On 21 December 2010 00:58, Benjamin K. Stuhl <[email protected]> wrote:
> Here is a patch that optionally uses PyMinuit > (http://code.google.com/p/pyminuit/) to do fitting rather than the > current simple L-M routine. Minuit's great advantage is that it provides > high-quality nonlinear error estimation as well as the more standard > covariance matrix-based error estimates. > > The patch is probably a bit ugly right now, in that it might be worth > wrapping most of the code that adapts to Minuit up in a utility module, > but the larger reason this patch is an RFC is that PyMinuit is GPLv2 only. > Veusz is GPL v2 or later, so the licenses are entirely compatible -- but > using PyMinuit would mean that binary releases of Veusz would effectively > be GPLv2 only. Thanks for the patch. As long as it is optional, this should be fine. I think the minuit fitting should be in its own function however - I can have a look at that. In the longer term I think the fitting probably belongs in its own dialog box to make it easier to visualise the error bars, etc. It might also be good to have an option to use scipy for the fitting too, but I don't know how well that compares to pyminuit numerically (the license is less strict of course). Jeremy _______________________________________________ Veusz-discuss mailing list [email protected] https://mail.gna.org/listinfo/veusz-discuss
