15:16 < gopala> Can density estimation trees handle multivariate distributions /

15:56 < rcurtin> gopala: a density estimation tree models a piecewise constant distribution, so I'm not sure entirely what is meant by your question

15:56 < rcurtin> if you mean whether or not a DET can model a multi-dimensional dataset, the answer is yes

15:57 < gopala> yes that was my question..

15:57 < rcurtin> ok, it sounds like I answered it ok then :)

15:57 < gopala> also it wasnt clear how i should input the predictee values and predictor features

15:57 < gopala> all in a csv ?

18:07 < win_dts> hi! So, i was wondering if mlpack applications/starter code can be tried successfully on windows system?

18:08 < win_dts> Also if one plans to contribute to mlpack, can they choose to do so by developing code on a windows system?

18:11 < zoq> win_dts: Hello, I'm not sure what is meant by applications/starter code, but if you mean whether or not you can build the mlpack code and tutorials on windows, yes that's possible. https://github.com/mlpack/mlpack/wiki/WindowsBuild might be helpful

18:13 < win_dts> zoq: I am checking out those articles. I think mlpack is targeted to unix systems. So, will I be able to write code for mlpack without much hiccups on windows?

18:13 < win_dts> Even in case of new updates?

18:19 < win_dts> ok, a noobier question: Can i compile mlpack code using cygwin compiler on windows system?

18:20 < win_dts> I mean how will a windows installation differ from simply using cygwin?

18:22 < zoq> Let me put it this way, I think most mlpack user uses some kind of Linux distribution, but we make sure it builds on windows. In fact, we use appveyor to check every commit on windows. I have never used Cygwin before to compile mlpack, but instead Visual Studio. But cygwin should work.

18:25 < zoq> You could also use the precompiled libraries: https://ci.appveyor.com/project/mlpack/mlpack/build/artifacts

18:27 < win_dts> ok, i need to ask a tonn of more questions, if you don't mid

18:27 < win_dts> *mind

18:27 < zoq> sure go ahead

18:28 < win_dts> ok so mlpack uses armadillo for super fast matrix and linear algebra calculations right

18:28 < zoq> right

18:29 < win_dts> now on searching about his i am hit with all these things: lapack, blas, openblas, intel mlk

18:29 < win_dts> and something about armadillio allwoing us to choose what to use

18:29 < win_dts> what is all this?

18:30 < win_dts> also a comment on (http://stackoverflow.com/questions/17639155/fast-lapack-blas-for-matrix-multiplication) "Note that high performance linear algebra is generally easier to accomplish on Linux and Mac OS X than on Windows." Why is it so?

18:34 < zoq> MKL, OpenBLAS etc. are all backends you can use to perform the math operations; e.g.: MKL is especially optimized for Intel CPU architecture.

18:35 < win_dts> ok, so lapack, mlk, blas are all libraries to solve various math problems?

18:35 < win_dts> is an optimized and efficient manner i suppose

18:36 < win_dts> And a wrapper to use these math solving code as c++ functions is made available via armadillo?

18:37 < zoq> yeah, as pointed out in the stackoverflow post, Intel MKL is faster on Windows, not sure that's true. I haven't done any research in that direction.

18:37 < win_dts> ok and wht about : "And a wrapper to use these math solving code as c++ functions is made available via armadillo?"

18:37 < zoq> armadillo does a lot more: http://arma.sourceforge.net/armadillo_joss_2016.pdf

18:37 < win_dts> ahh ok

18:38 < win_dts> it is so different once the vanilla c++ is taught and over and you now have to use professional open world software

18:39 < win_dts> college courses sure didnt prepare me for this lol

18:43 < zoq> You could also do e.g. matrix multiplication by using an array of arrays but isn't it much easier if you could use a library that provides an object to do matrix multiplication?

18:44 < win_dts> sure a library would make is much easier

18:44 < Wasp_> hey!

18:45 < zoq> Wasp_: Hello there!

18:46 < win_dts> now if instead of downloading lapack/blas , if I go for higher optimized packages like intel MKL / OpenBLAS, will Armadillo use the appropriate package?

18:46 < win_dts> Or should I be configuring it somewhere for it to do so...

18:46 < win_dts> ok never mind, i found the FAQ page about the same

18:49 < win_dts> So cmake links these lapack/blas to armadillo while building it?