IRC logs of #shogun for Sunday, 2011-08-28

--- Log opened Sun Aug 28 00:00:32 2011
-!- f-x` [~user@] has quit [Ping timeout: 260 seconds]00:07
shogun-buildbotbuild #134 of java_modular is complete: Success [build successful]  Build details are at
blackburnsonney2k: ^ fuck yeah00:16
blackburnruby is not working still00:31
-!- f-x [~user@] has joined #shogun00:49
shogun-buildbotbuild #135 of java_modular is complete: Failure [failed test_1]  Build details are at  blamelist: blackburn91@gmail.com01:00
-!- f-x [~user@] has quit [Remote host closed the connection]01:31
CIA-3shogun: Sergey Lisitsyn master * r417d6c9 / (14 files in 7 dirs): #ifndef WIN32 to #ifdef HAVE_LAPACK transition -
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]01:34
shogun-buildbotbuild #136 of java_modular is complete: Success [build successful]  Build details are at
-!- alesis-novik [~alesis@] has quit [Read error: Connection reset by peer]04:29
-!- alesis-novik [~alesis@] has joined #shogun04:31
CIA-3shogun: Soeren Sonnenburg master * r5ab50a6 / src/Makefile.template : add . to ruby path -
CIA-3shogun: Soeren Sonnenburg master * rc5ea1a3 / (3 files): remove tmp generated files from java examples -
-!- shogun-buildbot [] has quit [Quit: buildmaster reconfigured: bot disconnecting]06:35
-!- shogun-buildbot [] has joined #shogun06:42
CIA-3shogun: Soeren Sonnenburg master * r04d6769 / (10 files): add Daniels first batch of .cs examples -
shogun-buildbotbuild #137 of java_modular is complete: Failure [failed test_1]  Build details are at  blamelist: sonne@debian.org07:21
shogun-buildbotbuild #138 of csharp_modular is complete: Failure [failed test_1]  Build details are at  blamelist: sonne@debian.org07:31
shogun-buildbotbuild #134 of ruby_modular is complete: Success [build successful]  Build details are at
-!- mrsrikanth [~mrsrikant@] has joined #shogun07:35
-!- f-x [~user@] has joined #shogun08:03
-!- mrsrikanth [~mrsrikant@] has quit [Quit: Leaving]09:07
CIA-3shogun: Shashwat Lal Das master * r620710c / (data src/shogun/features/StreamingSparseFeatures.h): expand_if_required now more accurate and sort_features modified for correct memory freeing in StreamingSparseFeatures. -
CIA-3shogun: Shashwat Lal Das master * r44936ba / src/shogun/features/StreamingSparseFeatures.h : Better doc for StreamingSparseFeatures. -
CIA-3shogun: Shashwat Lal Das master * r5219f4d / data : Reverted unwanted commit in data. -
CIA-3shogun: Soeren Sonnenburg master * r0deb8ca / src/shogun/features/StreamingSparseFeatures.h :10:49
CIA-3shogun: Merge pull request #318 from frx/streaming_vw10:49
CIA-3shogun: Better documentation and some issues resolved in StreamingSparseFeatures -
-!- blackburn [~blackburn@] has joined #shogun11:44
blackburnsonney2k: does ppwm fails on your machine11:45
@sonney2kpython? no11:48
shogun-buildbotbuild #138 of java_modular is complete: Success [build successful]  Build details are at
blackburnnot python, java11:50
blackburnit fails *sometimes*?!11:50
blackburnsonney2k: I have transitioned all of them back to "main" instead of "run"11:51
blackburnand fixed all those fail on my machine11:51
@sonney2kwhy that? run <-> main thing?11:51
@sonney2kblackburn, we will need to transition back to run when we want tests11:52
@sonney2ks/want/have working/11:52
blackburnyes but it was ~20% of tests11:53
blackburnbetter they all have identical structure11:53
blackburnsonney2k: could you please run java_modular/
blackburnbetter a few times11:55
@sonney2kcurrently running all the tests11:55
blackburnsonney2k: so when are you planning to run tests, etc? I would say we won't get it in time for all of langs11:57
@sonney2kblackburn, the *ppwm* example crashes with 50% probability12:00
blackburnoh I really like probability crash12:00
blackburnI doesn't crash on my machine12:01
blackburnsonney2k: have run ~50 times12:05
blackburnsonney2k: I have to ask you to find the error ;)12:08
@sonney2kfound it already12:08
@sonney2khow do I call logi ?12:08
@sonney2kthere is one line commented in that example12:08
@sonney2kwhich needs to be in there12:09
@sonney2kgot it12:09
blackburnshould work12:10
blackburnso what's up?12:10
blackburnwhat was wrong?!12:10
CIA-3shogun: Soeren Sonnenburg master * ra74aaf0 / (2 files): fix ppwm example and add option to check to run just a few examples -
@sonney2kblackburn, DoulbeMatrix.logi didn't work12:11
@sonney2kit is static so I it needed another static import12:11
blackburnbut crash?12:11
blackburnis it because set_pwm wasn't called12:11
@sonney2kanyways what was wrong was that that the pwm was not set12:12
@sonney2kso computing w couldn't work12:12
blackburnI see12:12
@sonney2kcheck missing in libshogun of course12:12
@sonney2kbut hey... user error too12:12
blackburnyeah but we have pretty much errors of this kind12:13
blackburne.g. yesterday I was trying to debug mkl multiclass12:13
@sonney2kblackburn, really?12:13
@sonney2kI wouldn't say there are that many12:13
blackburnit was using labels directly12:13
blackburnso tried to get -1 svm12:13
blackburnall of labels usage is pretty unreliable12:14
blackburnuhoh how to extract code :D12:14
@sonney2kblackburn, well you could just create a big diff12:14
blackburnis it ok?12:14
@sonney2kso all code you contributed12:14
@sonney2kblackburn, it seems but better check
blackburnI don't understand the purpose of this operations12:15
@sonney2kblackburn, I think that some projects might not have the source *that* open as we do12:15
@sonney2kthey might have some internal repository12:15
@sonney2kand only from time to time release things12:15
blackburnAll students will need to add a zipped tarball, which can include source files, a single .diff file, multiple .diff files, binary files12:15
@sonney2kso google just wants to ensure that the code the students wrote is open12:16
blackburnbinary files ahah12:16
blackburnvery opensourcy12:16
@sonney2kof course also check if they would want to hire the student12:16
@sonney2kblackburn, binary files == data / images / ...12:16
blackburnah ok12:16
CIA-3shogun: Soeren Sonnenburg master * r60ad972 / src/shogun/distributions/PositionalPWM.cpp : assert that m_pwm is set -
@sonney2kbugfix ^12:18
blackburnsonney2k: but ruby?12:20
@sonney2kblackburn, what is w/ ruby?12:20
blackburndid it pass?12:20
@sonney2kblackburn, could you please ls ../examples/undocumented/csharp_modular/*.cs12:20
blackburnyesterday wasn't12:20
@sonney2kthe buildbot has some wierd (1).cs in there wtf ?12:21
blackburnno (1) on my machine12:21
@sonney2kand it is trying to compile distance_geodesic_modular12:21
blackburnit didn't find distance_geodesic_modular too12:22
@sonney2kthere shouldn't be any12:22
@sonney2kand it should only check *.cs files12:22
blackburnstarted that script for my ~350 commits12:25
@sonney2kblackburn, you might want to post that to the other students / ml12:28
@sonney2kthat will help them12:28
@sonney2kbtw we have more warnings again :)12:28
blackburnyes if it have success12:28
@sonney2kbecause I managed to convince the buildbot to check for Warning / warning12:29
@sonney2kso now we see swig warnings too12:29
@sonney2kbtw if you can come up with a list of warnings to ignore12:29
@sonney2k(because we cannot fix them)12:29
@sonney2kthen I can also suppress them12:29
blackburn/usr/include/python2.6/numpy/__multiarray_api.h:1187:1: warning: 'int _import_array()' defined but not used [-Wunused-function] /usr/include/c++/4.6/backward/backward_warning.h:33:2:12:30
blackburnwarning: #warning This file includes at least one deprecated or antiquated header which may be removed without further notice at a future date. Please use a non-deprecated interface with equivalent functionality instead. For a listing of replacement headers and interfaces, consult the file backward_warning.h. To disable this warning use -Wno-deprecated. [-Wcpp]12:30
blackburnin python_modular12:30
@sonney2kblackburn, a real list - like an email with the lines12:31
blackburndeprecated warning raises from larank12:31
blackburnbtw we can't fix most of them12:32
blackburnsonney2k: do we need CMath available in modular interfaces?12:32
-!- Netsplit *.net <-> *.split quits: @sonney2k12:39
-!- Netsplit over, joins: @sonney2k12:39
-!- Netsplit *.net <-> *.split quits: serialhex12:41
-!- Netsplit over, joins: serialhex12:42
blackburnsonney2k: uploaded tarball, forwarded an email how to that12:53
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]13:13
-!- mrsrikanth [~mrsrikant@] has joined #shogun14:08
-!- f-x [~user@] has quit [Remote host closed the connection]14:52
-!- blackburn [~blackburn@] has joined #shogun15:19
-!- mrsrikanth [~mrsrikant@] has quit [Quit: Leaving]16:34
CIA-3shogun: Soeren Sonnenburg master * r973318c / (2 files in 2 dirs):19:52
CIA-3shogun: Fix tests for java and c#19:52
CIA-3shogun: - remove duplicate geodesic example in c#19:52
CIA-3shogun: - fix java's to exclude -
@sonney2kblackburn, great thanks19:53
@sonney2kblackburn, Re warnings - I know - that is why I would like to ignore them19:53
@sonney2kblackburn, c# failed because of some duplicate geodesic file named geodesic (1).cs19:54
@sonney2k(space in filename!)19:54
@sonney2kso java and c# should finally run through now19:54
blackburnsonney2k: nice19:56
blackburnsonney2k: ruby to go?19:56
@sonney2kno ruby works too19:56
blackburnI mean there are ~5 examples or so?19:57
@sonney2kblackburn, yeah - I hope serialhex does some more19:58
@sonney2kat least Daniel is doing some for c# now :)19:58
@sonney2kand serialhex promised - so I have high hopes he will do some19:58
blackburnsonney2k: so what can I do this evening?19:59
@sonney2ke.g. port examples / warning list to ignore for the buildbot20:02
@sonney2kapart from that we still have release script / make doc20:04
@sonney2kdebian package20:04
@sonney2kand heikos migrations (that is where you can help next week - I hope heiko manages to have sth running by monday)20:04
@sonney2kblackburn, ^20:05
@sonney2ka new web page would also be cool / release notes thanking gsoc - so you could also write the first draft of those20:05
@sonney2kis anything of the above OK for you?20:06
blackburnsonney2k: yeah first ones are ok ;)20:10
@sonney2kso which one will you do today?20:10
shogun-buildbotbuild #142 of lua_modular is complete: Failure [failed compile]  Build details are at  blamelist: sonne@debian.org20:10
blackburnI'm currently packing as I'm going back to samara20:10
blackburnI would create warning list and do sth with examples20:11
blackburnno idea how to do release script and heiko's migrations20:11
@sonney2kerm - why did gcc crash20:14
blackburnsonney2k: there was something with "bus error" or so20:15
shogun-buildbotbuild #144 of csharp_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #142 of java_modular is complete: Success [build successful]  Build details are at
CIA-3shogun: Soeren Sonnenburg master * r2f0a2c8 / (2 files):20:21
CIA-3shogun: Merge pull request #319 from frx/streaming_vw20:21
CIA-3shogun: Made VW output more exact. -
@sonney2kblackburn, that sounds more like a memory error...20:22
@sonney2kquite scary ...20:22
@sonney2kand not reproducible ...20:23
@sonney2kso either kernel bug (unlikely) or hardware problem20:23
@sonney2kblackburn, you could port a few examples from python_modular to lua20:29
@sonney2kblackburn, btw did you ever check whether *documented* examples work20:32
@sonney2kand btw did you document your examples in descriptions/modular/ ?20:32
@sonney2kthen you should do that20:32
@sonney2kjust run make in the doc dir20:32
@sonney2kerr examples dir20:32
blackburnyeah that is exactly what I should do20:32
@sonney2kI would say you go through the generated documented/python_modular/*.py files and document all examples that are not yet documented and that you can make sense of20:34
blackburnyeah okay20:42
blackburnwill start soon20:43
shogun-buildbotbuild #143 of lua_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #13 of nightly_none is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #13 of nightly_all is complete: Success [build successful]  Build details are at
@sonney2kblackburn, how about git log --name-status --graph for generating the new ChangeLog?21:44
blackburnhmm yeah21:45
blackburnprobably it should be ok21:45
@sonney2kabsolutely amazing21:49
@sonney2kall tests ran through!21:49
@sonney2kthe green lantern ;-)
-!- alesis-novik [~alesis@] has quit [Quit: Leaving]21:52
blackburn"battle station fully operational"21:53
blackburnsonney2k: is here again?!22:00
@sonney2kwhy should it go?22:00
@sonney2kdid you git rm it?22:00
blackburnyou did22:00
@sonney2kdoing again22:00
blackburnsonney2k: oh make made a list of undocumented examples!22:03
@sonney2kblackburn, regarding the release process - I would update ChangeLog, documentation then git commit / push22:08
@sonney2kthen create a shogun-1.0 branch22:09
@sonney2kthen add lib/versionstring.h to that branch22:09
blackburnhow to use that?22:09
@sonney2kupdate docu in there22:09
@sonney2kand then tag version 1.0.022:09
@sonney2kthen create a tarball out of it22:09
@sonney2kdoes that make sense?22:10
blackburnyeah but how to do that technically22:10
@sonney2kI dont' understand the question22:11
@sonney2kmake release will do it22:11
@sonney2kit won't create the branch if things are done from the shogun 1.0 branch - this way we can create bugfix releases22:12
blackburnsonney2k: why do we need documented/undocumented paths in examples?22:16
@sonney2kto create documented examples from undocumented ones22:17
@sonney2kblackburn, not sure if this was what you were asking for22:17
blackburnwhy not to store all of them in one dir?22:18
@sonney2kstore what in one dir?22:18
@sonney2kdocumented and undocumented?22:18
@sonney2kthen one would overwrite them when generating them22:19
blackburnI didn't notice there are description integrated22:19
@sonney2kof course we could have a  <begin descr> </end descr> at the beginning and replace everything inside but that is hmmhh annoying too22:20
@sonney2kof course22:20
@sonney2kthat is why you need to write descriptions :)22:20
@sonney2kbtw please do make doc22:20
@sonney2kthen open ^22:20
@sonney2kand you will see that some classes don't yet have a @brief description22:21
@sonney2kso please add one22:21
CIA-3shogun: Soeren Sonnenburg master * r931e913 / src/ChangeLog : generate changelog based on git history -
CIA-3shogun: Soeren Sonnenburg master * r95f5e1e / src/configure : enable csharp autodetection in configure -
CIA-3shogun: Soeren Sonnenburg master * rda0f5b4 / src/Makefile : generate changelog based on git history -
CIA-3shogun: Soeren Sonnenburg master * reb40088 / Makefile : some progress on make release magic -
CIA-3shogun: Soeren Sonnenburg master * rb2d77a9 / examples/ : remove example -
blackburnsonney2k: do you have any other lang in mind we can interface to?22:38
@sonney2kisn't what we have enough?22:38
blackburnyeah that's why I'm asking22:39
@sonney2kbut since you ask - r_modular, matlab_modular would be cool22:41
blackburnI have absolutely no idea how can one use say lua :D22:43
@sonney2kblackburn, lua is ultra lightweight + fast22:43
@sonney2kit is used in games22:43
@sonney2kfor high-level scripting22:43
blackburnshogun isn't22:43
@sonney2kgames arent' either22:44
blackburnyeah I know it is being used in games22:44
blackburnbut who would ever use shogun in games?!22:44
@sonney2klook at torch522:44
@sonney2kanother big ml toolbox22:44
@sonney2kit interfaces to lua only22:44
@sonney2kwhy what?22:45
blackburnwhy only to lua?22:45
blackburnheh torch5 is crazy22:46
blackburnC with classes22:46
@sonney2kblackburn, looks like we are the only ones to use swig...22:47
@sonney2k(only ml people that is)22:47
@sonney2kso writing interfaces is very expensive w/o22:47
blackburnI can't mind any other lang than python to use for some scientific shit :D22:48
@sonney2kC++ :)22:49
@sonney2kI don't think a lot of ruby/lua/c# exists for that22:49
@sonney2kbut R is *very* popular22:49
@sonney2kand java/C++ are too22:50
blackburnI don't know.. but that's crazy for me22:50
blackburnpython scientific stack is very powerful thing now22:51
blackburnmatplotlib, scipy, scikits, etc22:51
@sonney2kblackburn, yes, but octave and R have that already for many many years22:52
@sonney2kand some want more performance / have other projects to integrate things22:53
@sonney2kother projects in other languages that is22:53
blackburnI see22:53
@sonney2kanyway better continue the doc22:53
blackburnsonney2k: I hope it will be possible to read :D22:59
blackburnI still don't know how to say some complex things using english22:59
@sonney2kask serialhex when you are very unsure23:02
-!- blackburn1 [~blackburn@] has joined #shogun23:03
-!- blackburn [~blackburn@] has quit [Ping timeout: 245 seconds]23:04
serialhexblackburn1: "some complex things in english" :P23:10
-!- blackburn1 is now known as blackburn23:11
serialhextonight is gonna be one of those nights blackburn...23:12
blackburnserialhex: what?23:13
serialhexredbull & 5-hour energy :P  beef jerky & potato chips for "real" food too23:13
blackburn?! :)23:14
serialhexyes, lots of work... classes started, shogun programs, work in the am.... :-/23:15
blackburnI see23:15
@sonney2kserialhex, please ask now if the current examples don't work for you23:23
serialhexsonney2k: the current examples work fine for me (the ones included in upstream) i'm going to be spending all my free time getting lots more done23:27
CIA-3shogun: Sergey Lisitsyn master * rf949f36 / (21 files): Example doc updates -
CIA-3shogun: Sergey Lisitsyn master * r4887ddc / src/shogun/preprocessor/KernelPCA.h : Added reference for kPCA -
@sonney2kserialhex, thanks - please closely follow the ones in python - you did that already and I think kernel_gaussian_modular.rb is a good template23:28
serialhexyes, they are almost direct translations from one to the other (it helps that the syntax is very similar)23:28
CIA-3shogun: Sergey Lisitsyn master * ra126380 / (2 files): Renamed wrong called laplacian eigenmaps example description -
@sonney2kyeah thats true - it shouldn't be too hard23:31
@sonney2kI guess python borrowed from ruby or vice versa23:32
serialhexwell, most programming langs are ~similar, the difference is in the details (like ruby methods don't need to end in (), python ones do, even w/o args)23:33
serialhexthere is more, btu +- the same23:34
CIA-3shogun: Sergey Lisitsyn master * r76ed357 / (12 files in 2 dirs): Preproc -> preprocessor naming transition for modular examples -
@sonney2kthere we go23:50
blackburnsonney2k: did you update it23:50
@sonney2kversion 1.0.023:51
blackburnah I see23:51
blackburnSHOGUN is implemented in C++ and interfaces to Matlab(tm), R, Octave and Python23:51
CIA-3shogun: Soeren Sonnenburg master * r6898577 / (35 files in 5 dirs):23:54
CIA-3shogun: Documentation / Release adjustments for shogun 1.0.023:54
CIA-3shogun: - add put-online target to makefile23:54
CIA-3shogun: - fix install documentation23:54
CIA-3shogun: - include new lua/java/ruby/csharp examples in doc23:54
CIA-3shogun: - rename dirs and symlinks to x -> x_static -
@sonney2kblackburn, feel free to fix whatever mistakes you find23:54
@sonney2kblackburn, that looks pretty impressive though:
@sonney2kblackburn, we will now keep the documentation under doc/<lang>/version23:55
blackburnnot bad23:55
@sonney2kand have symlinks doc/<lang>/current pointing to the currentl release23:56
@sonney2kand latest to the latest devel thingy that the buildbot will create (hopefully)23:56
@sonney2kblackburn, the good thing is that this is all autogenerated stuff so these examples work ;-)23:57
blackburnthe estimated cost of shogun is23:59
blackburn??$ 2,392,85123:59
--- Log closed Mon Aug 29 00:00:20 2011

Generated by 2.10.0 by Marius Gedminas - find it at!