Beta distribution in AK-MCS

UQ with UQLab Community Q&A and How To
,
1

Hello,
I know that the input variables when applying the Kriging model should conform to the normal distribution.
Should I transform the input variables to a normal distribution in UQlab if my input variables are Beta distribution or Triangular distribution? Or I just write them as follows and don’t need to do anything when using AK-MCS:

InputOpts.Marginals(1).Name = ‘X1’;
InputOpts.Marginals(1).Type = ‘Beta’;
InputOpts.Marginals(1).Parameters = [5 10 0.25 1];

Thank you.

2

Dear Yuf Sun

There is a separate section in the Kriging documentation about scaling (2.10 Performing Kriging on an auxiliary space (scaling)) :

Please let me know if this does not properly answer your question.

Best regards
Styfen

3

Hi @YUF,

As far as I know, the input variables to the Kriging model does not need to conform to a normal distribution. It is the output distribution that are modeled as Gaussian random process with respect to the inputs.

For details regarding scaling the input space for other reasons, please refer to @styfen.schaer’s answer.

I hope this helps!

4

Thank you for your kind reply. My problem has been solved.

5

Thank you for your reply. I had a wrong understanding with Kriging before.