Drawing from multivariate normal distribution

[DJDuque]

Background

I have looked at the documentation, and it looks like this is currently not possible (?)

What is your motivation?
I keep going back to Python and Numpy's multivariate_normal whenever I need this, and I was wondering if it was within the scope of the rand_distr crate.

Feature request

I would like to be able to draw samples from a multivariate normal distribution given a covariance matrix.

[dhardy]

I believe you are correct; neither rand_distr or statrs support this. rand_distr has a few multivariate distributions (e.g. Dirichlet, unit sphere).

Wikipedia explains how to sample from this distribution — but requires the Cholesky decomposition (available in some other crates).

Yes, ideally we'd implement this in rand_distr. I'll leave this issue open, but don't expect much progress (unless you'd like to provide).

It also brings up the question of whether rand_distr should depend on a linear algebra library. IIRC this may have been brought up before. Possibly it makes sense to start a new crate for multi-variate distributions. @vks?

[TheIronBorn]

Is this from statrs not sufficient? https://docs.rs/statrs/latest/statrs/distribution/struct.MultivariateNormal.html#impl-MultivariateNormal it supports a covariance matrix