There is now an Eigenvalue method using the cholesky method you put in the example. I do not really use that, but it seems like a logical inclusion.
I gave the standard fit an option to seht the number of fit parameters by hand, because it does not work automatically
if the function is not consistent between calls.