TEST_APPROX
Test Data for Approximation Algorithms


TEST_APPROX is a C library which provides sets of test data for approximation algorithms.

TEST_APPROX contains a number of vectors of data (X(1:N),Y(1:N)) for which no underlying functional relationship is given.

The task of interpolation software is to find, from some given class of functions, the function G(X) which exactly matches the known data values. That is, G(X(1:N)) = Y(1:N).

The task of approximation software is to find, from some given class of functions, the function H(X) for which some approximation error is minimized. There are many forms of error measurement. For instance, the error might simply be the sum of the differences of the function and the data at the data abscissas:

l1(X) = sum ( 1 <= I <= N ) abs ( H(X(I)) - Y(I) )
or the square root of the sum of squares
l2(X) = sqrt ( sum ( 1 <= I <= N ) ( H(X(I)) - Y(I) )^2 )
or the maximum pointwise error:
l_inf(X) = max ( abs ( H(X(I)) - Y(I) ) )
In cases where a functional form is given, the error might be measured in terms of the integral of the absolute value of the difference over some interval:
L1(X,A,B) = integral ( A <= X <= B ) abs ( H(X) - F(X) ) dx
and so on.

The problems available include:

  1. p01_data.png: DeBoor example, Mars position data
  2. p02_data.png: DeBoor example, roughly linear data
  3. p03_data.png: The pulse data, 0 0 0 0 0 1 0 0 0 0 0;
  4. p04_data.png: The jump data, 0 0 0 0 0 1/2 1 1 1 1 1;
  5. p05_data.png: DeBoor's Titanium Property data;
  6. p06_data.png: The Sawtooth data;
  7. p07_data.png: Concavity test data;
  8. p08_data.png: Extrapolation test data;
  9. p09_data.png: Sunspot data, 1700-1960;
  10. p10_data.png: 100 samples of y=2+5x+10*N(0,1), where N(0,1) is a random normal value with 0 mean and unit variance;

Licensing:

The computer code and data files described and made available on this web page are distributed under the GNU LGPL license.

Languages:

TEST_APPROX is available in a C version and a C++ version and a FORTRAN90 version and a MATLAB version.

Related Data and Programs:

BERNSTEIN_POLYNOMIAL, a C library which evaluates the Bernstein polynomials, useful for uniform approximation of functions;

CHEBYSHEV, a C library which computes the Chebyshev interpolant/approximant to a given function over an interval.

DIVDIF, a C library which includes many routines to construct and evaluate divided difference interpolants.

SPLINE, a C library which includes many routines to construct and evaluate spline interpolants and approximants.

TEST_APPROX, a dataset directory which contains sets of data (x,y) for which an approximating formula is desired.

TEST_INTERP_1D, a C library which defines test problems for interpolation of data y(x), depending on a 1D argument.

Reference:

  1. Samuel Conte, Carl deBoor,
    Elementary Numerical Analysis,
    Second Edition,
    McGraw Hill, 1972,
    ISBN: 07-012446-4,
    LC: QA297.C65.
  2. Carl deBoor,
    A Practical Guide to Splines,
    Springer, 2001,
    ISBN: 0387953663,
    LC: QA1.A647.v27.
  3. Max Waldmeier,
    The Sunspot-Activity in the Years 1610-1960,
    Shulthess, 1961,
    LC: QB525.W34.

Source Code:

Examples and Tests:

TEST07 creates data files for an Overhauser spline interpolant to data set 7.

TEST08 creates data files for a cubic spline interpolant to data set 7.

TEST10 creates data files for a B-spline approximant to data set 7.

TEST11 creates data files for a beta spline approximant to data set 7:

TEST12 creates data files for a Bernstein approximant to data set 5.

TEST13 creates data files for a cubic spline interpolant to data set 5.

List of Routines:

You can go up one level to the C source codes.


Last revised on 08 February 2012.