Message while building chaos polynomials

Methodology
1

image
image1013×46 2.51 KB

Hello

I am currently building with openturns chaos polynomial metamodels
I have 20 input parameters
See screenshot joined for particular message I see during metamodel building
I build metamodel with a sample of 200 computations (LHS)
I compute predictivity factor Q2 with test sample of 20 other computations
I obtain a Q2 factor very near of 1 so it seems good

I build my metamodel with basic instructions :

chaosalgo = ot.FunctionalChaosAlgorithm(input_sample, output_rsample[:,outputIndex])
chaosalgo.run()
chaosalgo_result = chaosalgo.getResult()
chaosalgo_metamodel = chaosalgo_result.getMetaModel()

thanks in advance for your advises

2

Hi Flore,
This message comes from the part of the algorithm dedicated to the automatic construction of a joint distribution of your input sample. It is done by trying all the continuous parametric models and some of them fail e.g. because you have negative values and try positive models. At the end, if no parametric model is accepted a nonparametric model is built (Histogram). So these messages are harmless and you can simply ignore them.

Cheers

Régis

3

thanks Régis !