08:37 < travis-ci> manish7294/mlpack#8 (lmnn - 4851b6d : Manish): The build has errored.

08:37 < travis-ci> Change view : https://github.com/manish7294/mlpack/compare/bd8e65d22874...4851b6d6bac6

08:37 < travis-ci> Build details : https://travis-ci.com/manish7294/mlpack/builds/74196903

09:45 < travis-ci> manish7294/mlpack#9 (lmnn - f61e947 : Manish): The build passed.

09:45 < travis-ci> Change view : https://github.com/manish7294/mlpack/compare/4851b6d6bac6...f61e947dfae0

09:45 < travis-ci> Build details : https://travis-ci.com/manish7294/mlpack/builds/74198275

09:48 < jenkins-mlpack> Yippee, build fixed!

09:48 < jenkins-mlpack> Project docker mlpack nightly build build #327: FIXED in 2 hr 34 min: http://masterblaster.mlpack.org/job/docker%20mlpack%20nightly%20build/327/

21:20 < gman> Hi there! I have been using mlpack KFoldCV and SimpleCV for a while and after reading the tutorials, going through the API and studying the code I still don't know for sure if CV is randomizing the dataset split or not...

21:20 < gman> ...somebody has a clue on that?

21:51 < rcurtin> gman: hang on, let me double check

21:53 < gman> sure

21:55 < rcurtin> looking at kfold_cv_impl.hpp:300, I am pretty sure that the CV classes don't randomize the dataset split

21:56 < rcurtin> if you wanted to do that, you could call ShuffleData() from mlpack/core/math/shuffle_data.hpp beforehand

21:56 < gman> uhmm.. ok thanks for the clarification

21:56 < rcurtin> the CV code is pretty complex since it is all template metaprogramming that has to work with lots of different algorithms

21:56 < gman> yep I noticed that

21:56 < gman> I think the advantage of KFoldCV would be to actually randomize the dataset

21:57 < gman> otherwise... I have seen issues when calculating some metrics such as precision/recall

21:57 < rcurtin> I think you're completely right and that would be a nice constructor argument to add to KFoldCV

21:57 < gman> good... I will try to work on something like that and let you know if it works

21:57 < rcurtin> implementationally speaking, it would be hard to change the internals of the class itself and how it extracts the dataset, but easy to add an if(shuffle) ShuffleData(...) to the constructor

21:58 < rcurtin> sure, if you'd like to do that feel free, otherwise we can open a Github issue and I'll get to it when I have a chance (maybe tomorrow? should be easy)

21:58 < gman> cool, as you wish - I can't promise I will have it working so soon :)

21:58 < rcurtin> ok---let me go ahead and open an issue, and we can track its progress there

21:59 < gman> arma:shuffle is giving me many issues - so will try to understand why's that first

21:59 < rcurtin> yeah, arma::shuffle() will shuffle all elements, not just columns (if I am remembering right)

21:59 < rcurtin> take a look at the shuffle_data.hpp file and maybe you can use either the ShuffleData() function or see what the internal implementation of that function is and it could help

22:00 < gman> will do, thanks for the help!

22:00 < rcurtin> sure, glad that you are finding the CV code useful!

22:00 < rcurtin> if you have any issues in the future don't hesitate to ask

22:00 < gman> fantastic, thanks

22:00 < rcurtin> no problem, happy to help :)

22:02 < rcurtin> https://github.com/mlpack/mlpack/issues/1409

22:34 < zoq> ShikharJ: Do you need any help with the debug process, we can work on that part together? Starting with a really small sample is a good idea.

23:39 < travis-ci> mlpack/mlpack#4951 (master - 949ab83 : Ryan Curtin): The build has errored.

23:39 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/0853affad63d...949ab83ef5a3

23:39 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/382917286