Using non diagonal noise.

[TyannDumerchat]

Is it possible to parametrize a the noise component of the kernel with something else than a diagonal white noise ?
I have some knowledge about the noise structure in my training set and I know it is correlated.
Would it be possible to use for exemple a parametrized RBF kernel as the noise part of my model ?

More precisely I would like to use the GPy.models.gp_kronecker_gaussian_regression class to train a multioutput GP, and as I know the structure of the noise matrix for one of my subspaces, I would like to use something else than just kappa.I which is white noise.

Many thanks.