In applied mathematics, polyharmonic splines are used for function approximation and data interpolation. They are very useful for interpolating and fitting scattered data in many dimensions. Special cases include thin plate splines[1][2] and natural cubic splines in one dimension.[3]

Definition

edit

A polyharmonic spline is a linear combination of polyharmonic radial basis functions (RBFs) denoted by   plus a polynomial term:

  (1)

where

 
Polyharmonic basis functions
  •   (  denotes matrix transpose, meaning   is a column vector) is a real-valued vector of   independent variables,
  •   are   vectors of the same size as   (often called centers) that the curve or surface must interpolate,
  •   are the   weights of the RBFs,
  •   are the   weights of the polynomial.

The polynomial with the coefficients   improves fitting accuracy for polyharmonic smoothing splines and also improves extrapolation away from the centers   See figure below for comparison of splines with polynomial term and without polynomial term.

The polyharmonic RBFs are of the form:

 

Other values of the exponent   are not useful (such as  ), because a solution of the interpolation problem might not exist. To avoid problems at   (since  ), the polyharmonic RBFs with the natural logarithm might be implemented as:

 

or, more simply adding a continuity extension in  

 

The weights   and   are determined such that the function interpolates   given points   (for  ) and fulfills the   orthogonality conditions

 

All together, these constraints are equivalent to the symmetric linear system of equations

  (2)

where

 

In order for this system of equations to have a unique solution,   must be full rank.   is full rank for very mild conditions on the input data. For example, in two dimensions, three centers forming a non-degenerate triangle ensure that   is full rank, and in three dimensions, four centers forming a non-degenerate tetrahedron ensure that B is full rank. As explained later, the linear transformation resulting from the restriction of the domain of the linear transformation   to the null space of   is positive definite. This means that if   is full rank, the system of equations (2) always has a unique solution and it can be solved using a linear solver specialised for symmetric matrices. The computed weights allow evaluation of the spline for any   using equation (1). Many practical details of implementing and using polyharmonic splines are explained in Fasshauer.[4] In Iske[5] polyharmonic splines are treated as special cases of other multiresolution methods in scattered data modelling.

Discussion

edit

The main advantage of polyharmonic spline interpolation is that usually very good interpolation results are obtained for scattered data without performing any "tuning", so automatic interpolation is feasible. This is not the case for other radial basis functions. For example, the Gaussian function   needs to be tuned, so that   is selected according to the underlying grid of the independent variables. If this grid is non-uniform, a proper selection of   to achieve a good interpolation result is difficult or impossible.

Main disadvantages are:

  • To determine the weights, a dense linear system of equations must be solved. Solving a dense linear system becomes impractical if the dimension   is large, since the memory required is   and the number of operations required is  
  • Evaluating the computed polyharmonic spline function at   data points requires   operations. In many applications (image processing is an example),   is much larger than   and if both numbers are large, this is not practical.

Fast construction and evaluation methods

edit

One straightforward approach to speeding up model construction and evaluation is to use a subset of   nearest interpolation nodes to build a local model every time we evaluate the spline. As a result, the total time needed for model construction and evaluation at   points changes from   to  . This can yield better timings if   is much less than  . Such an approach is advocated by some software libraries, the most notable being scipy.interpolate.RBFInterpolator. The main drawback is that it introduces small discontinuities in the spline and requires problem-specific tuning: a proper choice of the neighbors count,  . Recently, methods have been developed to overcome the aforementioned difficulties without sacrificing main advantages of polyharmonic splines.

First, a bunch of methods for fast   evaluation were proposed:

  • Beatson et al.[6] present a method to interpolate polyharmonic splines with   being a basis function at one point in 3 dimensions or less
  • Cherrie et al. [7] present a method to interpolate polyharmonic splines with   as a basis function at one point in 4 dimensions or less

Second, an accelerated model construction by applying an iterative solver to an ACBF-preconditioned linear system was proposed by Brown et al.[8] This approach reduces running time from   to  , and further to   when combined with accelerated evaluation techniques.

The approaches above are often employed by commercial geospatial data analysis libraries and by some open source implementations (e.g. ALGLIB). Sometimes domain decomposition methods are used to improve asymptotic behavior, reducing memory requirements from   to  , thus making polyharmonic splines suitable for datasets with more than 1.000.000 points.

Reason for the name "polyharmonic"

edit

A polyharmonic equation is a partial differential equation of the form   for any natural number  , where   is the Laplace operator. For example, the biharmonic equation is   and the triharmonic equation is  . All the polyharmonic radial basis functions are solutions of a polyharmonic equation (or more accurately, a modified polyharmonic equation with a Dirac delta function on the right hand side instead of 0). For example, the thin plate radial basis function is a solution of the modified 2-dimensional biharmonic equation.[9] Applying the 2D Laplace operator ( ) to the thin plate radial basis function   either by hand or using a computer algebra system shows that  . Applying the Laplace operator to   (this is  ) yields 0. But 0 is not exactly correct. To see this, replace   with   (where   is some small number tending to 0). The Laplace operator applied to   yields  . For   the right hand side of this equation approaches infinity as   approaches 0. For any other  , the right hand side approaches 0 as   approaches 0. This indicates that the right hand side is a Dirac delta function. A computer algebra system will show that

 

So the thin plate radial basis function is a solution of the equation  .

Applying the 3D Laplacian ( ) to the biharmonic RBF   yields   and applying the 3D   operator to the triharmonic RBF   yields  . Letting   and computing   again indicates that the right hand side of the PDEs for the biharmonic and triharmonic RBFs are Dirac delta functions. Since

 

the exact PDEs satisfied by the biharmonic and triharmonic RBFs are   and  .

Polyharmonic smoothing splines

edit

Polyharmonic splines minimize

  (3)

where   is some box in   containing a neighborhood of all the centers,   is some positive constant, and   is the vector of all  th order partial derivatives of   For example, in 2D   and   and in 3D  . In 2D   making the integral the simplified thin plate energy functional.

To show that polyharmonic splines minimize equation (3), the fitting term must be transformed into an integral using the definition of the Dirac delta function:

 

So equation (3) can be written as the functional

 

where   is a multi-index that ranges over all partial derivatives of order   for   In order to apply the Euler–Lagrange equation for a single function of multiple variables and higher order derivatives, the quantities

 

and

 

are needed. Inserting these quantities into the E−L equation shows that

  (4)

A weak solution   of (4) satisfies

 

(5)

for all smooth test functions   that vanish outside of   A weak solution of equation (4) will still minimize (3) while getting rid of the delta function through integration.[10]

Let   be a polyharmonic spline as defined by equation (1). The following calculations will show that   satisfies (5). Applying the   operator to equation (1) yields

 

where     and   So (5) is equivalent to

  (6)

The only possible solution to (6) for all test functions   is

  (7)

(which implies interpolation if  ). Combining the definition of   in equation (1) with equation (7) results in almost the same linear system as equation (2) except that the matrix   is replaced with   where   is the   identity matrix. For example, for the 3D triharmonic RBFs,   is replaced with  

Explanation of additional constraints

edit

In (2), the bottom half of the system of equations ( ) is given without explanation. The explanation first requires deriving a simplified form of   when   is all of  

First, require that   This ensures that all derivatives of order   and higher of   vanish at infinity. For example, let   and   and   be the triharmonic RBF. Then   (considering   as a mapping from   to  ). For a given center  

 

On a line   for arbitrary point   and unit vector  

 

Dividing both numerator and denominator of this by   shows that   a quantity independent of the center   So on the given line,

 

It is not quite enough to require that   because in what follows it is necessary for   to vanish at infinity, where   and   are multi-indices such that   For triharmonic     (where   and   are the weights and centers of  ) is always a sum of total degree 5 polynomials in     and   divided by the square root of a total degree 8 polynomial. Consider the behavior of these terms on the line   as   approaches infinity. The numerator is a degree 5 polynomial in   Dividing numerator and denominator by   leaves the degree 4 and 5 terms in the numerator and a function of   only in the denominator. A degree 5 term divided by   is a product of five   coordinates and   The   (and  ) constraint makes this vanish everywhere on the line. A degree 4 term divided by   is either a product of four   coordinates and an   coordinate or a product of four   coordinates and a single   or   coordinate. The   constraint makes the first type of term vanish everywhere on the line. The additional constraints   will make the second type of term vanish.

Now define the inner product of two functions   defined as a linear combination of polyharmonic RBFs   with   and   as

 

Integration by parts shows that

 

(8)

For example, let   and   Then

 

(9)

Integrating the first term of this by parts once yields

 

since   vanishes at infinity. Integrating by parts again results in  

So integrating by parts twice for each term of (9) yields

 

Since   (8) shows that

 

So if   and  

 

(10)

Now the origin of the constraints   can be explained. Here   is a generalization of the   defined above to possibly include monomials up to degree   In other words,   where   is a column vector of all degree   monomials of the coordinates of   The top half of (2) is equivalent to   So to obtain a smoothing spline, one should minimize the scalar field   defined by

 

The equations

 

and

 

(where   denotes row   of  ) are equivalent to the two systems of linear equations   and   Since   is invertible, the first system is equivalent to   So the first system implies the second system is equivalent to   Just as in the previous smoothing spline coefficient derivation, the top half of (2) becomes  

This derivation of the polyharmonic smoothing spline equation system did not assume the constraints necessary to guarantee that   But the constraints necessary to guarantee this,   and   are a subset of   which is true for the critical point   of   So   is true for the   formed from the solution of the polyharmonic smoothing spline equation system. Because the integral is positive for all   the linear transformation resulting from the restriction of the domain of linear transformation   to   such that   must be positive definite. This fact enables transforming the polyharmonic smoothing spline equation system to a symmetric positive definite system of equations that can be solved twice as fast using the Cholesky decomposition.[9]

Examples

edit

The next figure shows the interpolation through four points (marked by "circles") using different types of polyharmonic splines. The "curvature" of the interpolated curves grows with the order of the spline and the extrapolation at the left boundary (x < 0) is reasonable. The figure also includes the radial basis functions φ = exp(−r2) which gives a good interpolation as well. Finally, the figure includes also the non-polyharmonic spline phi = r2 to demonstrate, that this radial basis function is not able to pass through the predefined points (the linear equation has no solution and is solved in a least squares sense).

 
Interpolation with different polyharmonic splines that shall pass the 4 predefined points marked by a circle (the interpolation with phi = r2 is not useful, since the linear equation system of the interpolation problem has no solution; it is solved in a least squares sense, but then does not pass the centers)

The next figure shows the same interpolation as in the first figure, with the only exception that the points to be interpolated are scaled by a factor of 100 (and the case phi = r2 is no longer included). Since φ = (scale·r)k = (scalekrk, the factor (scalek) can be extracted from matrix A of the linear equation system and therefore the solution is not influenced by the scaling. This is different for the logarithmic form of the spline, although the scaling has not much influence. This analysis is reflected in the figure, where the interpolation shows not much differences. Note, for other radial basis functions, such as φ = exp(−kr2) with k = 1, the interpolation is no longer reasonable and it would be necessary to adapt k.

 
The same interpolation as in the first figure, but the points to be interpolated are scaled by 100

The next figure shows the same interpolation as in the first figure, with the only exception that the polynomial term of the function is not taken into account (and the case phi = r2 is no longer included). As can be seen from the figure, the extrapolation for x < 0 is no longer as "natural" as in the first figure for some of the basis functions. This indicates, that the polynomial term is useful if extrapolation occurs.

 
The same interpolation as in the first figure, but without the polynomial term

See also

edit

References

edit
  1. ^ R.L. Harder and R.N. Desmarais: Interpolation using surface splines. Journal of Aircraft, 1972, Issue 2, pp. 189−191
  2. ^ J. Duchon: Splines minimizing rotation-invariant semi-norms in Sobolev spaces. Constructive Theory of Functions of Several Variables, W. Schempp and K. Zeller (eds), Springer, Berlin, pp. 85−100
  3. ^ Wendland, Holger (2005). Scattered Data Approximation. Cambridge University Press. p. 9. ISBN 0521843359.
  4. ^ G.F. Fasshauer G.F.: Meshfree Approximation Methods with MATLAB. World Scientific Publishing Company, 2007, ISPN-10: 9812706348
  5. ^ A. Iske: Multiresolution Methods in Scattered Data Modelling, Lecture Notes in Computational Science and Engineering, 2004, Vol. 37, ISBN 3-540-20479-2, Springer-Verlag, Heidelberg.
  6. ^ R.K. Beatson, M.J.D. Powell, and A.M. Tan: Fast evaluation of polyharmonic splines in three dimensions. IMA Journal of Numerical Analysis, 2007, 27, pp. 427–450.
  7. ^ J. B. Cherrie; R. K. Beatson; D. L. Ragozin (2000), Fast evaluation of radial basis functions: methods for four-dimensional polyharmonic splines
  8. ^ Damian Brown; Leevan Ling; Edward Kansa; Jeremy Levesley (2000), On Approximate Cardinal Preconditioning Methods for Solving PDEs with Radial Basis Functions
  9. ^ a b Powell, M. J. D. (1993). "Some algorithms for thin plate spline interpolation to functions of two variables" (PDF). Cambridge University Dept. Of Applied Mathematics and Theoretical Physics Technical Report. Archived from the original (PDF) on 2016-01-25. Retrieved January 7, 2016.
  10. ^ Evans, Lawrence (1998). Partial Differential Equations. Providence: American Mathematical Society. pp. 450−452. ISBN 0-8218-0772-2.
edit

Computer Code