08:46 < ShikharJ> zoq: Are you there?

09:43 < ShikharJ> zoq: Can we merge the BatchSupport and DCGAN PRs now? Both of them are complete, and once they are in, we can start testing the code with the GANOptimizer class as well.

10:23 < jenkins-mlpack> Project docker mlpack nightly build build #354: STILL UNSTABLE in 3 hr 9 min: http://masterblaster.mlpack.org/job/docker%20mlpack%20nightly%20build/354/

10:39 < zoq> ShikharJ: Unfortunately we have to wait: https://github.com/mlpack/mlpack/blob/master/CONTRIBUTING.md#pull-request-process, but what about working locally we can still open a PR, the build will fail since not all files are there but we can still discuss the code.

11:48 < Atharva> sumedhghaisas: You there?

11:48 < Atharva> zoq: I had a doubt.

11:49 < Atharva> The gradient check is failing if we use mean squared error instead of negative log likelihood, I tried to figure out why but couldn't.

11:55 < manish7294> zoq: There's a matlab implementation of LMNN, but it's available as a part of separate toolkit, so can you help in incuding that?

11:58 < zoq> Atharva: Ahh, this is because https://github.com/mlpack/mlpack/blob/master/src/mlpack/methods/ann/loss_functions/mean_squared_error_impl.hpp#L42 doesn't match with the Forward pass, I'll see to open a PR later today to fix that.

11:58 < zoq> manish7294: Sure, I guess all we have to do is to add it to the search path.

11:58 < manish7294> zoq: This is the toolkit https://lvdmaaten.github.io/drtoolbox/

12:06 < zoq> manish7294: So basically all we have to do is to add the lib to https://github.com/mlpack/benchmarks/blob/master/libraries/package-urls.txt

12:06 < zoq> manish7294: and write a simple install script that add's the files to the matlab path: addpath(genpath(‘installation_folder/drtoolbox’)) in the form of: https://github.com/mlpack/benchmarks/blob/master/libraries/mlpy_install.sh

12:07 < manish7294> zoq: Thanks I will try that.

12:54 < Atharva> zoq: Yes, I made the changes in the Backward function and it works now.

12:56 < zoq> Atharva: Nice, do you think you can open a PR with the modifications? I guess you could also use the existing PR. Perhaps, I'll have to implement another one, if I remember right the RNN network doesn't work so well with the loss function.

12:59 < Atharva> zoq: Sure, I can commit it in the sampling PR if that's fine. Or are you going to implement mean square error again?

12:59 < zoq> Atharva: the sampling PR is jsut fine, thanks!

13:00 < Atharva> zoq: Okay then.

13:39 < manish7294> zoq: I tried figuring out more details and what I found is we can use toolbox by just adding path to toolbox at start of the matlab. So, we can do that by calling a subprocess at starting of matlab's method script. Does this sound reasonable? We can have a drtoolbox_inststall.sh to just check if matlab & drtoolbox exists.

13:40 < manish7294> zoq: And is it possible to have matlab installed on slake as I guess matlab_install.sh just check for existence of matlab on system.

13:42 < zoq> manish7294: hm, I guess all the install script has to do is to unpack the archive. And not sure but I think adding the libraries folder to https://github.com/mlpack/benchmarks/blob/master/Makefile#L77 should also work: https://de.mathworks.com/help/matlab/matlab_env/add-folders-to-matlab-search-path-at-startup.html

13:42 < zoq> manish7294: matlab should be available on slake

13:43 < manish7294> zoq: It doesn't look like that slake having matlab

13:44 < manish7294> zoq: Right, that seems possible.

13:44 < zoq> manish7294: It's in '/opt/matlab/bin'

13:47 < zoq> manish7294: You could use SSH X-Forwarding to open the GUI. 'ssh -X server' and afterwards you can start the application.

13:47 < manish7294> zoq: Thanks! found it

13:59 < manish7294> rcurtin: Look like we found it https://github.com/shogun-toolbox/shogun/blob/f55d13fe3aa1ba2efcb715a9df91f4739bbf6237/src/shogun/multiclass/KNN.h#L102

14:38 < rcurtin> manish7294: that looks perfect if we can call that from Python

14:48 < manish7294> rcurtin: It looks like we are able to make calls to nearest_neighbors(), I tried printing them :)

14:49 < manish7294> and as expected, first column is same as index

14:57 < rcurtin> ok, that sounds just fine, we can use that then

16:39 < Atharva> sumedhghaisas: rcurtin: I saw the comments on the PR, will make changes soon

16:40 < Atharva> sumedhghaisas: You there?

16:53 < manish7294> rcurtin: I made the changes for accuracy calculator. Hopefully, it's working fine https://github.com/mlpack/benchmarks/pull/123/commits/fa381795a71afef2b4812266353310e957ceb00d

16:59 < manish7294> zoq: It seems letters and balance datasets still have labels in first column. Please check them out.

17:28 < rcurtin> manish7294: if the datasets have a problem, you can supply a fixed version and I can drop them into place on the server

17:29 < manish7294> rcurtin: sounds good :)

17:32 < rcurtin> just to let you know that I can help too, that's all I mean. :)

17:38 < manish7294> rcurtin: no worries, I have it ready --- just sending you a link :)

17:44 < manish7294> rcurtin: Here are the links: https://www.pastefile.com/4o6xkf & https://www.pastefile.com/4nbBRg

18:06 < rcurtin> manish7294: thanks, dropped them into place on mlpack.org/datasets/