Jonathan Feinberg

I apologies for being vague, but I prefer not to have my e-mail address listed in a public site. I see I used HTML tags `` in my post, which...

I recieved your e-mail now. I'l take a look at in the next couple of days.

Okay, side note to the original question: Am I understanding you correctly that you are trying to improvise MV-Gaussian mixture model by summing MV-Gaussian random variables? I ask, but that...

[Sum of random variables](https://en.wikipedia.org/wiki/Sum_of_normally_distributed_random_variables) should not be confused with [sum of distributions](https://en.wikipedia.org/wiki/Mixture_distribution). In a nutshell: `f_{Z=X+Y}(z) != f_X(z)+f_Y(z)`. But you are in luck, GMM and KDE is basically the same...

It is not in 4.0.0, it is in the branch `kde` which is what #272 is a PR of. To install chaospy from the `kde` branch without an release run:...

Thank you for thinking so.

For this work I perfer that you cite both the chaospy paper (because that is the framework you have used), and this paper: https://www.researchgate.net/publication/322691700_Multivariate_Polynomial_Chaos_Expansions_with_Dependent_Variables This outlines (amongst a couple of...

If the data results in non-positive definite covariance matrix, you likely have linear dependencies in your data. Run: ```python lambdas, _ = np.linalg.eig(data) ``` (or if your data is transposed:...

If the data columns are linear dependent of each other, they do not contribute any information, and you can safely remove them: ```python data = data[lambdas != 0] ```

As you are hinting to, if you want to do multivariate, you need to be handling dependencies. Getting GMM and KDE with dependencies supported in chaospy is non-trivial. I am...