03:23 < metahost> rcurtin: Ah! I presumed IRC to be the primary channel because of the logs. :)

05:26 < tavishjain> Can someone guide me over some articles of using and developing ML and DL algos with C++ ?

05:26 < tavishjain> I have plenty experience of building models with python, but would love to learn it in C++ and contributing to MLPACK this summer

05:46 < hunting-party10[> Hi rcurtin

05:51 < hunting-party10[> Hi rcurtin zoq I was trying to fix the memory leak issue and i think i have fixed it , But the test which was designed calls a function `DeleteModules()` which is no longer needed since the destructor is fixed . So am i allowed to change the test itself or no ?

11:19 < zoq> hunting-paty10: Feel free to change it.

13:22 < hunting-party10[> Cool

13:53 < metahost> zoq: Hi Marcus I have updated ensmallen#149 to include the bounds as optimizer args itself as per our discussion. Please have a look once you can! :)

14:22 < zoq> metahost: Okay great, will take a look once I have a chance.

15:32 < rcurtin> metahost: yeah, we are still using the IRC logging system I set up years ago :)

15:32 < rcurtin> it might be possible to switch to Matrix's logging system now; I am not sure

15:32 < rcurtin> haven't looked into it :)

16:25 < metahost> zoq: Thanks!

16:27 < metahost> rcurtin: Ryan will switching to Matrix logging retain the historical logs too?

16:32 < metahost> (did we have matrix from the get go?)

16:34 < zoq> metahost: No, Ryan added that like 2 months ago.

16:36 < zoq> Haven't checked the matrix logging, does the matrix logging provide any features that we don't have with the IRC logs?

16:36 < zoq> Unless there is anything, I don't see the need to update that part.

16:37 < rcurtin> yeah, my opinion is that if someone opens a PR to mlpack.org, I'd be happy to merge it in

16:37 < rcurtin> maybe matrix.org provides better logging and search functionality

16:38 < TavishJainGitter> Can someone guide me over some articles of using and developing ML and DL algos with C++ ?

16:38 < TavishJainGitter> I have plenty experience of building models with python, but would love to learn it in C++ and contributing to MLPACK this summer

16:39 < zoq> TavishJain: There are some links to websites and books on the mlpack GSoC page.

16:40 < zoq> TavishJain: They are more or less about C++ concepts, not necessarily about ML.

16:40 < tavishjain> zoq I read some articles and documentations of Armadillo in the evening today (luckily C++ was the first programming language I learnt so understood the code)

16:41 < tavishjain> But as i was reading the mlpack codebase today to understand and start buiding things the main thing that concerned me was to understand which aprt of code does what

16:41 < zoq> TavishJain: Great, I heard some good things about Coursera.

16:42 < tavishjain> Like I dont have much experience in building C++ libraries, do you recommend reading about that first zoq?

16:45 < zoq> TavishJain: A general idea should be helpful. The way I would appreoach the code is to pick on method/algorithm that you are already familair with and see how it was implemented; PCA is a good example.

16:47 < tavishjain> Ohh thank you for that zoq. Looking forward for some good times ahead :-)

16:51 < hunting-party10[> Is there a way to run a specific subtest under a test eg . Run the `SimpleResidue` test under the `ANNLayer Test`

16:53 < rcurtin> hunting-party10[: yeah, try 'bin/mlpack_test -t ANNLayerTest/SimpleResidue'

16:53 < Nakul[m]> bin/mlpack_test -t yourtest suites

16:53 < rcurtin> exactly :)

16:55 < hunting-party10[> ah didnt know we had to use the `\` operator

16:55 < hunting-party10[> * ah didnt know we had to use the `/` operator

17:12 < rcurtin> yep, not too difficult. you can look at the boost unit test framework documentation for more options (or run bin/mlpack_test -help too)

17:15 < hunting-party10[> ah ic , Im new to the boost lib as well, will take some time to get used to things :)

17:21 < rcurtin> sounds good, I think the documentation for the parts of Boost that we use is pretty good

17:24 < hunting-party10[> Yes i was checking out the varient documentation seemed pretty simple

17:25 < hunting-party10[> <rcurtin "sounds good, I think the documen"> Templating is a bit confusing due to the large amount of templating types being used

18:39 < metahost> zoq: yeah keeping on using IRC logs as is makes sense. As for search getting a dump of the logs are easy enough. :)

18:58 < metahost> rcurtin: This issue (https://github.com/vector-im/riot-web/issues/2630) for riot.im mentions missing IRC style logs. :))