00:01 < kris2> or maybe rcurtin that would be great

08:24 < nadarb> Hello everyone

08:25 < nadarb> I am new to this chatroom, can anyone explain to me how it works?

08:25 < nadarb> Thanks :D

13:30 < vss> hi ! :p

13:41 < sahilc> I realized that there isn't a implementation for any linear program solvers in mlpack

13:42 < sahilc> we do have a LDSDP implementation

13:42 < sahilc> would it make sense to include a linear program solver too?

13:49 < vss> Trion : Submitted your proposal ?

13:51 < Trion> yep, but just a draft :-) waiting for feedback

13:52 < vss> you sent it to your mentor for feedback ?

13:53 < Trion> Nope, on GSoC

13:53 < vss> cool

13:54 < vss> where are you from ? XD

13:55 < Trion> India :D

13:55 < vss> which part ?

13:55 < Trion> Punjab

13:55 < vss> damn

13:55 < vss> i live in chennai XD

13:57 < Trion> Wow, lots of distance

13:58 < vss> i have a bunch of friends who are from chandigarh

13:58 < vss> you're doing ur btech now ?

13:59 < Trion> yep

14:18 < travis-ci> mlpack/mlpack#2139 (master - 54ac930 : Ryan Curtin): The build passed.

14:18 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/81dc1e5ee824...54ac930b26e0

14:18 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/213831270

15:42 < Trion> What is the double that FFN::Evaluate returns? I thought it will give the model output instead i.e. arma::mat(elements, 1)

15:46 < zoq> Trion: It returns the error, this function is basically only used for training since the existing optimizer expect an implementation of Evaluate and Gradient, what you like to use is Predict

15:48 < Trion> I am trying to evaluate the current frame from game to get a vector of possible reward for all actions, select the one with maximum reward, get the actual reward for that action from environment and optimize the accuracy loss with SGD.

15:55 < Trion> Oh! I understand now! :P Train and Predict prototypes looked similar so I was confused.

15:59 < Trion> so the responses in Train prototype are the expected outputs for inputs and the responses in Predict is the output that we get after evaluation... shouldn't it be better renamed as output?

16:02 < zoq> Maybe you're right, I used the same name scheme as for logistic regression or nbc.

16:02 < Trion> :P Can it be my first mlpack PR?

16:03 < vss> lol XD

16:04 < zoq> Trion: I've checked nbc and it it uses results, do you think that would also work?

16:05 < Trion> yes that is good too

16:06 < zoq> okay, great

16:38 < Trion> Is there any make shortcut for ann method? I tabbed all possibilities and nothing seemed referring to ann

16:41 < zoq> Trion: The shortcut you see are for executables, since the ann code has no executable there is none. But if you run make it should build modified files only.

16:42 < Trion> :P I had 'make clean'-ed myself

17:16 < rcurtin> zoq: I see the mlpack-jenkins account, when is it supposed to send the 'can one of the admins please verify the patch email'? if the build fails?

17:34 < zoq> rcurtin: Ah, I see actually only if someone opens a new PR that isn't whitelisted, but since the PR was already there it adds the comment.

17:35 < rcurtin> a whitelisted PR would be one that passes whatever mlpack-jenkins' checks are?

17:45 < zoq> rcurtin: A PR is whitelisted if it's from one of the users mentioned in the configuration. So, if someone open's a PR, Jenkins get's a message (webhook) and will comment 'Can one of the admins verify this patch?' that's the default message on the PR, I guess we could remove that if you like.

17:45 < zoq> From that point nothing will happen until some whitelisted member writes a message "@mlpack-jenkins test this please", that message triggers the webhook to run the jobs associated with the Pull requests.

17:45 < zoq> So basically nothing has changed, just that we can tell jenkins to run additional tests if we think they would provide some useful information.

18:04 < rcurtin> ah, ok, right; I dunno if the "can one of the admins verify this patch" message is necessary, at least for me---I already get emails for every PR that is opened

18:04 < rcurtin> but I am not sure if that is the case for everyone

18:07 < zoq> Yeah, I guess it might be a good indicator for someone who opens a PR, that we get a notification, but I'm not certain that's true. If we let the message in, it might be a good idea to filter that mail out.

18:12 < rcurtin> yeah, fair enough

18:38 < travis-ci> mlpack/mlpack#2143 (master - 8b55c58 : Ryan Curtin): The build passed.

18:38 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/54ac930b26e0...8b55c58a72b9

18:38 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/213944410