08:02 < hacsheng> I would like to contribute to this organisation. And I have a great interests in Essential Deep Learning modules project idea especially Bidirectional Recurrent networks (BRN).

08:02 < hacsheng> I traced IRC and issue and I found RBM,SpikeSlabRBM and GAN was were done. Is there any the pending work about DBN,RBFN, and BRN.

08:03 < hacsheng> If not, Can I propose modules that I wish to implement?

12:13 < zoq> hacsheng: Hello, there is an open PR for RBM and GAN, so this is already taken up. However, you are free to propose anything you like to implement. The models listed on the ideas page are suggestions.

13:14 < hacsheng> zoq: Thanks!! I prefer to implement BRN. Hope to discuss with you after I organize my thoughts clearly in detail!

13:38 < zoq> hacsheng: Good plan, let's do this.

17:05 < ytka> hello wanna ask a question. I download mlpack-master and g++ recurrent_network_test.cpp. There are two main error below:

17:07 < ytka> 1.this file requires compiler and lib support for the ISO C++2011 standard . This support must be enable with the std=c++11 or -std=gnu++11 compiler options

17:09 < ytka> 2.In file included from /usr/include/mlpack/prereqs.hpp:110:0, from /usr/include/mlpack/core.hpp:238, from recurrent_network_test.cpp:12: /usr/include/mlpack/core/util/arma_config_check.hpp:19:27: fatal error: arma_config.hpp: No such file or directory compilation terminated.

17:11 < ytka> There is no arma_config_check.hpp in mlpack. (mlpack-master) But it seem to be necessary

17:25 < zoq> ytka: Hello, arma_config is generated in the build process (CMake). About the issues with recurrent_network_test.cpp, did you build mlpack? Once you build the code you could run the test with: bin/mlpack_test -t RecurrentNetworkTest. Or you can build against the lib and run your own code.

18:36 < ShikharJ> zoq: I was thinking whether mlpack should have a requirement for code coverage (codecov), on the incoming PRs. What do you think?

18:46 < zoq> ShikharJ: I'm not sure it should be a requirement, at least not a hard requirement. Writing a good and meaningful test should be the number one priority, however, it might be useful as an indicator, depending on how flexible it is, can we exclude something like: size_t parameter { return paraemeter}.

18:48 < ShikharJ> Sure, code coverage is anyways only intended as a guideline. Plus I find it very useful to view the generated reports to lookout for areas where the code remains untested.

19:06 < zoq> If you like we can test it out, there is already an coverage script https://github.com/mlpack/mlpack/blob/master/CMake/mlpack_coverage.in: cmake -DBUILD_WITH_COVERAGE

19:08 < sshkhrnwbie> This is regarding our discussion about implementing the variance scaling initializer (for further use in Xavier and He initializers). I went through the comments on #939 and your design guidelines and think it will be suitable to implement two classes VarianceScalingUniformInit and VarianceScalingGaussianInit as I agree that there will be little code overlap in them

19:09 < sshkhrnwbie> How do you suggest I incorporate the mode (fan-in / fan-out / fan-average) in the initializer ?

19:14 < zoq> sshkhrnwbie: An easy option would be to let the user provide the parameters at contruction time, another option would be to pass the constructed network itself.

19:14 < zoq> W ecould do both and start with the first option.

19:16 < sshkhrnwbie> ok I will try and get an implementation working

19:18 < zoq> sshkhrnwbie: great

19:52 < travis-ci> ShikharJ/mlpack#68 (Test - 4a15164 : Shikhar Jaiswal): The build has errored.

19:52 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/a6e44437ac61...4a15164bd1b4

19:52 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/340216922

22:01 < Leon23333> Hi! I've built mlpack on my windows10 but after building solutions, there are nearly 119 compile errors, is this normal?

22:02 < zoq> Errors or warnings?

22:03 < zoq> VS complains a lot about size_t, which is super strange.

22:04 < zoq> But there are usually no errors.

22:05 < Leon23333> errors and warnings

22:06 < Leon23333> the build result says 88 succeeded and 42 failed, 6 skipped

22:06 < zoq> hm, maybe you can post the output? But use pastebin.com or something like that.

22:11 < Leon23333> the output message is huge...

22:11 < zoq> Let's start with the first error.

22:14 < Leon23333> https://pastebin.com/vLPxTniZ

22:15 < Leon23333> the first several errors look like same, and all from python_copy

22:16 < zoq> Okay, can you rerun the configure step (CMake) with -DBUILD_PYTHON_BINDINGS=OFF.

22:17 < zoq> The python bindings are somewhat new, and not tested on windows.

22:18 < Leon23333> it was OFF

22:19 < Leon23333> Sorry,I'll try

22:20 < Leon23333> https://pastebin.com/u/Leon233333

22:20 < Leon23333> other errors

22:21 < zoq> Looks python related, which is strange if DBUILD_PYTHON_BINDINGS is set to off.

22:25 < Leon23333> yeah I set it to OFF and I'm rebuilding solution now

22:25 < zoq> Okay, let me know if you get the same errors.

22:26 < Leon23333> Sure

22:38 < Leon23333> no errors, lots of warnings

22:38 < Leon23333> thanks!!!!

22:38 < Leon23333> finally

22:39 < zoq> Nice!

22:40 < Leon23333> Does setting PYTHON to OFF mean that can not use mlpack on python?

22:41 < zoq> Yeah, we have to look into using python on windows.

22:41 < Leon23333> yeah sure

22:41 < Leon23333> thank u!

22:42 < zoq> Here to help :)

23:44 < travis-ci> ShikharJ/mlpack#69 (ResizeLayer - 90f2c68 : Shikhar Jaiswal): The build has errored.

23:44 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/3f203913db0c^...90f2c6824e90

23:44 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/340284639

23:50 < travis-ci> ShikharJ/mlpack#70 (ResizeLayer - 71d55d4 : Shikhar Jaiswal): The build has errored.

23:50 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/90f2c6824e90...71d55d42fab3

23:50 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/340285656