[SciPy-User] Alternatives to scipy.optimize

[Erik Petigura]

Thanks for all the suggestions!  The discussion was very enlightening.

I wound up writing the following wrappers:

https://gist.github.com/1927518

I think Matt has the right idea with lmfit:

> This approach gives a user a lot of flexibility in setting up fitting
> models, and can allow the fitting function to remain relatively static
> and written in terms of the "physical" model.  Of course, it is not
> going to be as fast as numexpr for array calculations, but it is more
> general, and faster ufuncs is not the issue at hand.


In the end, I decided to fit my linear parameters (polynomial trend) separately from my box parameters, by just excluding the region with the box.

Erik

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20120227/83857a57/attachment.html>