I have written a software program for the analysis and optimization of computer models (or simulations). It is called "space" (Stochastic Process Analysis of Computer Experiments) and grew out of my doctoral dissertation under the guidance of William Welch, University of Waterloo.

The code is useful for analysis and global optimization of very expensive functions. The major functions are :

Several responses can be specified, it is possible to minimize one response variable subject to simple constraints on other response variables. Currently, the program does not handle categorical variables, missing values or constraints on x -variables (however, constraints on x-variables might be handled by specifying a second response that represents the x-variable constraint).

The program is time consuming and therefore you may not want to use it with functions that are inexpensive to evaluate. The program is excessively time consuming if the data consists of more than a few hundred points. We have used it with 10 dimensional data. The limiting factor however are not the number of dimensions but the number of data points.

Available for download:

The code works under windows and various unix platforms. I have heard that someone also successfully compiled it under LINUX.

This program was written during my dissseration. It is functional, but is no longer being actively developed.

Licencing Information
The program can be used and distributed free of charge. The distribution must credit Matthias Schonlau, Ph.D. as the author and reference this website ( or ). If distributed through a website the website must contain a hyperlink to my website.

Return to Home Page
Remove navigation bar on the left