Hello,
I completed the modeling with the Kriging model, but I think the eval function is too slow ，is there a way to get the analytical formula of the model?
Greetings
Frank
Dear @JUN_LAI
I am afraid that this is not possible in general and not just a limitation of UQLab. It is a known problem that as the number of samples increases, not only the time for training the Kriging model increases, but also the time to evaluate it.
However, usually the evaluation is still quite quick. Therefore, I wonder how you evaluate the model. Do you evaluate it in a vectorized way (all samples at once) or one sample at a time?
Best regards
Styfen
Dear Styfen
Sorry, you may have misunderstood me. I want to ask if it is possible to replace the uq_evalModel() function with a mathematical analysis formula. For example, I have established a kriging model, and I want to use the model to calculate, but I don’t want to use the uq_evalModel() function, but want to express it in a mathematical analytical formula. Is this function currently available?
Greetings
Frank
Dear @JUN_LAI
In this case, it is not clear to me what exactly you want to do or achieve. There is not much overhead due to uq_evalModel, i.e. you will hardly gain speed by bypassing it. If the evaluation of your Kriging model is slow, it means that either your experimental design is very large or you are evaluating it on many samples (or both). In such scenarios, almost all of the time is spent doing vectorized operations on large matrices.
However, if you still want more control over this, you have access to all the parameters of the Kriging model to evaluate it yourself. See Table 25 in the manual for more information:
Best regards
Styfen
Dear Frank,
I am afraid there is no explicit mathematical analysis formula for the uq_evalModel(). I think it is a black box. Actually I think uq_evalModel() is very quick as you have the Kriging metamodel, and then u can get the output with 1s, why do you think it is slow?
Best,
TT