Hi everyone. I am pretty psyched that I got so much feedback :) I think planning for a road towards 1.0 and also beyond would be good.
Thanks Gilles for sharing your insights. They match quit well what I had in mind :) Also, I think getting an idea about the userbase would be great - though it is a bit hard. We *could* put a survey on the website. Not sure if there is any other method? Also, I totally agree with Vlad: we must focus on user interests and also on our own qualifications. I'm not entirely happy with the state of HMM, GP and DPGMM/VBGMM currently. So there are two goals I heard a lot: sparse matrix support everywhere and Python 3 support. These sound also pretty good to me. As I think grid search an model evaluation is one of the core features of sklearn, I have been working on that quite a bit lately and totally agree with Olivier that there is still some worthwhile work to do. Also the "write your own estimator" docs and general sklearn API should be a pretty high priority, I think. As I see one of the applications of sklearn in teaching, I think it would be great if we could get all the "ML 101" algorithms together. We might disagree somewhat on what those are, but I think we can find some common ground. Neural networks, and general ensembles are the first that come to my mind. And also logistic regression. I think it is a bit weird that we don't have multinomial logistic regression in sklearn. Ok so what do we do now? I think we should decide if and how we want to ask our users. Then we should find a way to set up a roadmap. I am not sure if the github issue tracker is the right way to do it. Should we instead use the github wiki? Best, Andy ------------------------------------------------------------------------------ Master HTML5, CSS3, ASP.NET, MVC, AJAX, Knockout.js, Web API and much more. Get web development skills now with LearnDevNow - 350+ hours of step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122812 _______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
