08:39 < deepanshu_> Hey! I wanted to start contributing to mlpack.

08:39 < deepanshu_> Would "Use pass-by-value instead of two overloads for copy and move #1021" be a good first bug?

09:54 < psycho-valker> hi i am facing issues while building ml pack on my ubuntu

09:55 < psycho-valker> psycho-valker psycho-build $ bin/mlpack_test

09:55 < psycho-valker> Running 765 test cases...

09:55 < psycho-valker> *** 1 failure is detected in the test module "mlpackTest"

09:56 < psycho-valker> on gmm_test.cpp line 234: in "GMMTest/GMMTrainEMSingleGaussianWithProbability": difference{0.0570898} between g.Component(0).Mean()[0]{0.47299670939187022} and 0.5{0.5} exceeds 5%

09:56 < psycho-valker> what should i do?

11:49 < rcurtin> deepanshu_: sure, that could work

11:49 < rcurtin> psycho-valker: that is probably a random test failure

11:50 < rcurtin> take a look through the issues for one like 'fix commonly failing tests' (I can't easily get a link right now, I am using a phone)

11:50 < rcurtin> are you using mlpack 2.2.5 or git master? I believe the tolerances for that test have been adjusted in the git master branch

11:56 < psycho-valker> ok thanks i am using mlpack 2.2.5

12:05 < rcurtin> cult-: you can manually set the initial transition matrix with this constructor:

12:05 < rcurtin> http://mlpack.org/docs/mlpack-2.2.5/doxygen/classmlpack_1_1hmm_1_1HMM.html#ac173a0c0e119519329525a892f8848e7

12:06 < rcurtin> then if you are training with unlabeled data (which I believe you are otherwise the initial transition matrix doesn't matter because it is directly estimated from labeled data),

12:06 < rcurtin> the training will start from the specified initial matrices

12:06 < rcurtin> so you could pass in the old-style 2.0.x initial matrices

12:06 < rcurtin> however like you said if there is no problem with 2.2.x maybe no need to do that :)

12:08 < sourabhvarshney1> @rcurtin Can I get the link to any pdf related to knowledge of all algorithms

12:13 < cult-> rcurtin: thanks

13:50 < zoq> sourabhvars: There is no single pdf for all methods, if you search for an overview I can recommend "Pattern Recognition and Machine Learning" by Christopher M. Bishop.