Alternative solvers

[behinger]

especially thinking of penalized least squares to automatically fit the number of spline parameters, the penalized toolbox could be interesting
https://github.com/w-mcilhagga/penalized/blob/master/penalized.m

together with the direct formulations for b-spline penalty matrix one could hack up their own version of the penalty function.

In general I find it interesting that by this way one could punish high parameter values more severly than others (e.g. punish very flexible splines, but keep the others intact).