Anthony Padula, Michael W. Trosset, and Virginia Torczon
While developing numerical algorithms intended to work on complex computer simulations, it is necessary to run many tests on these algorithms in order to ensure their validity and compare the results of different algorithms. However, it is often difficult to obtain many different computer simulations on which to test algorithms. Further, such simulations are often expensive and each tends to have a unique interface.
Thus, it would be useful to be able to create pseudorandom functions that behave like complex computer simulations, but which are fast, have a standard interface, and can be created with specified properties such as taking a given number of inputs and possessing an undelying quadratic trend. To this end we have written the krigifier. It models a stochastic process to generate a set of data with a specified covariance structure, and uses kriging to interpolate this data - thus producing an approximate realization of a stochastic process.
The software posted here is part of an ongoing project to develop a complete, self-contained, C++ class-based implementation of the krigifier. This software was written by Anthony Padula as part of a 2000 Honors Thesis in Mathematics under the direction of Michael W. Trosset and Virginia Torczon.
|Last Updated 6/18/2003||[krig.tgz]|
We suggest obtaining the necessary F2C
library files from the Netlib
repository. Otherwise, you can simply grab the zipped file which
contains all the necessary libraries.
We have some documentation for this software. The documentation is still under review, but we are happy to provide it in its current form. There is compressed postscript file discusses using the Krigifier, Kriging Approximation Class, and the Radial Basis Function Approximation Class. We also have a version of the documentation created using latex2html. Not all of the recent features are discussed, but the basics are covered.
|[Krigifier Documentation (GZIPPED PS)]||[Krigifier Documentation (HTML)]|
|Kriging Approximation with Quadratic Trend||[gls.h]||[gls.cc]|
|Sample Data Files||[krig.dat]||[krig.dat.good]|
|S. Park's Random Number Generator||[rngs.h]||[rngs.c]|
|C.M. Siefert's Parameter Estimation Class||[ParamEstimate.h]||[ParamEstimate.cc]|
|E. Dolan's Pattern Search Class||[PatternSearch.h]||[PatternSearch.cc]|
|E. Dolan's Compass Search Class||[CompassSearch.h]||[CompassSearch.cc]|
|P.L. Shepherd Direct Search Class||[DirectSearch.h]||[DirectSearch.cc]|
|Vector and Matrix Classes||[cppmat.h]||[vec.h]|
|Extra header files||[maps_general.h]||[Dyn_alloc.h]|
|Plot generation utility (requires gnuplot)||[plotpoints.c]|
|Radial Basis Function Approximation Class||[rbf.cc]||[rbf.h]|