Please could I have more detail on the 'quasirandom' sequence in surrogateopt?

Hi,

I'm investigating the effectiveness of the function 'surrogateopt' by applying it to a series of test tasks involving integer optimisation variables, and constraints. Considering that it's effectiveness is dependent on the intial points, I would like to know more about the 'quasirandom' sequence by which these are sometimes selected. Digging through the code for surrogateopt.m in R2020a, I can see that this sequence is probably generated at line 139 of BlackboxModelOptimizer.m in the call:

nextTrial = globaloptim.bmo.solver(self.trialRequest);

However, I cannot access the code for this function since 'solver' is contained within a 'p file' - solver.p

Is anyone able to give me some more information on this quasirandom sequence?

Is it dependent on the state of the random number generator at the call of surrogateopt()?

Any help appreciated. Thanks

Regards,

Bob

Answer 1

[Rakesh Kumar]

Please take a look at this doc page (section Definitions for Surrogate Optimization)

https://www.mathworks.com/help/gads/surrogate-optimization-algorithm.html

surrogateopt uses Sobol set to create random samples. It is not dependent on the state of MATLAB random number generator.

Thanks,

Rakesh