Re: lsqminnorm vs lsqnonneg

[Juan Pablo Carbajal]

> Octave provides already the "lsqnonneg" optimzation function, yet misses 
> "lsqminnorm". I assume ``just'' solving the least-squares problem would be 
> easier than a constraint least-squares. So I wonder why the more complicated 
> function is implemented, and the potentially easier one is not. Is there a 
> reason for that, for example algorithmic issues? Or is this maybe a 
> low-hanging (m-file) fruit?

I cannot answer for the developers of the function, but I would guess
that the creation of a function is mainly driven by need, and secondly
by the first Octave directive (matlab drop-in compatible). It is
likely that functions are developed by people that need them, but
since they use octave, they might not be motivated to spend time
implementing a matlab compatible API. This second task usually falls
in the hand of highly motivated core developers or people who need
that their code also runs in matlab (collaborators, visibility,
etc...).
If you can create an m-file that provides a matlab compatible API for
lsqminnorm functions (it is probably just a wrapper around an already
existing function, like curvefit, etc), I assume it will be very
welcomed!