Julien Schueller

we could also handle multivariate scipy distributions: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.vonmises_fisher.html but there seem to be no way to get the range ``` import scipy.stats as st d = 3 mu=[d**-0.5]*d scipy_dist =...

scipy says also different results depending on the interpolation method, but I thought we had some kind of linear method but scipy says 5.2 for the linear method, but maybe...

ok, I can reproduce with the (modified) tinybuild example, when numpydoc is enabled it yields a missing ref of an excluded attribute: ``` numpydoc_test_module.py: class MyClass: """A class.""" thisown =...

2 problems: - same attribute for normalization factor in natural + log scale

unfortunately general optimization algorithms are just not capable of solving these efficiently ; it can go wrong with just a few slack variables, for example: https://github.com/openturns/openturns/issues/3031

I'm still skeptical but happy to be proven wrong. If you want to try solving calibration algorithms with evolutionary algorithms you may want to comment this condition: ``` void Pagmo::checkProblem(const...

there is a problem to use pagmo here as NonLinearLeastSquaresCalibration yields an unbound problem, which is not supported by pagmo

you can do: ``` git checkout upstream/master [filepath] ``` then commit (the command stages the changes)

you could create a dedicated repo with these then link it from https://github.com/openturns/openturns/wiki/Bibliography for example

@mbaudin47 do you want to take Mathieu comments into account before we include this ?