Régis LEBRUN

Hi @mbaudin47 , You are facing the curse of dimensionality! Let's consider S, it is a sum of two continuous atoms `(x3**2 * X4**2 / 16)` and `(X5**2)` with no...

Hi @mbaudin47 , Nice trip in distribution arithmetic! Your suggestion to use a ChiSquare distribution whenever possible is a very good idea. For now, almost all the simplifications are based...

S is definite positive, but not its transform as a correlation for a normal copula. To make things clearer, given a Spearman correlation S, if it can be obtained as...

The first step is to ask yourself from where this Spearman correlation comes from. If it has been estimated from data, there is a ball in which any correlation matrix...

Yes, it is because one have to classify the origin wrt the failure domain. The approximation of the probability depends on this classification, see #850 for another example of this...

@josephmure Initially we were relying on R to implement the draw() method of graphs, and within R these point styles are different. Unfortunately, matplotlib does not allow to set the...

So the View class should be updated.

Hi, You should read https://www.esaim-m2an.org/articles/m2an/pdf/2012/02/m2an110045.pdf to have an in-depth understanding of what happen when you try to use the orthogonal polynomial family associated to the log-normal distribution. To sum up...

I found many more or less important ways to speed-up things with respect to your code. I started to implement it as a complement of your branch, but as I...

Hi, When I run your benchmark on my laptop (HP ZBOOK with Intel(R) Xeon(R) E-2286M CPU @ 2.40GHz, numpy 1.26.1 for python 3.10.11) I get: ``` Time numpy: 8.23 Time...