How exactly is the merit function weight 'w' cycled in the surrogateopt function algorithm?

Hi everyone,

I'm using the surrogateopt function and my question relates to the manner in which additional evaluation points are chosen by cycling through different values of a weighting factor.

I know that, for the selection of additional evaluation points, surrogateopt utilizes a merit function which returns a weighted combination of two measures. I also understand that surrogateopt cycles through four specific weights (0.3, 0.5, 0.8 and 0.95). The one thing I am still slightly unsure about (and also wasn't able to find in the relevant publications) is whether cycling means that (i) for every sample point one of these four weights is used to compute its merit function value or whether (ii) all four weights are used to compute four merit function values, the best of which is then retained for the current sample point (or some other action is carried out using these four values).

I hope that this question is clear and easy to answer.

Answer 1

[Alan Weiss]

Each sample point uses one weight (one merit function at any one time). The weights cycle after every new point is generated.

Alan Weiss

MATLAB mathematical toolbox documentation