[SciPy-User] Least-squares fittings with bounds: why is scipy not up to the task?

Eric Emsellem eemselle@eso....
Thu Mar 8 17:12:17 CST 2012

Dear Gael

thanks for the feedback.

Well yes, I thought of the fact that an email (with all the drawbacks of 
such a medium) may not be the right way, and that my message may be 
misinterpreted. I obviously didn't mean to be offensive here. I am one 
of the biggest fan of python et al., open source in general etc, and I 
try to contribute whenever I can with my very limited expertise (I 
recently opened a github account and am still fighting to finalised my 
badly-written package to release it in case it could be useful to anyone).

I believe it is fine to react as you did, because it does at least show 
that people care (seeing the bright side).  I apologize: if I offended 
you, I probably offended others. This was definitely not my initial 
intent. And I was, for sure, not condescending, by very very far, as 
again I am impressed by what has been achieved (I did use python at a 
time when most modules were very buggy or hard to handle, and you had to 
tweak your system to make it work at all, not talking about early 
version of e.g. linux distribs).

I am  putting energy (at my own, very modest, level) to have more users 
of e.g., python. This means organising tutorials and providing advice 
when relevant. I try to do that as much as I can.

The question about a leastsq I relayed is often what I hear from 
non-experts when looking at what is available. I always first try to 
look for an easy solution, to convince them that something like that 
exists, or that you can solve the problem anyway. In the specific case I 
describe, I didn't find a way out (besides using the very useful 
implementation of mpfit!). I do believe that, considering the amazing 
amount of extensive development in scipy and related packages, such a 
routine should be available and directly linked with the main, big, 
packages. This was basically my two cents (probably badly toned) to 
trigger a reaction. I did for sure get one, but not the one I expected. 
So once more: apologies to all I may have offended.

Let's see where this goes now. I see that there is a lmfit package 
(thanks Matt for the input!). I'll have a look at this asap and try to 
test it on my own minimisation problem. I do think that having a 
well-tested module integrated within scipy would be a big plus. Apart 
from testing these on my specific problems, I cannot offer much 
considering my limited expertise (in programming and maths).


On 03/08/2012 10:07 PM, Gael Varoquaux wrote:
> I am sorry I am going to react to the provocation.
> As some one who spends a fair amount of time working on open source
> software I hear such remarks quite often: 'why is feature foo not
> implemented in package bar?. I am finding it harder and harder not to
> react negatively to these emails. Now I cannot consider myself as a
> contributor to scipy, and thus I can claim that I am not taking your
> comment personally.

More information about the SciPy-User mailing list