing kernel Hilbert spaces ●  Sobolev spaces ●  Native spaces ●  Geometric properties of node sets

●  Optimal nodes on the sphere

Overview

Page 4: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels

Examples:

Page 5: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels

Examples:

Page 6: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels

Page 7: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels

Page 8: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 9: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 10: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 11: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 12: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 13: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Page 14: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

Key idea: linear combination of translates and rotations of a single radial kernel:

Linear system for determining the interpolation coefficients

Page 15: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

•  Some results on positive definite radial kernels.

Page 16: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

•  Some results on positive definite radial kernels.

Page 17: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

•  Some results on positive definite radial kernels.

Page 18: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

•  Some results on positive definite radial kernels.

Page 19: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

•  Some results on positive definite radial kernels.

Page 20: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

Examples: Gaussian

Inverse multiquadric

�(r) = exp(�("r)2)

Inverse quadratic

•  Some results on positive definite radial kernels.

Page 21: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

●  Results on dimensions specific positive definite radial kernels:

Page 22: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite radial kernels

●  Examples

Matérn

Wendland (1995)

Truncated powers

J-Bessel

Finite-smoothness

Infinite-smoothness

Platte

Page 23: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

●  Discussion thus far does not cover many important radial kernels:

●  These can covered under the theory of conditionally positive definite kernels.

●  CPD kernels can be characterized similar to PD kernels but, using generalized Fourier transforms. We will not take this approach; see Ch. 8 Wendland 2005 for details.

●  We will instead use a generalization of completely monotone functions.

Cubic Thin plate spline Multiquadric

Cubic spline in 1-D Generalization of energy minimizing spline in 2D

Popular kernel and first used in any RBF application; Hardy 1971

Page 24: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 25: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 26: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

k=1 k=2

Page 27: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

k=1 k=2

Page 28: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 29: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 30: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 31: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite kernels

Page 32: Radial Basis Functions for Scientific Computing

Supplementary material Radial basis function (RBF) interpolation

1999   2003  

2005   2007  

2004  

A Primer on Radial Basis

Functions with Applications to

the Geosciences

Bengt Fornberg Natasha Flyer  

2014:  SIAM  

●  Many good books to consult further on RBF theory and applications:

Page 33: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels (revisited)

Page 34: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation with kernels on the sphere

Page 35: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 36: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 37: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 38: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 39: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 40: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Page 41: Radial Basis Functions for Scientific Computing

Supplementary material SBF interpolation

Key idea: linear combination of translates and rotations of a single zonal kernel on

Linear system for determining the interpolation coefficients

Page 42: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

Page 43: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

Page 44: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

Page 45: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

Page 46: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

Page 47: Radial Basis Functions for Scientific Computing

Supplementary material Positive definite zonal kernels

●  Some references for the material to come:

Page 48: Radial Basis Functions for Scientific Computing

Supplementary material Spherical harmonics

●  A good understanding of functions on the sphere requires one to be well- versed in spherical harmonics.

●  Spherical harmonics are the analog of 1-D Fourier series for approximation on spheres of dimension 2 and higher.

●  Several ways to introduce spherical harmonics (Freeden & Schreiner 2008)

●  We will use the eigenfunction approach and restrict our attention to the 2-sphere.

●  Following this we review some important results about spherical harmonics.

Page 49: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 50: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 51: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 52: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 53: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 54: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

m=-3 m=-2 m=-1 m=0 m=1 m=2 m=3

Page 55: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

m=-4 m=-3 m=-2 m=-1 m=0 m=1 m=2 m=3 m=4

Page 56: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 57: Radial Basis Functions for Scientific Computing

Supplementary material Overview of spherical harmonics

Page 58: Radial Basis Functions for Scientific Computing

Supplementary material Theorems for positive definite zonal kernels

Page 59: Radial Basis Functions for Scientific Computing

Supplementary material Theorems for positive definite zonal kernels

Page 60: Radial Basis Functions for Scientific Computing

Supplementary material Theorems for positive definite zonal kernels

Page 61: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite zonal kernels

Page 62: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite zonal kernels

Page 63: Radial Basis Functions for Scientific Computing

Supplementary material Conditionally positive definite zonal kernels

Page 64: Radial Basis Functions for Scientific Computing

Supplementary material Spherical Fourier coefficients

Page 65: Radial Basis Functions for Scientific Computing

Supplementary material Examples of positive definite zonal kernels

Page 66: Radial Basis Functions for Scientific Computing

Supplementary material Error estimates

Page 67: Radial Basis Functions for Scientific Computing

Supplementary material Reproducing kernel Hilbert spaces

Page 68: Radial Basis Functions for Scientific Computing

Supplementary material Reproducing kernel Hilbert spaces

Page 69: Radial Basis Functions for Scientific Computing

Supplementary material Reproducing kernel Hilbert spaces

Page 70: Radial Basis Functions for Scientific Computing

Supplementary material Reproducing kernel Hilbert spaces

Page 71: Radial Basis Functions for Scientific Computing

Supplementary material Sobolev spaces

Page 72: Radial Basis Functions for Scientific Computing

Supplementary material Sobolev spaces

Page 73: Radial Basis Functions for Scientific Computing

Supplementary material Native spaces

Page 74: Radial Basis Functions for Scientific Computing

Supplementary material Native spaces

Page 75: Radial Basis Functions for Scientific Computing

Supplementary material Native spaces

Page 76: Radial Basis Functions for Scientific Computing

Supplementary material Native spaces

Page 77: Radial Basis Functions for Scientific Computing

Supplementary material Geometric properties of node sets

●  The following properties for node sets on the sphere appear in the error estimates:

(Only part of the sphere is shown)

Page 78: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation error estimates

Theorem. Target functions in the native space.

Notation:

●  We start with known error estimates for kernels of finite smoothness.

Page 79: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation error estimates

Theorem. Target functions twice as smooth as the native space.

Notation:

●  We start with known error estimates for kernels of finite smoothness.

Remark. Known as the “doubling trick” from spline theory. (Schaback 1999)

Page 80: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation error estimates

Notation:

●  We start with known error estimates for kernels of finite smoothness.

Remark. (1)  Referred to as “escaping the native space”. (Narcowich, Ward, & Wendland (2005, 2006)).

(2)  These rates are the best possible.

Theorem. Target functions rougher than the native space.

Page 81: Radial Basis Functions for Scientific Computing

Supplementary material Interpolation error estimates

●  Error estimates for infinitely smooth kernels (e.g. Gaussian, multiquadric).

Remarks: (1)  This is called spectral (or exponential) convergence. (2)  Function space may be small, but does include all band-limited functions. (3)  Only known result I am aware of (too bad there are not more). (4)  Numerical results indicate convergence is also fine for less smooth functions.

Notation:

Theorem. Target functions in the native space.

Page 82: Radial Basis Functions for Scientific Computing

Supplementary material Optimal nodes

●  If one has the freedom to choose the nodes, then the error estimates indicate they should be roughly as evenly spaced as possible.

Icosahedral Fibonacci Equal area

Minimum energy s=2 Minimum energy, s=3 Maximal determinant

Swinbank & Purser (2006) Saff & Kuijlaars (1997)

Hardin & Saff (2004) Womersley & Sloan (2001)

Examples:

Det

erm

inis

tic

Non

-det

erm

inis

tic

Page 83: Radial Basis Functions for Scientific Computing

Supplementary material What about the shape parameter?

●  Smooth kernels with a shape parameter.

�(r) = exp(�("r)2)Ex:

Linear system for determining the interpolation coefficients

RBF-Direct

Page 84: Radial Basis Functions for Scientific Computing

Supplementary material RBF interpolation in the “flat” limit

RBF interpolant:

Page 85: Radial Basis Functions for Scientific Computing

Supplementary material Base vs. space

Analogy: (Fornberg)

Vectors  

Polynomials  

Splines  

Page 86: Radial Basis Functions for Scientific Computing

Supplementary material Using a bad basis for flat kernels:

Page 87: Radial Basis Functions for Scientific Computing

Supplementary material Using a good basis for flat kernels:

Page 88: Radial Basis Functions for Scientific Computing

Supplementary material Behavior of interpolants in the flat limit

Page 89: Radial Basis Functions for Scientific Computing

Supplementary material Behavior of interpolants in the flat limit

Page 90: Radial Basis Functions for Scientific Computing

Supplementary material Uncertainty principle misconception

●  Schaback’s uncertainty principle:

Principle: One cannot simultaneously achieve good conditioning and high accuracy. Misconception: Accuracy that can be achieved is limited by ill-conditioning.

Restatement: One cannot simultaneously achieve good conditioning and high accuracy

when using the standard basis.

●  It’s a matter of base vs. space. ●  Literature for interpolation with “flat” kernels is growing:

Driscoll & Fornberg (2002) Larsson & Fornberg (2003; 2005) Fornberg, Wright, & Larsson (2004) Schaback (2005; 2008) Platte & Driscoll (2005) Fornberg, Larsson, & Wright (2006) deBoor (2006) Fornberg & Zuev (2007) Lee, Yoon, & Yoon (2007) Fornberg & Piret (2008) Buhmann, Dinew, & Larsson (2010) Platte (2011) Song, Riddle, Fasshauer, & Hickernell (2011)

Fornberg & Wright (2004) Fornberg & Piret (2007) Fornberg, Larsson, & Flyer (2011) Fasshauer & McCourt (2011) Gonnet, Pachon, & Trefethen (2011) Pazouki & Schaback (2011) De Marchi & Santin (2013) Fornberg, Letho, Powell (2013) Wright & Fornberg (2013)

Theory: Stable algorithms:

Page 91: Radial Basis Functions for Scientific Computing

Supplementary material

Better kernel bases for the sphere: RBF-QR algorithm

Page 92: Radial Basis Functions for Scientific Computing

Supplementary material

Better kernel bases for the sphere: RBF-QR algorithm

Page 93: Radial Basis Functions for Scientific Computing

Supplementary material

Better kernel bases for the sphere: RBF-QR algorithm

E =

2

66666664

1"2

"2

"2

"4

. . .

3

77777775

Page 94: Radial Basis Functions for Scientific Computing

Supplementary material

Better kernel bases for the sphere: RBF-QR algorithm

Page 95: Radial Basis Functions for Scientific Computing

Supplementary material

Better kernel bases for the sphere: RBF-QR algorithm

Page 96: Radial Basis Functions for Scientific Computing

Supplementary material Numerical example

Page 97: Radial Basis Functions for Scientific Computing

Supplementary material Remarks on RBF-QR

●  RBF-QR allows one to stably compute “flat” kernel interpolants on the sphere.

●  One can reach full numerical precision using this procedure (for smooth enough target functions and large enough N)

●  It is more expensive than standard approach (RBF-Direct).

●  Work has gone into extending this idea to general Euclidean space, but the procedure is much more complicated.

●  Matlab Code for RBF-QR is given in Fornberg & Piret (2007) and is available in the rbfsphere package.

Page 98: Radial Basis Functions for Scientific Computing

Supplementary material Concluding remarks

●  This was general background material for getting started in this area. ●  There is still much more to learn and many interesting problems:

o  Approximation (and decomposition) of vector fields. o  Fast algorithms for interpolation using localized bases o  Numerical integration o  RBF generated finite differences o  RBF partition of unity methods o  Numerical solution of partial differential equations on spheres. o  Generalizations to other manifolds.

v  If you have any questions or want to chat about research ideas, please come and talk to me.