markus-nilsson
commented
7 years ago Found two things when testing the code
1) Generated .nii uFA can become complex, which is not handled by many viewers. I think this needs to be managed, in particular since uFA is a "display parameter" and not a model parameter as such
2) It would be good and convenient to have the random guess initialization implemented as a msf_fit_something function.
Updated dtd_gamma to support negative variance