00:24 < rcurtin> Suryo: the thing is, CNE takes forever to actually run and it's a bit flaky. so I don't have a problem testing with CNE so long as the test time doesn't take too long (like hopefully only a handful of seconds) and the tests are reliable with different random seeds

00:24 < rcurtin> when I say it's flaky, I only mean that it is not guaranteed to actually get to the minimum reliably (because initial conditions and random new point selection, etc.)

00:35 < zoq> I think the idea right now is to test with CNE locally and later test the functions against PSO?

00:37 < mlJens> Hi. Just a question. Is there still a fix for the convolution model?

00:37 < mlJens> https://github.com/mlpack/mlpack/pull/1554

00:39 < Suryo> rcurtin: yeah I understand this. So that's why I'm selecting the initial points close to the global optima. I mean, it defeats the purpose of testing the optimizers by a lot, but for what it's worth, I want to test the functions as much as possible before resuming the actual development with PSO.

00:54 < rcurtin> Suryo: got it---sounds good to me

00:54 < Suryo> Thanks for the affirmation. I'll get with it.

00:55 < rcurtin> :)

02:17 < robertoh1eso> Hi! I'm working on the implementation of http://papers.nips.cc/paper/3539-fast-high-dimensional-kernel-summations-using-the-monte-carlo-multipole-method.pdf I've implemented KDE Monte Carlo summation for single trees but I can't get it to comply with the relative error tolerance in any case after many tests.

02:19 < robertoh1eso> Maybe I missunderstood the paper, but although it gets really close, it always exceeds relative error and I was wondering if it's supossed to comply. Take into account that I haven't implemented PCA trees yet, only summation.

03:06 < Suryo> zoq: don't worry, I think I figured it out.

03:32 < Suryo> zoq: all the new tests for the new functions have passed. Waiting on your review, overall. I'll look into the code styling in a bit. Hoping to merge soon!

04:49 < rcurtin> robertohueso: it's a little late here now but I'll take a look at the paper closely tomorrow afternoon (I am busy in the morning) and let you know what I think

05:20 < favre49> zoq: I'm sorry but my progress might be a bit slow for the next couple of days since I've fallen quite ill. I should be fine in a few days though, and I'll still be coding, probably just less.

05:32 < rcurtin> favre49: get some rest! I hope you feel better soon :)

08:05 < jenkins-mlpack2> Project docker mlpack nightly build build #344: STILL UNSTABLE in 3 hr 51 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/344/

09:04 < mlJens> Hi. I wanted to ask, if there is still a fix for the convolutional model available ==> https://github.com/mlpack/mlpack/pull/1554

10:25 < robertoh1eso> rcurtin: Thanks! :)

12:11 < toshal> zoq: Sorry, For the late reply. Yes we can give it more try.

12:11 < toshal> I would resent the mail.

12:13 < toshal> will*

14:06 < zoq> favre49: No worries, health always comes first.

14:07 < zoq> mlJens: Not sure I get what you mean.

14:08 < zoq> Suryo: Great, will take a look.

17:46 < mlJens> zoq: Hey, there was a bug reported and as far as I understood also confirmed.

17:46 < mlJens> More details here: https://github.com/mlpack/mlpack/pull/1554

17:46 < mlJens> If I run the DigitRecognizerCNN.cpp, I also get an error unless I change BATCH_SIZE to 1

17:47 < mlJens> Just wanted to know, if there are any news about that.

17:47 < mlJens> I installed the latest version available on your HP

21:52 < jenkins-mlpack2> Project mlpack - git commit test build #185: UNSTABLE in 45 min: http://ci.mlpack.org/job/mlpack%20-%20git%20commit%20test/185/

23:34 < ShikharJ> sakshamB: toshal: I was wondering if the meeting timings can be made a bit earlier? Maybe 7pm IST?