What is the difference between surrogateopt and bayeopt?

I would like to ask the general difference between these 2 algorithms in MATLAB. I understand that Bayesian Optimization uses a Gaussian Process modeling that uses probability theory. In Bayesian Optimization, to my understanding you also use radial basis function to build your gaussian process model (which is sometimes called a surrogate model?), while SurrogateOpt also does that?

It would be very helpful if I can understand the nuances and differences between the two, at a theoretical/MATLAB implementation.

Thank you!

Answer 1

[Alan Weiss]

The algorithms have fairly complete descriptions:

https://www.mathworks.com/help/stats/bayesian-optimization-algorithm.html

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

They are similar in some ways, but differ in a host of details. bayesopt assumes that evaluating an objective function gives a random answer. surrogateopt, like all other optimization algorithms, assumes that evaluating an objective function gives a deterministic answer. The algorithms have so many differences that it is not worthwhile listing them; please, just look at the links I gave. Note that the algorithms behave somewhat differently under parallel computation than serial computation.

Alan Weiss

MATLAB mathematical toolbox documentation