For non-linear least squares the Levenberg-Marquadt method works really well. However this is not directly applicable to non-normal distributions where we don't want to minimize the sum of squares. The standard optimizers work /ok/ but aren't great. I haven't found much literature about robust solvers for non-normal distributions, but I wanted to sit down at some point and think about options here.
For non-linear least squares the Levenberg-Marquadt method works really well. However this is not directly applicable to non-normal distributions where we don't want to minimize the sum of squares. The standard optimizers work /ok/ but aren't great. I haven't found much literature about robust solvers for non-normal distributions, but I wanted to sit down at some point and think about options here.
Note to self: minpak LM routine is here