IRC logs of #shogun for Friday, 2013-08-30

« Thursday, 2013-08-29 Index Saturday, 2013-08-31 »
--- Log opened Fri Aug 30 00:00:37 2013
@iglesiasghey gsomix_!00:01
@iglesiasggsomix_, how are you doing?00:01
@sonney2kiglesiasg, I think he must be sleeping by now00:05
@sonney2kit is like 2 or 3 am :)00:05
@iglesiasgsonney2k, aham, what is his timezone?00:05
@iglesiasgI thought it was somewhere in the US00:05
@sonney2kiglesiasg, smae like lisitsyn's00:05
@iglesiasgno idea what smae is :S00:06
@sonney2ksame :P00:07
@iglesiasgaah hehe00:07
@iglesiasgI checked http://acronyms.thefreedictionary.com/SMAE and I was like what?!00:08
@sonney2kSailor Moon Awards Engine of course00:09
@sonney2kman!00:09
@iglesiasgyeah that one was best00:09
@sonney2kiglesiasg, ohh man I finally ordered this damn sportswatch00:11
@iglesiasghehe good00:12
@sonney2kor maybe not00:12
@sonney2know I will have to go by bike to job and do all kinds of sports00:12
@iglesiasghehe00:13
@iglesiasgI did my personal record in a race last week btw!00:13
@iglesiasg5 km -- 20:3000:13
@sonney2kI don't run00:14
@iglesiasgrun is for weaks :D00:16
@sonney2kiglesiasg, no I am a lame duck on land00:17
@sonney2kiglesiasg, btw I think it might also be swig related00:17
@sonney2kiglesiasg, could be he has an old swig version that causes issues in the generated wrapper code00:17
@iglesiasgit could be, he mentioned the error happened in the Python part00:17
-!- hushell [~hushell@c-98-232-178-161.hsd1.or.comcast.net] has quit [Ping timeout: 256 seconds]00:40
@iglesiasgHeikoS, ping01:09
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has joined #shogun01:25
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has quit [Ping timeout: 246 seconds]02:04
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Ex-Chat]02:30
-!- hushell [~hushell@c-98-232-178-161.hsd1.or.comcast.net] has joined #shogun03:15
-!- sonne|osx_ [~sonne@f053036061.adsl.alicedsl.de] has joined #shogun03:46
-!- sonne|osx [~sonne@f053046168.adsl.alicedsl.de] has quit [Ping timeout: 240 seconds]03:48
-!- sonne|osx_ is now known as sonne|osx03:48
-!- HeikoS [~androirc@90.204.242.124] has quit [Ping timeout: 240 seconds]05:51
-!- gsomix_ [~gsomix@109.169.146.219] has quit [Ping timeout: 264 seconds]06:36
-!- gsomix_ [~gsomix@80.234.26.132] has joined #shogun06:48
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun []06:49
-!- foulwall [~zhengyang@114.255.40.7] has quit [Quit: leaving]06:57
-!- foulwall [~zhengyang@114.255.40.7] has joined #shogun07:07
-!- sonne|osx [~sonne@f053036061.adsl.alicedsl.de] has quit [Quit: sonne|osx]07:32
-!- sonne|osx [~sonne@89.204.139.233] has joined #shogun08:10
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun08:11
shogun-notifier-shogun: Fernando Iglesias :develop * 428be1e / src/shogun/ (18 files): https://github.com/shogun-toolbox/shogun/commit/428be1e1302ad2d95d65d0dd981a8b72e8ed72b408:11
shogun-notifier-shogun: Fix a few Doxygen warnings.08:11
shogun-notifier-shogun: Soeren Sonnenburg :develop * 9dfa747 / src/shogun/ (18 files): https://github.com/shogun-toolbox/shogun/commit/9dfa7474585f99f3707b50caf1e4cbbfb59098a008:11
shogun-notifier-shogun: Merge pull request #1497 from iglesias/develop08:11
shogun-notifier-shogun:08:11
shogun-notifier-shogun: Fix a few Doxygen warnings.08:11
shogun-notifier-shogun: Fernando Iglesias :develop * 894924d / doc/ipython-notebooks/metric/LMNN.ipynb: https://github.com/shogun-toolbox/shogun/commit/894924d09d01461a7a5602987281269bd31e176008:12
shogun-notifier-shogun: Add ipython notebook with LMNN applied to toy data set.08:12
shogun-notifier-shogun: Soeren Sonnenburg :develop * 9a4060e / doc/ipython-notebooks/metric/LMNN.ipynb: https://github.com/shogun-toolbox/shogun/commit/9a4060e25ba28bdfc7a97de30d40509fb61124de08:12
shogun-notifier-shogun: Merge pull request #1498 from iglesias/feature/lmnn08:12
shogun-notifier-shogun:08:12
shogun-notifier-shogun: First stage of the LMNN ipython notebook.08:12
shogun-buildbotbuild #1933 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1933  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>08:19
sonne|osxfoulwall: good morning!08:22
sonne|osxfoulwall: how is it going?08:22
sonne|osxbrb08:26
-!- sonne|osx [~sonne@89.204.139.233] has quit [Quit: sonne|osx]08:26
-!- gsomix__ [~gsomix@109.169.159.34] has joined #shogun08:26
-!- gsomix_ [~gsomix@80.234.26.132] has quit [Ping timeout: 264 seconds]08:29
lisitsynsonne|work: so when is your flight? when do you arrive?08:41
sonne|worklisitsyn: same like HeikoS08:41
lisitsynsonne|work: ah so you share the flight?08:41
lisitsynsonne|work: ok then I'll go for that lufthansa flight that arrives 12:2008:42
lisitsynI'll wait you there and then we go08:42
sonne|worklisitsyn: you will likely need 1 hour extra to enter the US08:42
lisitsynsonne|work: yes may be even more08:42
lisitsynso it is unclear who gonna wait08:42
sonne|worklisitsyn: and you will not allowed to use a mobile phone08:43
sonne|workeven though they have wifi on the airport08:43
sonne|workanyway we will somehow make it08:43
lisitsynsonne|work: well you know how I look like and I know how you look like08:43
sonne|worklisitsyn: luckily08:43
lisitsynthat's a plus08:43
lisitsynsonne|work: what's the deal with mobile phone?08:44
sonne|workalright got a new machine so reboot time08:44
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has left #shogun []08:44
thoralfHeyho.08:48
lisitsynthoralf: hey08:50
thoralflisitsyn:)08:54
shogun-buildbotbuild #1565 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1565  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>09:10
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has quit [Quit: Leaving.]09:14
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has joined #shogun09:38
shogun-buildbotbuild #1650 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1650  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>09:45
shogun-buildbotbuild #1349 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1349  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>09:45
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun09:55
-!- HeikoS [~androirc@90.204.242.124] has joined #shogun10:05
-!- mode/#shogun [+o HeikoS] by ChanServ10:06
-!- gsomix_ [~gsomix@95.67.177.128] has joined #shogun10:06
-!- gsomix__ [~gsomix@109.169.159.34] has quit [Ping timeout: 245 seconds]10:08
-!- van51 [~van51@ppp-2-87-4-38.home.otenet.gr] has joined #shogun10:24
shogun-notifier-shogun: Fernando Iglesias :develop * 797ca65 / src/shogun/metric/ (4 files): https://github.com/shogun-toolbox/shogun/commit/797ca650950de0aed31535017c5ce148b9830f5c10:43
shogun-notifier-shogun: Add option in LMNN to learn a diagonal transformation.10:43
shogun-notifier-shogun: Fernando Iglesias :develop * 65a4045 / tests/unit/metric/LMNN_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/65a404568fedc65529cd641da82aa2c4bfebed6010:43
shogun-notifier-shogun: Unit test for LMNN diagonal mode.10:43
shogun-notifier-shogun: Fernando Iglesias :develop * bc05ff8 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/bc05ff88b3e6e1d4b0196a2dc8201a48595665ab10:43
shogun-notifier-shogun: Merge pull request #1461 from iglesias/feature/lmnn_diagonal10:43
shogun-notifier-shogun:10:43
shogun-notifier-shogun: Add option in LMNN to learn a diagonal transformation.10:43
lisitsynsonne|work: https://code.google.com/p/googletest/downloads/detail?name=gtest-1.7.0-rc1.zip10:58
lisitsyngood timing10:58
sonne|workfinally...10:58
shogun-buildbotbuild #1566 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1566  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:03
-!- az_de [82954e22@gateway/web/freenode/ip.130.149.78.34] has joined #shogun11:07
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has quit [Remote host closed the connection]11:10
-!- az_de [82954e22@gateway/web/freenode/ip.130.149.78.34] has quit [Client Quit]11:12
-!- HeikoS1 [~heiko@nat-183-166.internal.eduroam.ucl.ac.uk] has joined #shogun11:29
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has joined #shogun11:29
-!- iglesiasg [~iglesias@2001:6b0:1:1041:49a7:bd41:6df2:b1e0] has joined #shogun11:32
-!- mode/#shogun [+o iglesiasg] by ChanServ11:32
shogun-buildbotbuild #1934 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/193411:35
shogun-buildbotbuild #1350 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1350  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:40
shogun-buildbotbuild #1567 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1567  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:42
-!- van51 [~van51@ppp-2-87-4-38.home.otenet.gr] has quit [Quit: Leaving.]11:47
shogun-buildbotbuild #1651 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/165112:01
shogun-buildbotbuild #1351 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1351  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>12:22
HeikoS1iglesiasg: pong12:31
lisitsynoh 90 mins12:32
lisitsyngood12:32
@iglesiasgHeikoS1, mm let me remember what was it I wanted to ask12:32
HeikoS1iglesiasg, lisitsyn: please have a look at the meeting draft I sent12:32
HeikoS1iglesiasg: haha :)12:32
@iglesiasgHeikoS1, I don't think I do remember fuck12:33
HeikoS1lisitsyn: also, do you have an idea about this:12:33
HeikoS1https://travis-ci.org/shogun-toolbox/shogun/jobs/1076360412:33
HeikoS1iglesiasg: haha :) dont worry12:33
@iglesiasgnext time I will just write it in the chat12:33
HeikoS1iglesiasg: I like your notebook12:33
HeikoS1iglesiasg: do we have some algorithms that can take a distance metric?12:34
HeikoS1iglesiasg: because would be cool to see that in there12:34
HeikoS1like you mention this knn one12:34
HeikoS1can ours do that?12:34
@iglesiasgyeah sure12:34
@iglesiasgthat's how I benchmark our LMNN12:34
@iglesiasgthe flow is like this12:35
lisitsynHeikoS1: yes I know12:35
@iglesiasgdata -> LMNN -> distance -> kNN -> classification12:35
HeikoS1iglesiasg: cool, so this would be nice to have in the toy example12:35
HeikoS1then one can compare classic knn vs metric knn12:35
@iglesiasgroughly, kNN also gets the data of course12:35
@iglesiasgHeikoS1, yeah sure, still to come :)12:35
HeikoS1lisitsyn: any ideas about that?12:35
@iglesiasgGeorg had a nice idea about plotting neighbours graphs12:36
lisitsynHeikoS1: fixing12:36
@iglesiasgand compare the neighbours found by kNN with Euclidean and kNN with LMNN-distance12:36
HeikoS1lisitsyn: how?12:36
HeikoS1iglesiasg: sweet12:37
HeikoS1this is exactly the stuff we need :D12:37
shogun-notifier-shogun: Sergey Lisitsyn :develop * c9aea34 / src/interfaces/modular/modshogun_ignores.i: https://github.com/shogun-toolbox/shogun/commit/c9aea344300ebee55feda78443c33228892a22ce12:37
shogun-notifier-shogun: Ignore sparsity structure in modular12:37
HeikoS1iglesiasg: do these things also work on manifolds?12:37
lisitsynHeikoS1: ^12:37
@iglesiasgHeikoS1, let me show you a new version of the notebook12:37
HeikoS1lisitsyn: cool, do you think its safe to push now? or should I ask travis again?12:37
@iglesiasgHeikoS1, yeah, you can actually do dim red with LMNN12:37
lisitsynHeikoS1: where do you want to push it to?12:38
HeikoS1iglesiasg: wow nice!12:39
HeikoS1lisitsyn: I updated my PR, lets see12:39
@iglesiasgHeikoS1, at the end I explained a bit this idea http://nbviewer.ipython.org/633421412:39
lisitsynHeikoS1: ah got it - I think it should work now12:39
HeikoS1lisitsyn: will wait for https://travis-ci.org/shogun-toolbox/shogun/builds/1079368312:40
@iglesiasgHeikoS1, instead of interpreting LMNN's result as a distance measure, there is another interpretation of it as a linear transformation12:40
@iglesiasgso in the last plot what I am showing is how the data looks like after applying this transformation to the data12:40
HeikoS1iglesiasg: nice, reading12:40
HeikoS1iglesiasg: btw you can just import pylab12:40
HeikoS1then you get numpy and matplotlib12:41
@iglesiasgI might change it12:41
HeikoS1iglesiasg: not really important, just wanted to mention it since its less typing :)12:41
@iglesiasgI am in this period of my life where I like to be implicit importing python modules :)12:41
@iglesiasgexplicit sorry, no implicit12:41
-!- lisitsyn [~lisitsin@mxs.kg.ru] has left #shogun []12:42
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun12:42
HeikoS1haha12:42
lisitsynHeikoS1: I am ok with your draft12:43
HeikoS1lisitsyn:  even with aksing for more and more things? :)12:43
lisitsynHeikoS1: yeah loooks ok12:43
HeikoS1alright12:43
lisitsynHeikoS1: have you seen sonne|work had some troubles with cheap flight?12:44
HeikoS1lisitsyn: ?12:44
HeikoS1he told me he booked12:44
HeikoS1939 euros or so12:44
HeikoS1but from berlin12:44
lisitsynHeikoS1: well it was more expensive than expected?12:44
HeikoS1yeah12:44
HeikoS1lisitsyn: saw that12:44
HeikoS1lisitsyn: well its a US flight, these are expensive12:44
@iglesiasglisitsyn, but do you have a limit in the money you can spend for the flight?12:44
HeikoS1mine is quite cheap12:44
lisitsyniglesiasg: IDK12:44
HeikoS1600 euros12:44
@iglesiasgI am assuming you get money from Google to do it12:45
HeikoS1yeah we should get it back12:45
HeikoS1if not we have the gsoc money12:45
lisitsyniglesiasg: well with some limits I guess12:45
@iglesiasgaham12:45
lisitsynmine is out of the limit12:45
HeikoS1lisitsyn deserves to come with us12:45
lisitsynI think12:45
lisitsyn:)12:45
@iglesiasgsure12:45
@iglesiasgI think it is definitely a good way of using that money12:45
HeikoS1really looking forward! :)12:45
lisitsynHeikoS1: once I get visa I buy that lufthansa that arrives to sfo 12:2012:45
HeikoS1ok12:46
HeikoS1lisitsyn: 6 sept right?12:46
lisitsynHeikoS1: and once they have all the penetration procedures done we meet12:46
lisitsyn:D12:46
HeikoS1fingers crossed!12:46
HeikoS1haha deep penetration12:46
lisitsynHeikoS1: yes 6 sept12:46
HeikoS1ok, doing some work now, see you in an hour12:47
lisitsynHeikoS1: see you12:47
shogun-buildbotbuild #1568 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1568  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>12:50
lisitsynohhh wasabi-chan has died http://icdn.lenta.ru/images/2013/08/30/13/20130830132651290/pic_df31d3d81263af49357f182a81b1c811.jpg12:52
thoralflisitsyn: wtf? ;)12:59
lisitsynthoralf: he was attacked by a raven so that cute thing is for recover :(13:00
@iglesiasgnoooooo13:14
@iglesiasgI just download Matlab source code for a metric learning method13:15
@iglesiasgand the readme is a .doc13:15
@iglesiasgfaith in humanity completely lost13:15
HeikoS1haha13:17
HeikoS1dont use matlab13:17
HeikoS1its evil13:17
HeikoS1:D13:17
@iglesiasgyeah.. but people use it so...13:17
@iglesiasgHeikoS1, I remember you were a pro-matlaber not so long ago though :P13:17
HeikoS1iglesiasg: I am still;)13:19
HeikoS1switched to python though13:19
HeikoS1but still massively use it13:19
HeikoS1unfortunately13:19
HeikoS1lack of proper shogun support still bugs me13:20
@iglesiasgI use it too quite a bit13:20
@iglesiasgit is where I get stuff working fastest13:20
@iglesiasgbut steadily my Python skills are improving13:21
HeikoS1iglesiasg: same here13:24
HeikoS1notebooks are great to prototype13:24
HeikoS1using them a lot13:24
shogun-buildbotbuild #1352 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1352  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>13:25
-!- az_de [82954e22@gateway/web/freenode/ip.130.149.78.34] has joined #shogun13:48
-!- Cheng [~Cheng@pa49-180-112-58.pa.nsw.optusnet.com.au] has joined #shogun13:50
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun13:51
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has joined #shogun13:52
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Ping timeout: 245 seconds]13:53
HeikoS1Meeting in 513:55
HeikoS1sonne|work: around?13:55
shogun-buildbotbuild #1652 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1652  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>13:56
HeikoS1wiking: are you there?13:56
HeikoS1who is still missing? van51, roman ..13:58
-!- gsomix_ is now known as gsomix13:59
gsomixhey13:59
HeikoS1hi!14:00
HeikoS1still waiting for some people14:00
HeikoS1maybe 5 more misn14:00
-!- votjakovr [2ef103d1@gateway/web/freenode/ip.46.241.3.209] has joined #shogun14:01
HeikoS1votjakovr: hi!14:01
votjakovrHeikoS1: hi!14:02
HeikoS1Ok, I think we are almost complete14:02
HeikoS1lets go :) everyone there and ready?14:03
-!- van51 [~van51@ppp-94-66-14-37.home.otenet.gr] has joined #shogun14:03
HeikoS1van51: hi!14:03
van51hello14:03
@iglesiasgready! :)14:03
HeikoS1ok then14:03
HeikoS1Welcome to the last Shogun GSoC meeting before the "pencils down"14:03
HeikoS1deadline, September 16.14:03
hushellHi~14:04
HeikoS1We will have one more post GSoC meeting in between then and September14:04
HeikoS123, which is the evaluation deadline, to celebrate, and to give14:04
HeikoS1details about the final evaluations (which are quite similar to14:04
HeikoS1mid-term)14:04
HeikoS1There are a little less than 4 weeks left. So everyone should be about14:04
HeikoS1to finish the coding parts of their project.14:04
HeikoS1The last 1-2 weeks are for documentation, polishing, reviewing (will soon explain), not for coding new stuff!14:04
HeikoS1This means coding has to be done in 2-3 weeks. Please be aware of that14:04
HeikoS1The topics for today are14:04
HeikoS11.) IPython notebooks.14:05
HeikoS12.) Examples/Web-demos14:05
HeikoS13.) Student Peer-reviewing of project documentation.14:05
HeikoS14) Project updates14:05
HeikoS1Any other suggestions?14:05
HeikoS1lisitsyn1, iglesiasg, sonne|work ?14:05
HeikoS1ok then14:06
HeikoS11.)  IPython notebooks.14:06
HeikoS1Again on the purpose: The notebooks should be a documentation of your complete GSoC project.14:06
HeikoS1So don't just do simple toy examples, but rather describe all classes that you implemented, when they are applicable, etc.14:06
HeikoS1Give a high-level overview of the classes you implemented (you know this, but others don't)14:06
HeikoS1Give an example for every method you implemented if possible. Include non-toy datasets.14:06
HeikoS1Include many plots to make them look cool.14:06
HeikoS1Give references to the papers that you implemented.14:06
HeikoS1The notebooks will be included in the upcoming Shogun 3.0 release (on our website), so your projects will get a lot of attention.14:06
HeikoS1Your notebooks will be the main starting point for people to try out your stuff, so make them as nice as possible.14:07
HeikoS1pickle27 and iglesiasg already have started on this, others please join14:07
@iglesiasg(HeikoS1 nothing to add, sorry the delay)14:07
-!- zeller_ [~zeller@embln.embl.de] has joined #shogun14:07
HeikoS1Some more inspiration: http://scikit-learn.org14:07
@iglesiasgzeller_, hi Georg! :)14:07
HeikoS1please start rather earlier than later on this process, as it takes some time14:08
HeikoS1On a technical side: Please submit the notebooks without result cells, as we will render them automagically14:08
pickle27HeikoS1: how do we do that?14:08
HeikoS1iglesiasg, pickle27 could you maybe post links to the current version of yours?14:08
@iglesiasgHeikoS1, question!14:08
HeikoS1pickle27: you can remove all result cells in the editor14:09
@iglesiasgsure, there you go http://nbviewer.ipython.org/633421414:09
HeikoS1iglesiasg: question14:09
@iglesiasgthere is a couple of typos in the last part, I will fix them14:09
@iglesiasgHeikoS1, so for the version we submit, we don't include the figures?14:09
@iglesiasgnot sure if figures are included as result cells14:09
HeikoS1pickle27: clear al14:10
HeikoS1exactly, just do clear14:10
HeikoS1cell->all output->clear14:10
HeikoS1the graphics are stored in binary format and we do not want that in the git14:10
pickle27http://nbviewer.ipython.org/urls/raw.github.com/shogun-toolbox/shogun/develop/doc/ipython-notebooks/ica/bss_jade.ipynb14:10
@iglesiasgoh nice, thanks for the clear shortcut14:11
HeikoS1We have a mechanism that renders the current git version every night and uploads the result (with plots) to our website from where people can view them14:11
pickle27HeikoS1: okay that makes sense, I was wondering about that14:11
HeikoS1iglesiasg: will sool provide a page on the website for that14:11
@iglesiasgyep, priority once GSoC period is over14:11
HeikoS1sure14:11
HeikoS1ok, again to emphasize: your whole projects with all its subtleties should be documented in one notebook14:12
HeikoS1any other questions?14:12
HeikoS1hushell, gsomix, van51 everything clear?14:12
HeikoS1votjakovr: ^ ?14:12
gsomixHeikoS, yep14:13
van51HeikoS1: so far yeah14:13
HeikoS1cool!14:13
HeikoS12.) Next point: Examples/Web-demos14:13
votjakovrHeikoS1: i've got one: many parts of my project are depends on last year project, should i document them?14:13
HeikoS1votjakovr: focus on your gsoc project in the norebook, and when thats done, it would be cool if you get a bit into the regression too14:14
votjakovrHeikoS1: ok14:14
HeikoS1votjakovr: as you fixed tons of bugs in there, it only works cause of your gsoc project :)14:14
HeikoS1ok then, examples/web-demos14:14
HeikoS1As said before, the most common use cases of your methods should be illustrated in a c++ example and at least one modular interface (python)14:14
HeikoS1so at least 1-3 examples14:15
HeikoS1If you want to do more, we greatly appreciate that.14:15
HeikoS1like pickle27, who did some static examples (and interfaces)14:15
HeikoS1keep in mind for the modular interfaces, you need to define interface files, start with that soon14:15
HeikoS1Now that was said before14:16
HeikoS1As some of you may have noticed, we have a growing web-demo framework, written by foulwall.14:16
HeikoS1foullwall could you share a link=?14:16
HeikoS1Everyone, so if you want to, and have time, include and example of your method in there.14:17
HeikoS1We think this is very cool for presenting Shogun. As for example for the NIPS workshop where we want to send things to14:17
HeikoS1Talk to S?ren and foulwall on how to do this.14:17
HeikoS1Note this is optional, but we would love to see more demos on the new projects14:17
HeikoS1foulwall: link?14:17
foulwallhttp://nn.7nn.de:8000/14:18
HeikoS1foulwall, as you don't have to write a notebook, please write a good documentation on how to use the web-framework.14:19
HeikoS1It should be easy for anyone to extend/add demos after reading it14:19
zeller_the link doesn't seem to work for me...14:19
foulwallOk HeikoS1 ,14:19
@iglesiasgzeller_, yeah, sorry about that. You have to complete the url with one of the examples you get listed in the 404 page14:20
@iglesiasgzeller_, for instance, http://nn.7nn.de:8000/clustering/entrance14:20
HeikoS1zeller_: try for example this one14:20
HeikoS1http://nn.7nn.de:8000/classification/binary/entrance14:20
zeller_ah alright thx14:20
foulwallzeller_: sorry I didn't add a entrance page14:20
HeikoS1foulwall: we plan to integrate this to our webpage at some point, s?ren will update you on that14:20
HeikoS1if we had the new gsoc methods in the demos, that would be awesome14:21
foulwallok HeikoS114:21
@iglesiasgfoulwall, the demos look really fancy :)14:21
foulwallhaha iglesiasg14:21
HeikoS1I like them a lot, too :)14:21
HeikoS1Questions on that?14:21
HeikoS1Ok then to something new this year's GSoC14:22
HeikoS13.) Student Peer-reviewing of project documentation.14:22
HeikoS1We think it is a good idea that someone who is not familiar with your projects takes a look.14:22
HeikoS1This is why we ask you to pair up and to review each others project a bit.14:22
HeikoS1We hope to get some synergistic effects from this, thats what open-source is about after all :)14:22
HeikoS1Note that we do not ask you to review code (although feel free to do so), but rather documentation and usability.14:22
HeikoS1Such as14:22
lisitsyn1oops guys sorry14:22
HeikoS1-Class documentation14:22
HeikoS1Are all methods documented with doxygen?14:22
HeikoS1Are all classes properly documented in @brief?14:22
HeikoS1How is the framework API used, what things can be combined?14:22
HeikoS1-IPython notebooks14:23
HeikoS1Try to understand things, and if not, tell each other what's not clear.14:23
HeikoS1Suggest extensions14:23
HeikoS1etc14:23
HeikoS1lisitsyn1: hello there ;)14:23
lisitsyn1hey14:23
HeikoS1Again,  for the web-demo framework of foulwall, there is also documentation/usability to review.14:23
HeikoS1Here is a random list of student pairs:14:24
HeikoS1van51 - gsomix14:24
HeikoS1hushell - lambday14:24
HeikoS1pickle27 - votjakovr14:24
HeikoS1iglesias - foulwall14:24
HeikoS1Please get in touch, ask for lists of classes and the notebooks and start giving each other feedback14:24
gsomixvan51, hey, maaaan14:24
HeikoS1haha ;)14:24
van51gsomix: yo :)14:24
HeikoS1It is also probably more fun sharing things than just to do everything on your own :)14:25
@iglesiasgfoulwall, demos for me! :)14:25
hushellI'll send an email to lambday :) seems he is not here14:25
HeikoS1iglesiasg: I tricked the random generator there since you will do the website stuff ;)14:25
HeikoS1hushell: yes please do, he could not be here today14:25
foulwalliglesiasg: demo for you:)14:25
HeikoS1pickle27, votjakovr you are also fine with that?14:26
votjakovrHeikoS1: sure :)14:26
pickle27HeikoS1: yeah sounds good!14:26
HeikoS1cool!14:26
pickle27votjakovr: I'll send you mail later today14:26
HeikoS1Everyone, try to start with that slowly and plan for some more intense reviewing the last 2 weeks of gsoc14:26
HeikoS1Questions/Suggestions?14:27
votjakovrpickle27: ok14:27
HeikoS1Sweet, so the last point for today is Project updates14:27
Chenghow about trying to get a user who is not a developer to check?14:27
thoralfHeikoS1: Do you know if require() will be optimized out when compiling without debug stuff?14:27
HeikoS1Cheng: thats a good idea14:28
@iglesiasgCheng, I think that is a very good idea too14:28
HeikoS1what would be the best way to do this.14:28
@iglesiasgHeikoS1, first we can try in the mailing list if there are volunteers14:28
HeikoS1Maybe every student sends a mail with a description of his project/notebook to the mailing list and asks?14:28
@iglesiasgotherwise, when we get a question in the mailing list, we say: to get an answer, please help us before reviewing ... :P14:29
HeikoS1haha :P14:29
@iglesiasgjust kidding14:29
HeikoS1thoralf: you can probably do one or two ;)14:29
HeikoS1thoralf: no reqiure stays in14:29
@iglesiasgwe can also ask friends who are not using Shogun, I guess we all one from school people interested in machine learning14:30
@iglesiasgone -> know14:30
HeikoS1iglesiasg: good idea14:30
pickle27iglesiasg: thats a good idea14:30
thoralfHeikoS1: Okay, since you asked me to replace assert() by require(), this will introduce some performance penalties.14:30
HeikoS1thoralf:  assert also stays in (I think at least), lets discuss later14:31
thoralfHeikoS1: assert() usually depends on -DNDEBUG14:31
HeikoS1Best thing would be to announce the finished notebooks somewhere and then people can have a look14:31
HeikoS1students, please also push your mentors you review your notebooks14:31
HeikoS1Cheng: thanks for the idea, we will think about it a bit more :)14:32
HeikoS1ok lets press on14:32
HeikoS14.) Project updates.14:32
HeikoS1Please every student/mentor (preferably mentor since they haven't talked at all) give a brief update on the project.14:32
HeikoS1In particular, where in the schedule you are, and whether its feasible to finish your project within the next 2-3 weeks.14:33
HeikoS1I dont know all the mentor nicknames, so using students14:33
HeikoS1van51 want to start?14:33
van51sure14:33
van51So when I started I was supposed to implement two GSoC ideas from your page14:33
HeikoS1other please prepare texts :)14:33
van51namely the hashing trick and the random fourier features, along with random kitchen sinks and fast food14:34
van51the first project (the hashing trick) is pretty much done14:34
van51and also suppors quadratic features14:34
van51what remains for that part are some modular examples, a web demo and the python notebook14:34
van51from the second part I am atm implementing random fourier features14:35
HeikoS1van51: how are the benchmarks going?14:35
van51HeikoS1: pretty good so far14:35
van51HeikoS1: will include all results in the notebook14:35
HeikoS1van51: looking forward to see that!14:35
HeikoS1van51: yeah mention the benchmarks in the notebook, then people can have a look14:36
-!- Cheng [~Cheng@pa49-180-112-58.pa.nsw.optusnet.com.au] has quit [Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org]14:36
van51I believe it is feasible to complete everything else in 2-3 weeks14:36
van51but in case I see that it won't be good work14:36
van51I will do as much as possible14:36
-!- Cheng [~Cheng@pa49-180-112-58.pa.nsw.optusnet.com.au] has joined #shogun14:36
van51and continue after GSoC the rest14:36
HeikoS1van51: try to pick most important things and make them work nice,14:36
HeikoS1thats better than having lots of things which dont work well14:37
van51HeikoS1: yeah exactly :)14:37
HeikoS1van51: and keep in mind, last 1-2 weeks is documenting ;)14:37
HeikoS1van51: you are very welcome to stay after gsoc, everyone is in fact! but we will talk about this another time14:37
@iglesiasgHeikoS1, sorry, I am not sure about one thing14:37
HeikoS1iglesiasg: ?14:38
@iglesiasglast 1-2 weeks or documenting period includes also time to finish the notebook?14:38
@iglesiasgor it should be already finished by then?14:38
@iglesiasgsorry if you already mentioned it, and I missed it14:38
HeikoS1iglesiasg: yes, notebook, class dox, maybe finishing/tuning examples14:38
HeikoS1iglesiasg: the point is: no more coding of algorithms or so14:38
@iglesiasgHeikoS1, all right, got it. Thanks!14:38
HeikoS1iglesiasg: thanks for pointing that out!14:39
HeikoS1ok next mentor/student pair14:39
HeikoS1gsomix14:39
gsomixOk.14:39
HeikoS1(and then  iglesias, pickle27, votjakovr, hushell, lambday, foulwall)14:39
gsomixSo I finished most of CSV and LibSVM readers.14:40
gsomixNeed to prepare PR with examples and to refactor old IO code.14:40
HeikoS1gsomix: did you test them on some files? Already usable?14:40
gsomixHeikoS, yep. Soere helps me with csv examples, now it tested on shogun's train data.14:41
gsomix*Soeren14:41
gsomix*it is tested14:41
HeikoS1cool, looking forward to see the examples, in fact I will use them soon :)14:41
HeikoS1gsomix: how is the project status apart from that?14:41
gsomixHm, I should add support of protobuf file format.14:42
gsomixAnd MATLAB file format.14:42
shogun-notifier-shogun: Heiko Strathmann :develop * f4368f0 / src/interfaces/modular/ (4 files): https://github.com/shogun-toolbox/shogun/commit/f4368f03772e3425f8ba239842e0dee4795e1bd714:42
shogun-notifier-shogun: added log-det framework (partly) to modular interfaces14:42
shogun-notifier-shogun: Heiko Strathmann :develop * 832dd29 / src/interfaces/modular/ (4 files): https://github.com/shogun-toolbox/shogun/commit/832dd2952282b187c4ab4d1d6fe6d842aa6d0a4614:42
shogun-notifier-shogun: Merge pull request #1496 from karlnapf/develop14:42
shogun-notifier-shogun:14:42
shogun-notifier-shogun: added log-det framework (partly) to modular interfaces14:42
HeikoS1gsomix: will that work in the next weeks?14:42
HeikoS1sorry for the merge, travis just went green14:43
gsomixYep. I hope I can implement it before "pencils down".14:43
gsomixotherwise I will stay after GSoC with this issues.14:43
gsomixI also plan improve parsing and make it more fast.14:43
HeikoS1gsomix: ok, cool, also keep in mind that documenting them is very important, so a notebook with examples will be very helpful14:44
gsomixOK14:44
HeikoS1gsomix: same for you, please dont code until the last minute, rather pick things and make them waterproof14:44
HeikoS1much more useful14:44
HeikoS1gsomix: thanks! continue?14:45
az_deok, I will do this14:45
az_depickle27 implemented all planned ICA /AJD algorithms and is currently doing an extra implementation of FastICA for comparison.14:45
HeikoS1az_de: go ahead !14:45
az_deThere are several examples and ipython notebooks on toy data, on cool audio data14:45
az_deand an application to fetal ECG, where the goal is to separate the heart beat signals of the mother and the baby.14:46
HeikoS1yep, I liked them :)14:46
az_deAll of those work essentially and just need some polishing and further documentation /testing...14:46
az_deIn summary, very good work!14:46
HeikoS1az_de, pickle27 I havent seen the baby one yet, is it public?14:46
pickle27az_de: thanks!14:46
pickle27HeikoS1: I haven14:46
pickle27't pushed it yet, Im still playing with it14:46
HeikoS1pickle27: ok, looking forward to see this14:47
HeikoS1pickle27: you added some static interfaces right?14:47
pickle27HeikoS1: yes the ICA algs are available from the static interface14:47
pickle27I haven't testes python static but it works in Octave and R14:47
HeikoS1pickle27: maybe you can also assist others in doing some small bits, since its very cool to have things available in R and Matlab until we find a solution to modular14:48
HeikoS1if someone wants to do static interfaces, pickle27 is your man :)14:48
pickle27HeikoS1: yes certainly I can lend a hand14:48
@iglesiasgI would like to make LMNN accessible in static too, so I will come to you pickle2714:48
HeikoS1az_de, pickle27 thanks! continue?14:48
HeikoS1iglesiasg: next?14:48
pickle27iglesiasg: sounds good14:48
zeller_that'd be great14:48
zeller_alright14:48
zeller_Fernando's LMNN project is well on track: the main algorithm is in shogun, it seems to be correct (but a bit more testing on larger data sets will be done before the end of GSoC) and slightly faster than the reference implementation in Matlab (if time permits additional optimization could be done, I guess).14:48
zeller_On top of that he has already implemented a variant that learns a diagonal metric transformation (in contrast to the full LMNN, it essentially only rescales feature coords, without rotation), and we will discuss another L1-regularized version of this later today, I hope it's feasible to include this too during the next weeks.14:49
HeikoS1zeller_, iglesiasg you should add the speed tests to benchmarks/14:49
shogun-buildbotbuild #1937 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1937  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:49
zeller_wher would you add these?14:50
@iglesiasgHeikoS1, I think I have never seen these, I will check14:50
HeikoS1zeller_, iglesiasg see van51 recent additions there14:50
HeikoS1https://github.com/shogun-toolbox/shogun/pull/1466#discussion_r608261814:50
@iglesiasgyeah, I just found the dir14:50
zeller_looks good14:50
HeikoS1cool14:51
@iglesiasgdidn't know about this, I will use it as reference14:51
HeikoS1what about the real-data for the notebook?14:51
@iglesiasgvan51, thank you for this!14:51
zeller_I hope there's a bit of time left for comparing a few multiclass classifiers to LMNN14:51
@iglesiasgyes14:51
HeikoS1zeller_: yes that would be cool14:51
van51iglesiasg: hehe np14:51
zeller_I'm working on compiling a few bioinformatics data sets maybe also for the notebook14:52
@iglesiasgregarding real data, so far I have used the wine data set and a portion of MNIST14:52
HeikoS1ah thats nice14:52
HeikoS1zeller_, iglesiasg finished?14:52
zeller_yes14:52
@iglesiasgyes, I think that is pretty much it14:52
HeikoS1ok then, votjakovr please go on14:52
votjakovrHeikoS1: ok14:53
votjakovrThis GSoC i've finished Laplacian and EP binary GP classifiers.14:53
votjakovrAlso i added some numerical integration stuff (it probably can be used by some other parts of shogun).14:53
shogun-buildbotbuild #1569 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1569  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:53
votjakovrAlso i've done many fixes and improvements of existing regression part of GP framework.14:53
votjakovrNow i'm fixing gradient model selection, since it's very important for GP classification/regression.14:54
HeikoS1If time permits, we also want to add multiclass GP methods (starting with Laplace approximation)14:55
HeikoS1for the notebook, I am currently scanning a few bio classification datasets14:55
HeikoS1romand has already written a raphical python example which will probably also be the basis of a notebook14:55
HeikoS1Roman, sorry14:55
votjakovrHeikoS1: yeah14:56
HeikoS1and a possible extension is a variational multiclass GP classifier14:56
votjakovrHeikoS1: but it seems like after GSoC14:57
HeikoS1votjakovr: yep, I meant that :)14:57
HeikoS1One problem was that last year's GP implementation was quite buggy and was refactored significantly14:57
HeikoS1but now everything is tested and much more stable14:58
HeikoS1ok, that's about it for us, hushell could you continue?14:58
hushellokay14:58
hushellFew words for reminder. My project is about large scale general strucured output models. So my work mainly focuses on structured prediction and online SOSVM solvers.14:58
hushellI have done: 0) refine current SOSVM framework in Shogun. 1) Factor Graph framework. 2) MAP inference framework. 3) max product for tree graph. 4) learning stuffs of factor graph model (loss-augmented decoding etc.) This is the PR what I am working on.14:58
hushellTODO list: 1) OCR demo (with ipython notebook) 2) stochastic subgradient descent SOSVM solver. 3) block coordinate Franke-Wolfe SOSVM solver14:59
hushellI want to make the demo works first, then move on implementing more solvers. I am a bit lagging behind the schedule, but I think at least first 2 items in the TODO list will be finished on time. I'll make all the things left after GSOC.14:59
hushellAfter GSOC: 1) refine SO learning code. 2) more MAP inference methods, such as graphcuts, TRW-S, LPR etc. 3) pictorial structure demo.14:59
HeikoS1hushell: wow thats a lot :)15:00
hushellHeikoS1: I got a lot of time to prepare :)15:00
HeikoS1hushell: I agree with you, rather finish things with demos/tests before starting new stuff15:00
HeikoS1hushell: for your stuff, examples will be quite important to understand whats going on as its not so common :)15:01
HeikoS1then, finally, foulwall, please go ahead15:01
hushellyes, I'll speed up myself to get the demo working15:01
@iglesiasghushell, one question about the OCR demo15:01
@iglesiasghushell, will you be using the FactorGraphModel for that? I have seen OCR before done with HMM, so I wondered if you planned to use factor graph or the hm-svm15:02
hushelliglesiasg: I am going to use the tree max-product to do the inference15:02
@iglesiasghushell, all right!15:03
hushelland compare the results with different solvers, also with hm-svm15:03
@iglesiasgit sounds good15:03
lisitsyn1is that enough time15:03
lisitsyn1to finish things?15:03
hushellPatrick suggests this is easier than other stuff, he has an OCR demo in his code as well15:03
lisitsyn1alright15:04
HeikoS1hushell: try not to leave things unfinished, if you feel that its going to be too much, drop it for gsoc15:04
HeikoS1always can continue afterwards15:04
-!- zeller_ is now known as georg_zeller15:04
HeikoS1Everyone, please keep in mind to stop coding in 2-3 weeks15:05
HeikoS1ok, foulwall please give us an update for your stuff15:05
foulwallok HeikoS115:05
hushellHeikoS1: Sure. I'll make a waterproof gsoc, rather than crappy one ;)15:05
foulwallup to now , I almost finished the idea  'implementing interactive graphical demos'. Only arts/asp demos left, I can finish this week.15:06
sonne|workfoulwall: (and multiclass)15:06
foulwallsonne|work: :) That's already merged, pr sent soon.15:06
foulwallNow I'm working on simple mldata h5 downloader for shogun(debugging).15:07
foulwallI'm happy to help you with the web-based examples in next weeks.(just give me ipynb and I'll merge)15:07
HeikoS1foulwall, sonne|work would be great to have a tutorial how to extend the web things15:07
sonne|workHeikoS1: foulwall already wrote a README15:08
pickle27foulwall: I might be interesting in making a web demo15:08
HeikoS1sonne|work: thats good, please include it in the peer-reviewing then, iglesiasg15:08
@iglesiasgok15:08
pickle27foulwall: I'll send you a mail later and we can discuss15:08
foulwallOk pickle27,15:08
foulwallI believe it's feasible to complete the demos and downloader before pencil down15:09
@iglesiasgpickle27, CC the mailing list if you don't mind please :)15:09
HeikoS1votjakovr: the GPs should def. also be in the web-demos15:09
sonne|workfoulwall: van51 will also need your help15:09
pickle27iglesiasg: sure!15:09
@iglesiasgpickle27, I want to see how much effort that might take15:09
sonne|workfoulwall: so I guess will get some questions in the following weeks on IRC15:09
pickle27iglesiasg: haha me too lol15:09
HeikoS1votjakovr: there is already some basic stuff for gp regression, would be nice to have other cool things15:09
foulwallok:) I'll keep online15:09
HeikoS1foulwall: thanks!15:10
HeikoS1I'll close with a brief update for lambday as he could not make it15:10
foulwall:) HeikoS115:10
votjakovrHeikoS1: ok, i'll add :)15:10
HeikoS1He recently merged the final missing piece in his big framework15:10
HeikoS1Now is the time for testing as these estimators suffer numerical problems sometimes15:10
HeikoS1There is quite some stuff regarding linear solvers, eigen solvers, linear operators which will be accessible from shogun15:11
HeikoS1A big example with a GMRF on worldwide ozone data is waiting to be tested, too15:11
HeikoS1Since I do use these methods in research, I will also pass on this stuff to lambday15:12
HeikoS1In addition, he wrote this IndependentComputation framework15:12
HeikoS1Which is a general interface for parallel computation on multicore/cluster15:12
HeikoS1I am currently running a PBS version of this and it works fine15:12
HeikoS1So we might add that to Shogun after GSoC (hopefully)15:12
HeikoS1That's it15:13
HeikoS1Any remarks or final points?15:13
sonne|worknice15:13
sonne|workI have one remark if there is no further comment15:13
HeikoS1please go ahead15:13
sonne|workPlease fix warnings also on the buildbots e.g. http://shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1654/steps/compile%20python_modular/logs/warnings%20%2839%2915:13
sonne|workyou can see them all here http://shogun-toolbox.org/buildbot/waterfall15:14
sonne|worklets try to get at least python modular down to 0 warnings15:14
HeikoS1sonne|work: good point! I hope this will be massively reduced after the dox peer reviewing15:14
HeikoS1To explain, if you find your class in these warnings, you forgot to document member variables, functions15:14
HeikoS1Or you got a name clash with a modular language keyword15:14
sonne|worketc15:15
HeikoS1Students, BTW we will be releasing Shogun 3.0 shortly after gsoc, so if you want to help on that, let us know15:15
HeikoS1see the issue list for milestone 3.0 on github15:15
HeikoS1If there are not further remarks?15:16
lisitsyn1are we sure we need 3.0?15:16
sonne|workwell please help with that15:16
sonne|workthe 3.0 release I mean15:16
HeikoS1lisitsyn1: dont know, we should discuss if you dont agree I guess :) but many new features15:16
HeikoS1The meeting is over! Thanks for attending, I will send a summary to the mailing list15:16
sonne|workif you do all your work will be included and the next release will be massive15:17
lisitsyn1okay I don't agree ;) I don't think it is 3.015:17
sonne|workand due to the demos / notebooks etc a lot of people will look at it15:17
sonne|workand yes please stick with us if you had fun15:17
HeikoS1yep, we will get quite some attention I guess :)15:17
sonne|workwe definitely need you in the team :)15:17
HeikoS1sonne|work: we should have another "recruiting" meeting after gsoc and celebrate ;)15:18
sonne|workWell I guess we can invite people who stay to the next shogun workshop in berlin or so :)15:18
sonne|workbut lets first finish gsoc successfully15:19
hushellI will be happy to keep staying at Shogun :)15:19
sonne|workeveryone thanks for your excellent work15:19
sonne|workwe have come a long way this gsoc again15:19
sonne|workalright I am done :)15:19
pickle27yeah I plan on sticking around too! (although I'll be mia for a bit I'll be back in Dec!)15:20
sonne|workhushell, pickle27 would be very nice15:20
HeikoS1hushell, pickle27 would be very cool if you stayed!15:20
sonne|workand of course van51, votjakovr, foulwall, gsomix you are most welcome too15:20
HeikoS1sonne|work: and lambday :)15:21
HeikoS1ok I am getting lunch :) see you15:21
-!- az_de [82954e22@gateway/web/freenode/ip.130.149.78.34] has quit [Quit: bye, bye!]15:21
-!- Cheng [~Cheng@pa49-180-112-58.pa.nsw.optusnet.com.au] has quit [Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org]15:21
HeikoS1sonne|work: do you have an idea about https://github.com/shogun-toolbox/shogun/issues/149915:21
hushellokay, I go back to sleep, CU guys15:21
HeikoS1hushell: bye!15:21
sonne|workHeikoS1: yes sure (lambday)15:22
sonne|workbye everyone and hushell, foulwall have a nice sleep15:22
gsomixsonne|work, OK :)15:23
pickle27sonne|work: I have another swig question for you - I've been playing with the Ruby modular stuff, how does swig determine which "in" typemap is used15:23
pickle27sonne|work: like how would it decide between these two: TYPEMAP_SGMATRIX(float32_t, NUM2DBL, rb_float_new) and TYPEMAP_SGMATRIX(float64_t, NUM2DBL, rb_float_new)15:24
pickle27is it decided based on the required shogun type?15:24
sonne|workpickle27: there is a typecheck typemap to determine which one to use15:25
-!- georg_zeller_ [~zeller@embln.embl.de] has joined #shogun15:25
-!- georg_zeller_ [~zeller@embln.embl.de] has quit [Client Quit]15:27
pickle27sonne|work: and that works on the required shogun type?15:27
-!- georg_zeller_ [~zeller@embln.embl.de] has joined #shogun15:27
-!- georg_zeller_ [~zeller@embln.embl.de] has quit [Client Quit]15:27
pickle27like say I pass a numpy array of float32 to RealFeatures which uses float64 internally the typemap that gets called is for SGTYPE=float64_t right?15:28
-!- georg_zeller [~zeller@embln.embl.de] has quit [Ping timeout: 260 seconds]15:29
pickle27so my typemap would need to grab float32 from the pyobj and cast them to float64 for the shogun stucture15:29
shogun-buildbotbuild #1353 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1353  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>15:31
sonne|workpickle27: no it works on the e.g. ruby objects and if an object fits  - which you signal by returning 1 - the respective in typemap is used15:34
sonne|workpickle27: it cannot be done on an OUT typemap15:35
sonne|workpickle27: for OUT you would need to write simply another function returning another datastructure to distinguish things15:35
-!- van51 [~van51@ppp-94-66-14-37.home.otenet.gr] has quit [Quit: Leaving.]15:37
pickle27sonne|work: okay I see that now but how does it decide which version of the typemap?15:37
-!- travis-ci [~travis-ci@ec2-54-234-78-43.compute-1.amazonaws.com] has joined #shogun15:40
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1079772515:40
-!- travis-ci [~travis-ci@ec2-54-234-78-43.compute-1.amazonaws.com] has left #shogun []15:40
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has left #shogun []15:41
sonne|workpickle27: based on the typecheck15:43
sonne|workpickle27: you write one typecheck typemap per type to check15:43
sonne|workand you return either 0 or 115:43
sonne|workIIRC you can also prioritize15:43
pickle27sonne|work: The problem I am having is understanding the type of the Ruby NMatrix so that I can cast its data properly and iterate through it with pointer arithmetic (the NMatrix c api is pretty rough), I'll post a paste in 2 secs and explain the problem, Im trying to figure out if its an NMatrix api problem or a me not understanding swig well enough problem15:45
pickle27http://pastebin.com/rrtp7PBp15:46
pickle27so my typecheck happens in line 415:46
pickle27then in line 39 my NMatrix in map starts15:46
pickle27DENSE_STORAGE->elements is a void* so I need to cast it15:46
pickle27and my instincts tell me that I shouldn't need the switch case because that should be handled by which typemap is used15:47
pickle27but if I cast s->elements with SGTYPE it doesn't work because the Ruby array is 32 and its being cast to 6415:48
-!- votjakovr [2ef103d1@gateway/web/freenode/ip.46.241.3.209] has left #shogun []15:51
sonne|workdo some prints in the typcheck and in  typemap to see if this is ever reached15:55
shogun-buildbotbuild #1654 of deb3 - modular_interfaces is complete: Failure [failed test ruby_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1654  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>15:55
-!- van51 [~van51@ppp-94-66-14-37.home.otenet.gr] has joined #shogun15:57
-!- van51 [~van51@ppp-94-66-14-37.home.otenet.gr] has quit [Ping timeout: 240 seconds]16:01
sonne|workpickle27: ^16:12
shogun-notifier-shogun: Heiko Strathmann :develop * ebb6b6c / src/shogun/lib/computation/job/ (4 files): https://github.com/shogun-toolbox/shogun/commit/ebb6b6c2e7510c096acdd751f8ba1b926b21802016:24
shogun-notifier-shogun: moved more template instanciations from .h to .cpp16:24
shogun-notifier-shogun: Heiko Strathmann :develop * c58d6ff / src/shogun/lib/computation/job/ (4 files): https://github.com/shogun-toolbox/shogun/commit/c58d6ff2a10b66d12be94a456fe971bf37d1d9a316:24
shogun-notifier-shogun: Merge pull request #1500 from karlnapf/develop16:24
shogun-notifier-shogun:16:24
shogun-notifier-shogun: moved more template instanciations from .h to .cpp16:24
HeikoS1pickle27: https://travis-ci.org/shogun-toolbox/shogun/jobs/1079772616:25
HeikoS1one of your unit tests fails16:25
shogun-buildbotbuild #1540 of deb2 - static_interfaces is complete: Failure [failed compile libshogun]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1540  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>16:32
shogun-buildbotbuild #1570 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1570  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>16:34
-!- HeikoS1 [~heiko@nat-183-166.internal.eduroam.ucl.ac.uk] has left #shogun []16:43
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has quit [Quit: Leaving.]16:44
-!- FSCV [~FSCV@187.191.50.146] has joined #shogun16:49
-!- FSCV_ [~FSCV@50.7.50.60] has joined #shogun17:04
foulwallls17:04
foulwallping sonney2k17:05
shogun-buildbotbuild #1655 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/165517:05
-!- FSCV [~FSCV@187.191.50.146] has quit [Ping timeout: 260 seconds]17:07
shogun-notifier-shogun: Soeren Sonnenburg :develop * 789d48f / INSTALL: https://github.com/shogun-toolbox/shogun/commit/789d48f557a287c8ec80815c67a5a6aab8bfee5f17:09
shogun-notifier-shogun: improve install - mention how to specify additional library/include paths17:09
shogun-buildbotbuild #1354 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1354  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>17:12
-!- sonne|osx [~sonne@89.204.130.49] has joined #shogun17:14
-!- iglesiasg [~iglesias@2001:6b0:1:1041:49a7:bd41:6df2:b1e0] has quit [Quit: Ex-Chat]17:14
sonne|osxfoulwall: hmmhh I didn't see your PR for the multi class demo?17:16
shogun-buildbotbuild #1571 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1571  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>17:17
sonne|osxafk17:20
-!- sonne|osx [~sonne@89.204.130.49] has quit [Quit: sonne|osx]17:20
-!- van51 [~van51@athedsl-379804.home.otenet.gr] has joined #shogun17:21
-!- HeikoS [~androirc@90.204.242.124] has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]17:21
shogun-buildbotbuild #1572 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1572  blamelist: Soeren Sonnenburg <sonne@debian.org>17:21
shogun-buildbotbuild #1541 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/154117:21
-!- HeikoS [~androirc@90.204.242.124] has joined #shogun17:22
-!- mode/#shogun [+o HeikoS] by ChanServ17:22
shogun-buildbotbuild #1355 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1355  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>17:55
-!- HeikoS [~androirc@90.204.242.124] has quit [Ping timeout: 245 seconds]17:57
-!- travis-ci [~travis-ci@ec2-54-226-253-132.compute-1.amazonaws.com] has joined #shogun17:59
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1080201517:59
-!- travis-ci [~travis-ci@ec2-54-226-253-132.compute-1.amazonaws.com] has left #shogun []17:59
-!- HeikoS [~androirc@90.204.242.124] has joined #shogun18:04
-!- mode/#shogun [+o HeikoS] by ChanServ18:04
thoralfHeikoShi18:17
thoralfHeikoS: Got 10 minutes?18:17
thoralfsonney2k: You too?18:18
pickle27HeikoS: btw not sure why that test fails let me know if it continues to do so18:19
thoralfHeikoS, sonney2k: I just added some handy sanity checks to several places in the code using "REQUIRE()".18:20
thoralfBut it really slows down shogun, since it's not removed compiling with debug options.18:20
thoralf(Like assert when setting NDEBUG)18:20
thoralfI'd like to keep these checks, but it should be possible to disable them.  Otherwise shogun is not fun any more. ;)18:21
thoralfBut using the checks, developing gets more fun.18:22
thoralfDo you have any ideas?18:22
thoralfSomething like "SANITY", which gets disabled when NDEBUG is set.18:24
thoralfSince it's no real error handling but more a kind of parachute when developing. :)18:24
@HeikoSSorry thoralf I am in a car...18:24
@HeikoSBut I like the idea18:25
thoralfIf I introduce "SANITY" which is the same as REQUIRE, but can be disabled, would that be okay?18:27
thoralfOr shall we discuss that further?18:28
-!- sonne|osx [~sonne@f053036061.adsl.alicedsl.de] has joined #shogun18:30
thoralfHey S?ren.18:30
gsomixhttp://isocpp.org/files/papers/n3744.pdf18:30
@HeikoSThough one critique is that this should not be necessary.  If you cannot be sure what your code does you did something wrong.... and otherwise you assert or require. But require is mainly used to check wrong user input18:31
thoralfHeikoS: Just think of (implicit) bounds checking when using vec[i] (vec is of type SGVector).18:33
thoralfHeikoS: I did this and it breaks some unit-tests. ;)18:34
shogun-buildbotbuild #1356 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1356  blamelist: Soeren Sonnenburg <sonne@debian.org>18:37
thoralfgsomix: Cool.18:38
shogun-buildbotbuild #65 of osx1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/65  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>, Soeren Sonnenburg <sonne@debian.org>18:42
-!- travis-ci [~travis-ci@ec2-50-16-42-192.compute-1.amazonaws.com] has joined #shogun18:43
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1080388718:43
-!- travis-ci [~travis-ci@ec2-50-16-42-192.compute-1.amazonaws.com] has left #shogun []18:43
-!- HeikoS [~androirc@90.204.242.124] has quit [Ping timeout: 240 seconds]19:15
shogun-buildbotbuild #1656 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1656  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:32
@sonney2kthoralf, ASSERT is the same as REQUIRE in shogun20:04
@sonney2kthoralf, we don't remove ASSERTS with -NDEBUG20:05
@sonney2kthoralf, the way I usually do this when developing is to define some macro DEBUG_XXX which disables the debug output/assertions completely later on when defined empty20:06
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]20:09
-!- lisitsyn [~lisitsyn@109-226-90-47.clients.tlt.100megabit.ru] has joined #shogun20:13
thoralfsonney2k: Okay, thanks.20:58
thoralfsonney2k: I decided against a PR, not important any more. ;)21:02
thoralfBye.21:02
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!]21:02
-!- van51 [~van51@athedsl-379804.home.otenet.gr] has quit [Quit: Leaving.]21:06
-!- lambday [67157d36@gateway/web/freenode/ip.103.21.125.54] has joined #shogun21:15
lisitsynsonne|osx: the most convenient flight for me is21:18
lisitsynarrive 18 oct 12:2021:18
lisitsyndepart 24 oct 14:5521:18
sonne|osxlisitsyn: we go back one day later though21:34
sonne|osxoct 2521:34
sonne|osxso you would have to go back to the airport alone21:34
lisitsynsonne|osx: nah that's ok I guess21:34
lisitsynsonne|osx: well depends where we stay21:34
sonne|osxlisitsyn: well in the middle of nowhere :)21:35
lisitsynsonne|osx: the plan would be you just drop me at some bus stop21:36
lisitsyn;)21:36
sonne|osxsty like that must exist :)21:36
sonne|osxalright21:36
* sonne|osx off21:36
-!- lambday [67157d36@gateway/web/freenode/ip.103.21.125.54] has quit [Ping timeout: 250 seconds]22:03
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun22:25
-!- mode/#shogun [+o iglesiasg] by ChanServ22:25
-!- HeikoS [~androirc@dab-ell1-h-74-6.dab.02.net] has joined #shogun22:26
-!- mode/#shogun [+o HeikoS] by ChanServ22:26
-!- HeikoS [~androirc@dab-ell1-h-74-6.dab.02.net] has quit [Ping timeout: 264 seconds]22:39
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 256 seconds]22:40
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun22:42
-!- mode/#shogun [+o iglesiasg] by ChanServ22:42
--- Log closed Sat Aug 31 00:00:38 2013
« Thursday, 2013-08-29 Index Saturday, 2013-08-31 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!