01:42 < travis-ci> mlpack/ensmallen#715 (master - 29ff827 : Ryan Birmingham): The build passed.

01:42 < travis-ci> Change view : https://github.com/mlpack/ensmallen/compare/08994c52ee26...29ff82701937

01:42 < travis-ci> Build details : https://travis-ci.org/mlpack/ensmallen/builds/656809273

03:37 < kartikdutt18Gitt> Hi @zoq, Got my mlpack stickers. I absolutely love them.

03:37 < kartikdutt18Gitt> Thanks a lot.

05:11 < himanshu_pathak[> Hey rcurtin in https://ci.appveyor.com/project/mlpack/mlpack-wheels/builds/31128643/job/703b3or2uwmtwnxa can you tell what is the value of PYTHONPATH=

08:18 < jenkins-mlpack2> Project docker mlpack nightly build build #627: FAILURE in 3 hr 4 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/627/

08:34 < SaraanshTandonGi> What is the Deterministic value for a layer?

10:16 < robotcatorGitter> Hi, is anybody faces the STB_IMAGE_FOUND not works in Mac os? https://github.com/mlpack/mlpack/blob/master/CMakeLists.txt#L347

10:17 < robotcatorGitter> Only when I change the if statement into if (True), so when cmake will download the stb images dependency. https://github.com/mlpack/mlpack/blob/master/CMakeLists.txt#L347

10:20 < GauravSinghGitte> @saraansh1999 it determines whether the layer is in training or testing mode.

10:31 < TanviAgarwalGitt> i am not able to get armadillo on my system. hcan anybody tell alternative?

10:59 < SriramSKGitter[m> @tanviagwl98 : mlpack is built on armadillo and is a necessary component. What problems are you facing getting it on your system?

11:24 < TanviAgarwalGitt> I am not able to get a runnable form of armadillo @sriramsk1999 sir. Like I downloaded it but can't extract it.

11:28 < SriramSKGitter[m> No need to call me sir :) . What do you mean when you say can't extract it?

11:35 < TanviAgarwalGitt> can't extract is i am not getting zip file after download, the file after downloading is not supported on the system @sriramsk1999

11:43 < SriramSKGitter[m> Do you mean the `.tar.xz` file? If you are on Linux, `tar -xvf file.tar.xz` should do the trick. On Windows, I assume 7zip will be able to extract it.

11:49 < chopper_inbound[> Is there a concept of broadcasting in armadillo? (I need to operate on vectors and matrices)

12:56 < TanviAgarwalGitt> okay i will try again.

13:37 < rcurtin> himanshu_pathak[: for that step on line 9301 of the output, we do `SET PYTHONPATH=.`, which corresponds to build/src/mlpack/bindings/python/

13:37 < rcurtin> the problem I have with that build you can actually see just a few lines above: lmnn.pyx and cf.pyx are empty (they have size 0)

13:38 < rcurtin> this means that when the programs generate_pyx_lmnn and generate_pyx_cf were run, they failed to run properly

13:38 < rcurtin> on Windows this generally happens when the right .dlls can't be found

13:38 < rcurtin> and I know that's what's happening here because I used remote desktop to connect to the build system and see

13:38 < rcurtin> so my first step (which I've been trying to do for many days now...) is just to get those files to be non-empty

14:04 < AbishaiEbenezerG> hi mlpack! I've been learning neural networks for the past few weeks, and still have a lot to do. But i do feel that this is the right time to start contributing, atleast in a small way. I'm having some difficulty in finding some issues i can work on. Some guidance here would be really helpful. Thanks!

14:13 < AbishaiEbenezerG> I know i could open my own issue and claim it, but if anyone has some more guidance , that would be great

14:17 < himanshu_pathak[> rcurtin (@freenode_rcurtin:matrix.org): Also the problem is only with 32 bit not in 64 bit version I will also try to find out if I get something I will notify you

14:30 < kartikdutt18Gitt> Hi @abishaiema, Feel free to do so. Also there are some good first issues that you might want to look at.

14:31 < kartikdutt18Gitt> *you might want to take a look at.

14:31 < AbishaiEbenezerG> yes @kartikdutt18 , i did take a look at them.

14:31 < AbishaiEbenezerG> but most(if not all) have already been claimed...

14:32 < AbishaiEbenezerG> wanted to know how i need to approach this as i'm very new to this

14:32 < kartikdutt18Gitt> Hmm, Well if you find something interesting feel free to open a PR for the same.

14:33 < AbishaiEbenezerG> sure. Thanks tho !!!

14:36 < kartikdutt18Gitt> I think you could get familiar with the codebase and if you find something that you might need to use or something that's interesting you can open a PR for the same or if you need help with anything you can open an issue / discuss it here.

14:38 < AbishaiEbenezerG> i agree. I think i should just spend time with the codebase now and then take it from there....

15:00 < rcurtin> himanshu_pathak[: right, this has confused me from the outset---why are the libraries in the right place fpr 64 bit (even without any of the copying in .appveyor.yml)? I haven't figured that out

15:40 < rcurtin> mlpack has just been merged into the Julia package registry

15:40 < rcurtin> so now you can do

15:40 < rcurtin> julia> import Pkg; Pkg.install("mlpack")

15:40 < rcurtin> and then you can use the Julia mlpack bindings :)

15:41 < rcurtin> this unblocks the 3.3.0 release (which was waiting on this), so now we can finish up the other issues that are open and get the release done soon :)

15:43 < pickle-rick[m]> Awesome!!

16:05 < himanshu_pathak[> Yeah!!

16:08 < chopper_inbound[> wow

16:11 < kartikdutt18Gitt> Great.

16:43 < SaraanshTandonGi> What is layer -> Delta()

16:59 < SaraanshTandonGi> I think it is the upstream gradient. Is it? And where is the code that sets it?

17:34 < SaraanshTandonGi> same for layer -> Loss(), I see it being returned by the LossVisitor, but where is its value set?

17:37 < PrinceGuptaGitte> Hi @saraansh1999 Delta is the gradient matrix of that gradient, and loss is being set when we call `Backward()` function and pass in the loss of that class through visitor, and since backward function updates it.

17:37 < PrinceGuptaGitte> Of that layer*

17:37 < PrinceGuptaGitte> It gets updated

17:38 < PrinceGuptaGitte> That's the same way how `outputParameter` gets updated.

17:38 < PrinceGuptaGitte> In the forward function

17:49 < SaraanshTandonGi> yeah. Just saw it in the code. Thanks :)

18:01 < Pranshu54> Hi KhizirSiddiquiGi ,after referring SummerofCodeIdeas page I would like to know if we can add a module for "cyclic learning" in code base under "Improvisation and Implementation of ANN Modules". Cyclic Learning is used for faster training of neural networks.

18:10 < SaraanshTandonGi> @prince776 I see forward setting layer -> outputParameter(), and backward setting layer -> Delta().

18:10 < SaraanshTandonGi> Where is layer -> Loss() being set().

18:10 < SaraanshTandonGi> (edited) ... being set(). => ... being set()?

18:29 < PrinceGuptaGitte> You see backward() function of any layer class takes 3 parameter:

18:29 < PrinceGuptaGitte> In our activation, error, Delta

18:29 < PrinceGuptaGitte> Input*

18:30 < PrinceGuptaGitte> Error is the Delta of previous layer(while propagating in backward direction)

18:35 < PrinceGuptaGitte> And Loss() as far as I remember is only for certain layers, not every layer

18:35 < PrinceGuptaGitte> I might have went offtrack with explaination, but I wanted to clear it since it's really confusing sometimes

18:37 < SaraanshTandonGi> @prince776 Thanks. What you explained above is what i just took like 2hours to understand on my own. : P

18:47 < PrinceGuptaGitte> I also struggled understanding it, since I was totally new to boost and template programming

18:47 < PrinceGuptaGitte> But that's so beautifully done, I thought for this they definitely need inheritance but no, everything was done avoiding viable

18:50 < SaraanshTandonGi> yeah!

18:55 < SaraanshTandonGi> What could be causing this error in the Train function of ffn.

18:56 < SaraanshTandonGi> (edited) ... of ffn. => ... of ffn when I increase the MAX_ITERATIONS_PER_CYCLE value.

18:56 < SaraanshTandonGi> batch size = 50

18:56 < SaraanshTandonGi> train set size = 37800

19:10 < SaraanshTandonGi> MAX_ITERATIONS_PER_CYCLE = 40 giving this error

19:13 < SaraanshTandonGi> MAX_ITERATIONS_PER_CYCLE = 20 is not

19:32 < SaraanshTandonGi> Everything above MAX_ITERATIONS_PER_CYCLE=23 are failing.

19:33 < SaraanshTandonGi> (edited) ... MAX_ITERATIONS_PER_CYCLE=23 are failing. => ... MAX_ITERATIONS_PER_CYCLE=23 is failing.

19:41 < PrinceGuptaGitte> Vtables*, I meant. I just noticed typo

19:41 < PrinceGuptaGitte> Auto correct

19:43 < SaraanshTandonGi> @prince776 do you have any idea regarding this?

19:43 < SaraanshTandonGi> (edited) ... regarding this? => ... regarding this? ^^^

19:45 < SaraanshTandonGi> Also @prince776 if you have not already put in a lot of work I would like to work on TiripletMarginLoss

20:11 < PrinceGuptaGitte> Not really, can you share your code I might get an idea

20:14 < SaraanshTandonGi> will do

20:15 < SaraanshTandonGi> https://drive.google.com/drive/folders/1bHwfqIQcidgX8sOiwoMFdQWkWA9G4DnB?usp=sharing

20:20 < PrinceGuptaGitte> What is the batch size and number if columns kn your dataset?

20:21 < SaraanshTandonGi> > batch size = 50

20:21 < SaraanshTandonGi> > train set size = 37800

20:21 < SaraanshTandonGi> >

20:22 < SaraanshTandonGi> mnist data, so 784 cols

20:22 < SaraanshTandonGi> rows in mlpack

20:23 < PrinceGuptaGitte> I'll try to check it tomorrow. I have a maths exam tomorrow. Sorry

20:26 < SaraanshTandonGi> OK :)

20:26 < SaraanshTandonGi> Best of luck!

20:27 < SaraanshTandonGi> @zoq

20:27 < SaraanshTandonGi> @rcurtin

20:27 < SaraanshTandonGi> can you take a look at #2248 and #2249

20:29 < PrinceGuptaGitte> Thanks

20:51 < metahost> rcurtin: Ryan, I have added the documentation separately in doc/functions_types.md and made the other changes you suggested, please have a look when you can! :)

22:46 < travis-ci> shrit/models#8 (digit - 0ada054 : Omar Shrit): The build is still failing.

22:46 < travis-ci> Change view : https://github.com/shrit/models/compare/7cb1366e82d7...0ada054a4ef1

22:46 < travis-ci> Build details : https://travis-ci.com/shrit/models/builds/151274370