02:05 < rcurtin> Toshal: sorry for the slow response, sent you a PM

02:05 < rcurtin> oh wait, not online, oops

07:48 < jenkins-mlpack2> Project docker mlpack nightly build build #336: STILL UNSTABLE in 3 hr 34 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/336/

08:55 < ShikharJ_> rcurtin: zoq: Hmm, I was wondering if not for ML/AI, which area within CS would've been your preferred area of specialization?

13:11 < rcurtin> ShikharJ_: my original interests were in speech processing and speech synthesis; but when I got to grad school, those fields were being quickly supplanted by more pure machine leadning techniques so I ended up doing machine learning :)

13:46 < jenkins-mlpack2> Project mlpack - git commit test build #177: UNSTABLE in 47 min: http://ci.mlpack.org/job/mlpack%20-%20git%20commit%20test/177/

17:50 < mulx10> Is this method acceptable for image loading : https://pastebin.com/yP3bUe70?

17:51 < mulx10> it uses header only library (https://github.com/nothings/stb/blob/master/stb_image.h).

17:52 < mulx10> rcurtin, zoq: Please let me know.

17:52 < mulx10> Thanks

17:53 < zoq> mulx10: Is it compatible with the 3-Clause BSD license?

18:20 < rcurtin> mulx10: that's ok with me but let's avoid including stb directly in our project (perhaps CMake should simply download it if it is not available)

18:25 < MLJens> Hi. Can anybody tell me, which lib is containing the arma::mat<double> refernce? Just started with MLPack and got a linker error compiling my project.

18:32 < rcurtin> ShikharJ_: my original interests were in speech processing and speech synthesis; but when I got to grad school, those fields were being quickly supplanted by more pure machine leadning techniques so I ended up doing machine learning :)

18:33 < rcurtin> oops, didn't mean to send that message twice, sorry

18:33 < rcurtin> MLJens: maybe you need -larmadillo?

18:33 < rcurtin> or -lmlpack

18:36 < MLJens> still referenced mlpack.lib and armadillo.lib

18:36 < rcurtin> MLJens: hmm, can you provide the error in more detail/

18:36 < rcurtin> *?

18:37 < MLJens> ErrorLNK2001unresolved external symbol "class arma::Mat<double> data" (?data@@3V?$Mat@N@arma@@A)

18:37 < MLJens> I still know those errors, but I don't know, which file to reference in this case.

18:40 < rcurtin> does it give any more information than that? like which file defines the 'data' matrix

18:41 < rcurtin> and is it possible that you've defined some matrix 'data' as 'extern' somewhere perhaps?

18:43 < MLJens> Yes, external arma::mat data;

18:43 < MLJens> No, the file containing the reference is, what I need to know to link it :)

18:43 < rcurtin> ok, but if you have 'extern arma::mat data' somewhere, you also need to define that

18:43 < rcurtin> are you using one of the example programs in the documentation?

18:43 < rcurtin> the idea there is just that there exists some matrix you are running on, so you should replace the 'extern arma::mat data' with whatever is needed to load the data you want to use

18:45 < MLJens> Thank you, that is it. I still wondered. Usually, I use external to write an function for a Dll :)

18:46 < MLJens> Yes, trying the examples to get used to mlpack.

18:47 < MLJens> really looks nice...

18:47 < MLJens> Thank you for your help. Now, I have to make a break as the football final starts soon :)

18:50 < rcurtin> MLJens: sounds good, enjoy the game :)

18:51 < Toshal> rcurtin: Sorry, at first I didn't get that a PM in your message meant PM on Irc. So here I am.

18:59 < MLJens> Half of the match is still played. Remembered the wrong time of kickoff :(

19:12 < ShikharJ_> rcurtin: Interesting! And what about you zoq? Do you have an alternative field, that you were deeply interested at the time?

19:19 < zoq> ShikharJ_: Completed some really interesting physics courses. Would love to dive in more :)

19:21 < ShikharJ_> zoq: Quantum? Or Classical?

19:21 < zoq> classical

19:24 < ShikharJ_> Nice

19:32 < ShikharJ_> zoq: I'll be mostly finished with running the tests for https://github.com/mlpack/mlpack/pull/1823 by tomorrow, and I wish to merge it at the earliest. I'd love your input on that whenever you're free.

19:42 < ShikharJ_> Toshal: I have merged in the OReilly GAN fix PR, since I felt it was complete. After that, I'll merge in the Memory Sharing between GAN and FFN PR. Do you have a preference of me reviewing the dual optimizer PR first? Or can I focus on getting the Serialization PR merged (I personally feel that's more important)?

19:55 < rcurtin> ShikharJ_: if you're moving fast with the PRs, I can wait to release 3.1.1 until you have the serialization PR merged (I'm hoping to wait no later than tomorrow)

19:55 < rcurtin> the other two (memory sharing and dual optimizer) can wait until 3.2.0 since that would be adding functionality

19:57 < Toshal> ShikharJ: Okay I am fine with merging serialization first. It's upto you. But, I guess Dual Optimizer PR may change quite a lot of code. But I am fine with it.

19:59 < ShikharJ_> rcurtin: Let's wait on the GANs, I'd rather release them when we have the full functionality. We solved two issues with the convergence and memory usage, and two issues pertaining to (serialization and optimizer) need to be worked upon a bit. When are you planning on releasing 3.2.0?

20:00 < ShikharJ_> Toshal: Okay, then let's work on Serialization first, since it is a more pressing concern. Dual optimizer requires an API change, but isn't super-essential in the absolute sense. I'll merge the memory sharing work in the morning tomorrow, and maybe then you can continue with the serialization work.

20:01 < rcurtin> ShikharJ_: I had no particular date for 3.2.0, just whenever some more stuff is ready for release :)

20:02 < rcurtin> and ok, I won't wait for the GAN PRs, whenever the RF test fix is merge I'll release 3.1.1

20:02 < rcurtin> thanks for the info :)

20:04 < ShikharJ_> Cool, other than what I have mentioned above, I think the work is pretty complete and self contained. Maybe a 3.1.2 can happen (I think none of the GAN related work has made into any release).

20:05 < ShikharJ_> With 3.2.0, we can add a lot more functionality and evaluation metric support, and different variations of GANs.

20:05 < rcurtin> yeah; basically if there is a patch or tiny change, 3.1.2 can be the release number, if we add anything new then it can be 3.2.0

20:06 < ShikharJ_> rcurtin: Where can I find the release notes of previous releases?

20:06 < rcurtin> HISTORY.md should contain them all

20:07 < ShikharJ_> Okay, great.

20:07 < Toshal> ShikharJ: Okay. I will merge conflicts tommorow.

20:07 < ShikharJ_> saksham189_: Are you there?

20:07 < saksham189_> ShikharJ_: yes

20:07 < ShikharJ_> I assume you have seen the previous conversation?

20:08 < saksham189_> yes, I will work on adding minibatch discrimination layer

20:08 < ShikharJ_> Okay great. Try to formulate an idea of how it is going to be implemented, and we will discuss on Monday.

20:08 < ShikharJ_> 9.30 pm

20:09 < ShikharJ_> And what about your work on Highway networks? Do you feel that's ready for review?

20:09 < saksham189_> yes, I am waiting for a review on that.

20:11 < ShikharJ_> saksham189_: Cool, I'll take a look on that tomorrow as well. For the first phase, we'll focus purely on increasing GAN toolkit and making the existing framework more robust.

20:11 < ShikharJ_> And GAN varients will follow after that.

20:12 < ShikharJ_> *variants

20:13 < ShikharJ_> saksham189_: Did you happen to travel already? Let me know if you need some time off for doing that.

20:13 < saksham189_> ShikharJ_: thanks for the follow-up. sounds good to me.

20:13 < saksham189_> yes I got tired after travelling and couldn’t attend the meet.

20:14 < ShikharJ_> Ah, I see. Currently in Vietnam?

20:14 < saksham189_> yes

20:16 < ShikharJ_> Okay good, have a nice time. I'll be off for now. See you on Monday then.

20:17 < saksham189_> alright, see you then. :)

20:19 < ShikharJ_> rcurtin: Hmm, I can confirm GANs are not a part of the releases, so let's try to get them for 3.2.0.

20:19 < rcurtin> ShikharJ_: sounds good

20:20 < rcurtin> I had forgotten I pulled them out, I guess