11:36 < SinisterMJ> Cheers, the latest SVN Version has quite some issues with Visual Studio, as in template errors and Unresolved Linker Symbols

11:39 < SinisterMJ> ostream_extra.hpp causes error C4519

12:17 < SinisterMJ> error LNK2001: unresolved external symbol "public: static class mlpack::util::NullOutStream mlpack::Log::Debug" (?Debug@Log@mlpack@@2VNullOutStream@util@2@A)

12:31 < jenkins-mlpack> Project mlpack - nightly matrix build build #631: STILL FAILING in 7 hr 0 min: http://big.cc.gt.atl.ga.us:8080/job/mlpack%20-%20nightly%20matrix%20build/631/

15:08 < naywhayare> SinisterMJ: are you using VS2013 or VS2012?

15:08 < naywhayare> C++11 support in VS2012 isn't good enough for mlpack

15:10 < SinisterMJ> 2012

15:11 < SinisterMJ> I've fallen back to an older version now from SVN, which works fine

15:11 < naywhayare> which version from svn? probably before we switched to C++11 support?

15:12 < naywhayare> (I can't remember which revision that was... I think it was about two months ago)

15:12 < naywhayare> I'm going to get VS2013 installed today and test trunk against it, but it'll take an hour or two...

15:12 < SinisterMJ> 17228 atm

15:44 < naywhayare> SinisterMJ: ah, yeah, that's before C++11 was required

15:45 < SinisterMJ> I am currently having an issue with the CSV not loading properly. I remember there was some issue, but can't find the appropriate code

16:10 < naywhayare> SinisterMJ: can you tell me more? I can probably help

16:25 < SinisterMJ> Found it

16:26 < SinisterMJ> Is there a way to compile the framework via Linux, but have it result in .libs usable on VS2012? Or do I need to update my VS?

16:29 < naywhayare> I don't think that's possible... or, at least, I have no idea how to get a Linux compiler to produce working Windows binaries

16:29 < SinisterMJ> okay, so VS2013 it is

16:29 < naywhayare> yeah; sorry :( VS2012 doesn't have good C++11 support

16:30 < naywhayare> I'm installing VS2013 on a system right now so that I can hopefully fix potential issues...

16:30 < naywhayare> our Windows build slaves haven't worked in a long time. they're annoying and time-consuming to keep running correctly :)

16:30 < naywhayare> (although... given the state of the Linux builds too, maybe that is true of all build slaves...)

16:31 < SinisterMJ> lol

16:47 < naywhayare> this VS install takes forever... I don't know how much debugging I'll manage to do before I have to leave today

16:47 < naywhayare> so it may be Monday before I'm able to work out some of the issues I'm sure I'll find

16:48 < naywhayare> and I won't be available over the weekend unfortunately

16:55 < zoq> naywhayare: Btw. I fixed the build on OS X 10.10 by reinstalling the os x command line tools.

16:56 < naywhayare> zoq: that's better than a RAM failure or something... :)

19:43 < george___> Hello, has anyone used mlpack to perform KNN search on the torus? I tried to do this, but had some troubles. I perhaps erroneously thought that it would be sufficient to just write a torus metric.

19:46 < george___> But this did not quite work for me as planned, I think the binary space partitioning does not make a good tree if I am really thinking that I am on the torus.

19:58 < naywhayare> george___: try with the cover tree; that should work better with arbitrary metrics

19:59 < naywhayare> but I would think that BinarySpaceTree would at least give correct results, even if it wasn't fast

19:59 < naywhayare> can you tell me more about the issues that you had?

19:59 < george___> It is always possible that I did something wrong

20:00 < george___> I wrote a metric class which computed the L2 metric on the 2D torus

20:01 < naywhayare> so, if I understand right, you mean the L2 distance on the surface of a 3D torus? or maybe something else?

20:01 < george___> yes

20:01 < naywhayare> I'm not sure that that distance satisfies the triangle inequality

20:01 < george___> so S^1 X S^1

20:02 < naywhayare> yeah, I'm pretty sure it doesn't satisfy the triangle inequality, so it's not actually a metric

20:03 < george___> ok hold on, let me check

20:03 < naywhayare> I actually have to run for now (maybe be back later tonight? not sure) but I'll think about this

20:04 < george___> thanks, I'll examine this more, help is much appreciated

20:04 < naywhayare> if it isn't a metric, and there's no equivalent distance that can be shown to satisfy the triangle inequality, then it may be difficult or not possible to do this quickly with trees

20:04 < george___> right

20:04 < naywhayare> periodic coordinate systems are a thought, though; there was once support for this in mlpack but it wasn't maintained and didn't work (so... I guess there wasn't actually support, then)

20:05 < naywhayare> anyway, because it didn't work it got removed. but it could probably be resurrected in some form (though it would need some further thinking)

20:10 < george___> I'll think about this more, its been a while since I studied analysis but I was pretty sure that L2 satisfied the triangle inequality on the torus

20:31 < george___> In fact, I am sure that this satisfies the triangle inequality.

21:47 < george___> i.e. the torus is a metric space (with triangle inequality) because it is the product of the circle with itself, and since the circle is a metric space the induced metric on the torus satisfies the triangle inequality