Ivan Pribec

I forgot to add a question label. I didn't want to imply we should (or should not) follow the SciPy conventions. It was more a matter of where/how do we...

That's an argument against handling `**kwargs`, since it can be handled in the upstream SciPy interface with a little glue code. I guess it depends on what is the role...

Upon inspection `rwupdt` is only called in `lmstr`, which is the minimal storage driver for Levenberg-Marquadt. Given how abundant memory is today, I wonder how many users of this routine...

Thanks for the information. As long as the results have the same or better error tolerances, I don't think this should bother us. Is it okay (in terms of licenses)...

Looks like a problem related to nested procedures.

@arjenmarkus, personally, I'd prefer to keep the scope of MINPACK limited (at least initially) to non-linear least squares and systems of non-linear equation. While least squares is a form of...

After a bit of digging, I've found a few more interesting (nonlinear) least squares solvers that could be added to the comparison: * [RALFit](https://github.com/ralna/RALFit) is a recent Fortran code, algorithm...

One more NLS solver I could find is [`libdogleg`](https://github.com/dkogan/libdogleg) by @dkogan. An older version can be installed directly through the Debian/Ubuntu package manager (see https://tracker.debian.org/pkg/libdogleg) (not tested yet). The advantage...

Thanks for the reply. On the few examples I tested, libdogleg produced results matching other tools. When I try to upgrade with apt I don't get the latest version (>=0.15):...

QUADPACK was included in SLATEC and is thereby in public domain (at least in the USA). See [this](https://github.com/scipy/scipy/pull/7844#issuecomment-328317501) comment from the Scipy repository. The [website](https://nines.cs.kuleuven.be/software/QUADPACK/) from Leuven University also mentions...