ibayer

@ZeerakW Thanks for raising the issue. Has fastFM been working with 3.7 before or is a more recent cython need to compile fastFM for 3.7?

@ZeerakW > if it is in core then let me know and I can submit a PR later today. It's not, core contains only the C code. I'm just still...

@ZeerakW Ah, okay I think I understand the situation now. You only changed the Cython version to 0.29 to make source install work. Since we don't include the Cython generated...

@adrtod I'm currently completely rewriting and extending the solvers in c++ (currently C)https://github.com/ibayer/fastFM/pull/94#. The new version will be much easier to extend. The rewrite will include 3'rd FMs with the...

I don't like to change the API except for a very good reason, especial if this moves us further away for the sklearn API. I'm not sure if `easier to...

I just revisit this PR once again but still I can't make up my mind if replacing `fit_predict` by `fit_transform` is the right way to go. It improves consistency (I...

I don't see a difference, but please check that the `init_stdev` parameter is the same. Please have a look at my second comment above, this could explain your observation for...

@arogozhnikov Great comparison, I have a few suggestions that could make the evaluation even more useful for other people. 1. Provide the exact version of the software that you are...

@arogozhnikov It's possible to use a random seed with libFM. `"seed", "integer value, default=None"` https://github.com/srendle/libfm/blob/master/src/libfm/libfm.cpp#L93

@nicolasfguillaume Windows is not actively supported and I don't know if someone managed to compile fastFM on windows before. I would try to compile with g++ first. Feel free to...