10:22 < mentekid> rcurtin: take a look at my comment on PR #700 (https://github.com/mlpack/mlpack/pull/700) when you have time. Do you think we should implement our own (parallel) versions of unique and find?

11:45 < rcurtin> mentekid: sure, I am almost in the office, I will look when I get in

11:55 < mentekid> I'm not sure if what I did in CMakeLists makes any sense actually so I'll change that (hopefully travis will stop failing then...)

11:56 < mentekid> but my question is regardless of that

11:56 < mentekid> unrelated to*

12:16 < mentekid> I think I fixed it... I'm new to cmake so take a look

12:34 < Karl_> Hi, I'm new to MLpack. I'm trying to write a kernel based on Fisher's exact test (not sure if it will make sense in the end). My current goal is to use this kernel with kPCA. I've gotten to the point of a functioning kernel that compiles into mlpack_kernel_pca. My problem is how to present the data. I need two values for each point, and that adds a third dimension to the matrix. Is anyone aware of a good solution?

12:38 < mentekid> nope, didn't work... Travis's build crashes because it can't find openmp_max_num_threads(), but I thought my fiddling with CmakeFiles made him skip that line...

13:10 < mentekid> ok finally it seems to pass LSH without error :)

13:10 < mentekid> What I've done is a bit ugly, I'll try simplifying it a bit

13:51 < mentekid> rcurtin: Compiling with OpenMP off will produce a lot of warnings about unknown pragmas. Would disabling that warning with -Wno-unknown-pragmas make sense or should I leave them be?

14:12 < rcurtin> mentekid: sorry that I have not gotten back to you yet

14:12 < rcurtin> I have spent the past two hours digging myself out of an email hole, and I am not yet done with that

14:12 < rcurtin> I agree that -Wno-unknown-pragmas makes sense

14:13 < mentekid> no problem, take your time. I'll go ahead and add that and start implementing parallel find (easier than parallel unique)

14:14 < mentekid> Oh and I haven't made my post yet so I'll do that first

14:15 < rcurtin> ok, thanks. expect to hear more from me today, I am just not sure when. the problem with vacations is that when I come back, it takes days to get everything sorted out

15:13 < mentekid> how come my blog post doesn't appear? Do I have some weird syntax error in my markdown? Github seems to render it fine

15:16 < rcurtin> the date looks right to me... hang on

15:16 < rcurtin> ah, here we go

15:16 < rcurtin> http://big.mlpack.org/job/gsoc-blog/123/console

15:16 < rcurtin> there's the log output... it looks like your new post has the same title as last week :)

15:17 < mentekid> ah

15:17 < rcurtin> do you know how to navigate Jenkins? I can show you around if you like

15:17 < mentekid> I'm not that creative with my titles

15:17 < mentekid> no I don't, tell me :)

15:18 < rcurtin> sure, so, I should also say, the Jenkins server we have now is in a state of turmoil, since we are waiting on this big server to be scanned

15:19 < rcurtin> (they finally got it online, they are preparing to do the scan... I don't know how long that will take...)

15:19 < rcurtin> so we run Jenkins as a way to check the git master branch builds and run the benchmarking system

15:19 < rcurtin> but it turns out it is also useful for building the blog

15:19 < rcurtin> so if you go to http://big.mlpack.org/, you can see the main view with all of the jobs

15:20 < rcurtin> (many are failing, and you can see in the left column that many of the build slaves are offline---like I said, it is a bit of turmoil right now...)

15:20 < rcurtin> the job you are interested in here is the "z gsoc" blog (the "z" is so that the mlpack jobs show up first in the alphabetical list)

15:20 < rcurtin> so you can click on the "z gsoc" tab to expand it and then you can click on the "gsoc-blog" project

15:21 < rcurtin> on the left side you can see the build history, and if you click on a build, you can see the console output, list of changes, and other information (really only the console information is ever important)

15:22 < mentekid> I see

15:22 < rcurtin> you can explore the other projects in the same way. Jenkins even supports big matrix builds to test lots of different configurations, but the "mlpack - nightly matrix build" is neglected until the new system comes online

15:26 < mentekid> Aha... So how exactly is jenkins different from travis - or are they the same thing?

15:44 < rcurtin> they are similar tools... in this particular case, for Jenkins, we manage our own hardware

15:44 < rcurtin> whereas with Travis we do not

15:44 < rcurtin> Travis only gives us limited resources, so something like "build mlpack with all possible Armadillo versions, every night" is not realistic

15:44 < rcurtin> (we probably could do that, but we would need to pay a lot of money to do that)

15:45 < rcurtin> I think Jenkins is also somewhat more flexible and capable than Travis; I am not sure if Travis can even do matrix builds

15:46 < mentekid> I see. Cool, thanks for telling me all that. I only knew I should keep an eye on travis to see what I broke :)

15:46 < rcurtin> yeah, travis is nice because it builds PRs too

15:46 < rcurtin> I think you can set up Jenkins to do that, but I haven't done that

15:46 < rcurtin> I was reading, I think that Travis actually does do matrix builds, so maybe both tools are equally flexible

15:52 < mentekid> So, unrelated - I moved to UK and hired a co-working space for the next months

15:52 < mentekid> rented*

15:53 < mentekid> but it only stays open 8:30 - 17:30 GMT, so that's the hours I'll be most available

15:54 < mentekid> I do have internet at home though so I can arrange other working times as well

15:56 < rcurtin> that's fine with me, usually my mornings are what I devote to GSoC anyway so that will coincide with your afternoons

16:05 < travis-ci> mlpack/mlpack#1080 (master - 479eca0 : Ryan Curtin): The build passed.

16:05 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/37fda23945b4...479eca0c625c

16:05 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/140565215

18:25 < zoq> mentekid: great blog post :)

18:25 < mentekid> zoq: thanks :D

19:28 < travis-ci> mlpack/mlpack#1084 (master - 0caf99c : Ryan Curtin): The build has errored.

19:28 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/479eca0c625c...0caf99c9231b

19:28 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/140615737

19:45 < travis-ci> mlpack/mlpack#1085 (master - 9dd66c7 : Ryan Curtin): The build passed.

19:45 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/0caf99c9231b...9dd66c7312ff

19:45 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/140618347

19:49 < marcosirc> mentekid: Hi, in the title of your blog post, I think you were confused when writing "Week 2".

22:16 < vedantrathore> Hey Everyone I am a newbie here..I wanted to particitpate in Gsoc 2017 under mlpack so can someone give any guidance?

22:21 < zoq> vedantratho: Hello there, a good way to start would be to read this document: http://www.mlpack.org/involved.html

22:23 < vedantrathore> I've read it already and built mlpack from source code..I wanted to know that should I solve more bugs to increase my chances for next year..

22:30 < zoq> Getting in contact with the community is always a plus. If you find something interesting on the issue list, feel free to assign yourself. If you have some interesting idea or code, you can open a pull request or issue on Github and we will discuss it with you. We can also, discuss some ideas here.

22:32 < vedantrathore> Okay...Can I know what are the project ideas for the next gsoc so that I can research more on those topics

22:36 < zoq> We haven't really thought about it, I guess, you can use the current project ideas as starting point: https://github.com/mlpack/mlpack/wiki/SummerOfCodeIdeas; also new, interesting ideas are also suitable.

22:42 < vedantrathore> Okay..I look forward to contribute in mlpack.

22:43 < zoq> Sounds great, let us know if you need any help.

22:46 < rcurtin> Karl_: I forgot to respond to you earlier, can you give more details on what you are trying to do?

22:46 < vedantrathore> okay :)

23:28 < zoq> nilay: new pca timings: 0.453853s instead of 10.168902s, by using randomized SVD, I think that would work

23:33 < zoq> nilay: after some tuning: pca_timer: 0.243556s