This text is written around the core problem of attempting to learn a mapping y=f(x) that lives in a black box, which obscures from us the physics that converts the vector x into a scalar output y. This black box could take the form of either a physical or computer experiment, for example a finite element code, which calculates the maximum stress (f) for given product dimensions (x). The generic solution method is to collect the output values y(1),y(2),…,y(n) that result from a set of inputs x(1),x(2),…,x(n) and find a best guess ^f(x) for the black box mapping f, based on these known observations.
In this chapter we discuss the fundamentals and some of the technical minutiae of a number of specific surrogate model types capable of accomplishing this learning process. We begin, however, with a generic discussion of the key stages of the surrogate model building process.
To get started, download the Matlab toolbox, try one of the examples in the book and/or run an example script, e.g.
The Matlab code below creates a Kriging model of the ‘Branin function’ and produces the contour plot of the Kriging prediction at the top of this page.
% no. of dimensions
% no. of sample points
% set up global data structure
% space filling sample
% observed data
% maximise likelihood
% plot Kriging model