01:13 < stephentu> naywhayare: oh its pretty cool

01:26 < stephentu> https://github.com/mlpack/mlpack/wiki/SummerOfCodeIdeas#low-ranksparse-optimization-using-frank-wolfe

19:42 < vlad_gl> naywhayare: hi! what do you think about this https://github.com/mlpack/mlpack/pull/410 ? Do you like this idea? If so, I need to rewrite some of tests. :)

19:47 < vlad_gl> I look at https://github.com/mlpack/mlpack/wiki/SummerOfCodeIdeas#low-ranksparse-optimization-using-frank-wolfe , and i like this idea :) want to propose myself as GSoC student with it..:)

20:02 < naywhayare> vlad_gl: hi there, I haven't had a chance to look into PR #410 yet

20:02 < naywhayare> also, stephentu would be the person to ask about the low-rank sparse optimization project :)

20:05 < vlad_gl> It's about use c++11 random instead of boost::random. In c++11 used random module from one of the boost versions, but this allows to be independent of the current boost version.

20:06 < vlad_gl> I just want to know, what do you think about this idea :)

20:07 < naywhayare> yeah, I think it's a good idea, I'll probably pull it in a day or two when I have time to go over it

20:08 < naywhayare> C++11 standard features are probably a better choice than Boost, when possible

20:12 < vlad_gl> ok. so, I need to fix some tests now :)

20:16 < stephentu> vlad_gl: ya the frank wolfe stuff is awesome

20:18 < stephentu> martin jaggi is pretty much a badass

20:18 < stephentu> that paper is pretty much his phd thesis stuffed into an icml paper

20:18 < stephentu> so its actually like 2 papers in 1

20:27 < vlad_gl> stephentu: great :) start to read it. then write a couple of words about this in mailing list. and introduce myself, heh :)

20:56 < library_finder> Hi

20:56 < library_finder> I want to know if mlpack has code to learn decision trees to do binary classification?

21:02 < zoq> library_fin: Hello, unfortunately mlpack only comes with an implementation for a decision stumps (single level decision trees).

21:03 < library_finder> I see. Thank you. Would you have any recommendations for C++ libraries that implement decision tree classifiers?

21:09 < stephentu> shogun might?

21:10 < zoq> hm, maybe scala works for you too? https://github.com/stripe/brushfire :)

21:10 < stephentu> brushfire is pretty cool

21:11 < stephentu> is there any reason we dont have an implementation?

21:11 < stephentu> i think there has been some interest before of implementing it

21:14 < library_finder> ooh. is it easy to call scala code from C++?

21:15 < zoq> stephentu: yeah, mainly lack of time ...

21:20 < zoq> library_fin: It isn't that simple scala generates primarily JVM bytecode.

21:23 < stephentu> library_finder: the other direction is much easier

21:23 < library_finder> i see, thanks. doesn't look like shogun has a DT classifier..

21:23 < stephentu> zoq: ya that seems to be the primary issue :)

21:23 < library_finder> has anyone used mlpack?

21:25 < library_finder> sorry, that doesn't have a decision tree classifier either..

21:25 < library_finder> the only one I've found so far is openCV, but that is a vision library..

21:27 < stephentu> library_finder: most people using DT classifiers in industry are probably just using sklearn

21:28 < stephentu> so the C++ support is probably lacking

21:28 < stephentu> evne if you find a library, it probably is buggy

21:28 < zoq> library_fin: If I remember right, Parijat (shogun) implemented some fundamental algorithms (e.g decision trees) last year during the sommer of code.

21:29 < library_finder> does sklearn have C++ bindings?