07:42 < jenkins-mlpack2> Project docker mlpack nightly build build #368: STILL UNSTABLE in 3 hr 28 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/368/

08:41 < KimSangYeon-DGU> sumedhghaisas: I think I implemented QGMM in python and now, I'll research the accuracy of QGMM using original version and edited version like NLL + lambda * approx_constant

08:42 < KimSangYeon-DGU> I'll update our research repository and mlpack blog

08:47 < KimSangYeon-DGU> Oops.. my internet connection was broken..

08:48 < KimSangYeon-DGU> The trained parameters are converged now, and the yesterday's divergent problem that I said is because the alpha value updating problem, so I edited it.

08:49 < KimSangYeon-DGU> I'll also compare the several QGMM versions with the classical GMM.

08:50 < KimSangYeon-DGU> :)

08:53 < KimSangYeon-DGU> I think the main point for us to make our QGMM better is to find a neat algorithm to update the cos(phi) that is associated to alphas.

10:02 < favre49> zoq In line 197 of neat_test_tools.hpp, I keep getting the error "'ContinuousMultiplePoleCart’ does not name a type", despite including the header file.

10:02 < favre49> What can I do about that?

10:06 < zoq> favre49: The ContinuousMultiplePoleCart is in another namespace: rl; using namespace mlpack::rl; should fix the issue

10:12 < favre49> zoq Of course, can't believed I missed that. Thanks!

10:14 < zoq> favre49: Happy to help.

15:49 < zoq> favre49: I thought about the innovation IDs blow up issue in the speciation phase, this shouldn't happen.

15:55 < zoq> favre49: Btw. thanks for all the comments in the code, they are really helpful.

16:37 < favre49> zoq Thanks for the clarification!

16:37 < favre49> No problem, they're very helpful when I'm debugging too :)

17:31 < favre49> I noticed I've made a mistake in the multiple pole cart environments that we missed, and the dimensions of the `data` matrix are wrong, breaking it. Should I open a different PR for this, or can I include it in the current RL PR I'm working on?

17:53 < travis-ci> robertohueso/mlpack#24 (mc_kde_error_bounds - 4d0552b : Roberto Hueso Gomez): The build has errored.

17:53 < travis-ci> Change view : https://github.com/robertohueso/mlpack/compare/dbcd637c4b49...4d0552b2cbb4

17:53 < travis-ci> Build details : https://travis-ci.org/robertohueso/mlpack/builds/550908924

19:08 < zoq> favre49: Let's make it part of #1941.

22:51 < robb08> hey, is it possible to use a convolutional layer in a recurrent container (and only have it run once)? What's the difference between the containers