07:58 < nax_> Success, Works: https://www.nuget.org/packages/armadillo-code/7.800.2

13:24 < sgondala> I want to use mlpack in windows for office work, but for some support reasons, office environment is stuck with Visual Studio 2013. In order to use mlpack,I built mlpack using VS2015 on my personal machine (Followed this - keon.io/mlpack/mlpack-on-windows/) Using these libraries, I am planning to do this - https://github.com/mlpack/mlpack/wiki/WindowsBuild It didn't work

13:27 < sgondala> Can i create .dll files from headers (I can create .dlls from VS2015) and use that instead? I am not sure how to do this. Can some one guide me on this?

13:34 < naxalpha> zoq: I need a little help regarding nuget package

13:39 < kris1> zoq: I have updated both NAG and Xavier Init PR please have a look

14:38 < vivekp> Hi, I just submitted my draft proposal. Looking forward to helpful feeback to improve it further. I know I

14:41 < vivekp> Weirdly message got cut off, I guess.

14:43 < vivekp> I was just saying that I know I'm late by a few days now but unfortunately I got into the middle of a very busy week with a couple of assignments and quizzes lined up for the past Friday

14:48 < rcurtin> vivekp: you aren't late, the deadline has not passed yet :)

15:01 < vivekp> rcurtin: yes, indeed :) but I was just aiming to get it reviewed by Friday or so, so that I could work on its improvement over the weekend

15:11 < nish21> Hello, my proposal is almost ready and would like to get it reviewed once its done. Where do i upload it?

15:47 < shikhar> Hello! I was trying to isolate the issue which causes the CNN test to fail. I am able to consistently replicate the issue, with a particular seed value and a large enough training set size the network stops learning.

15:48 < shikhar> Some test runs : https://gist.github.com/shikharbhardwaj/6a8541f73e53ea3c27e8d8f373192659

16:02 < zoq> shikhar: Thanks for taking a look into the issue, might be interesting to take a look at the weights and probably test some other initialization strategies. (Really like the shell scheme).

16:06 < zoq> naxalpha: We are here to help; also you don't have to wait just ask right away, once we have a chance we will give you an answer.

16:06 < zoq> Maybe he checks the logs.

16:13 < shikhar> zoq: Omitting the first training example makes the network learn, with error 0.046. I'll look into the weights and init. The shell theme is a slight modification of the default one that comes with oh-my-zsh, here : https://github.com/shikharbhardwaj/dotfiles/blob/master/zsh/robbyrussell_mod.zsh-theme :)

16:43 < sagarbhathwar> zoq: I had a question about modern approaches to RBM. Are we particularly looking at spike and slab RBM or research more into modern variants of RBM?(Like Gaussian RBM and Mean Covariance RBM)?

16:43 < naxalpha> zoq: ok got it, I had to ask about nuget package include files, Is every hpp file include file for nuget? or how can I get list of include files?

16:55 < zoq> sagarbhathw: The great think about the ssRBM is that it can be used as building block for e.g. DBN, which isn't possible with cRBM or mcRBM. Also, not sure what you mean with more modern, the ssRBM is a modern approach at if I remember right it outperforms ccRBM and mmCRBM on a bunch of datasets like CIFAR-10. Anyway, I'm not against other approaches, if you like you can go for a different approach.

16:57 < zoq> naxalpha: Not sure what you mean here, every file that is in e.g. build/include once you build mlpack should be included.

17:01 < sagarbhathwar> zoq: Apologies, my question was worded wrongly by putting 'modern variants' instead of 'other variants'. Thanks for the feedback :) I shall concentrate on ssRBM

17:03 < zoq> sagarbhathw: No worries, don't get me wrong the mcRBM approach is also quite interesting, if you like the idea or feel more confident about it, go for it.

17:20 < sagarbhathwar> zoq: Not at all! I was mainly researching on modern RBMs and came across interesting papers on applications of cmRBM and GRBM. Just wanted to make sure ssRBM was the way to go. Thank you

17:21 < sagarbhathwar> mcRBM* i mean

20:09 < amoudgl> Hi!

20:09 < amoudgl> I am not able to build locally. CMake gets stuck after few steps. Log: http://pastebin.com/tGFjGdmk

20:10 < amoudgl> Just need some ideas on how to fix this issue.

20:10 < amoudgl> Tried uninstalling everything and building again, didn't work.

21:43 < rcurtin> amoudgl: I don't see any errors there, just warnings

21:44 < rcurtin> I think it configured just fine and you can probably go ahead and build

21:44 < amoudgl> Yes I agree. But after the last step, it doesn't terminate. It just get stuck there for hours.

21:46 < rcurtin> oh... interesting

21:47 < rcurtin> is the disk full or otherwise having a problem?

21:48 < amoudgl> I work on my institute server, not sure about disk space. I'll just check once.

21:51 < amoudgl> But yes, I strongly feel it may be the case. I had weight files (~150GB) on a separate dataserver which I shifted few days back to my home folder. I *guess* since then, I am not able to build.

22:44 < amoudgl> Moved everything back to dataserver. Still not working.