Kriging surrogate model in case of multiple outputs about smt HOT 5 CLOSED

Be careful when you have multiple outputs, Gaussian Process in SMT does not consider correlation between outputs, a independant GP is built for each of the output. You could do save the GP as you want (a list of model or a different name for each) for your PSO optimization.

from smt.

Exactly

from smt.

https://github.com/SMTorg/smt/blob/master/tutorial/SMT_EGO_application.ipynb
https://colab.research.google.com/github/SMTorg/smt/blob/master/tutorial/SMT_EGO_application.ipynb

from smt.

Dear @Paul-Saves

Thanks, but I don't see anything concerning the minimization of a multiple output function. Suppose I want to use PSO to minimize the kriging surrogate model, how can I implement this on a list of models? (section 11 of tutorial)
I have a 800 * 9 array as input and 800 * 14 one as output.

https://github.com/SMTorg/smt/blob/master/tutorial/SMT_Tutorial.ipynb

from smt.

Dear @NatOnera

If I understand correctly, in my case, I should build 14 kriging surrogate models, and compile predicted outputs to build the objective function for optimization, correct?

from smt.