Open in new window / Try shogun cloud
--- Log opened Wed Mar 27 00:00:16 2013
-!- shogun-notifier- [] has quit [Quit: transmission timeout]00:31
-!- FSCV_ [~FSCV@] has quit [Quit: Leaving]01:07
shogun-buildbotbuild #334 of nightly_default is complete: Failure [failed test]  Build details are at
-!- n4nd0 [] has joined #shogun07:05
-!- sumit [75e1a693@gateway/web/freenode/ip.] has joined #shogun07:13
-!- shogun-notifier- [] has joined #shogun07:23
shogun-notifier-shogun: Sergey Lisitsyn :dev * b5ce0f2 / examples/undocumented/libshogun/ (14 files):
shogun-notifier-shogun: Made some examples work without eigen307:23
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]07:42
-!- hoijui [] has joined #shogun07:50
-!- blackburn [] has joined #shogun08:24
shogun-notifier-shogun: Soeren Sonnenburg :master * 8a992ee / src/Makefile:
shogun-notifier-shogun: revert commit 001790288e12ada4bd2d4956007b4f87834d84c7 for main Makefile08:32
shogun-notifier-shogun: Soeren Sonnenburg :master * 3e79d94 / examples/undocumented/libshogun/ (14 files):
shogun-notifier-shogun: make examples fail nicely when eigen3 is not available08:47
blackburnsonney2k: ??08:47
shogun-buildbotbuild #381 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>08:49
blackburnsonney2k: why do you work on master08:49
blackburnand why did you duplicate my commit :)08:49
shogun-buildbotbuild #382 of rpm1 - libshogun is complete: Success [build successful]  Build details are at
sonne|workblackburn: please drop the dev branch for now09:03
sonne|workblackburn: I didn't see your commit unfortunately09:03
n4nd0sonne|work: why do you prefer to use master directly?09:03
sonne|workbecause we need to get things back to work before we can switch09:03
blackburnsonne|work: that's a secret plan of you - please let us know next time :)09:04
sonne|workerm I didn't know we switched already09:07
sonne|workI was all the time only doing preparations for the switch09:08
sonne|workshogun-buildbot: force build 'nightly_default'09:09
shogun-buildbotThe build has been queued, I'll give a shout when it starts09:09
shogun-buildbotbuild #964 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Soeren Sonnenburg <>09:18
shogun-buildbotbuild #965 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Soeren Sonnenburg <>09:35
shogun-buildbotbuild #335 forced09:35
shogun-buildbotI'll give a shout when the build finishes09:35
-!- n4nd0 [] has quit [Quit: leaving]09:44
-!- n4nd0 [] has joined #shogun09:50
sumitblackburn: Just A small doubt , How far has Shogun 2.1.0 progressed in implementing RandNLA algorithms,.?I mean I could trace Lapack.cpp eigen3 though,,any clues for ERROR ANALYSIS?MATRIX MULTIPLICATION...MATRIX Decomposition??Divided Differences?Integration formulae,EigenValue Algos etc etc?09:52
n4nd0sumit: where are you getting that error?09:53
-!- hoijui [] has quit [Read error: Connection reset by peer]09:55
sumitWell I had a glance on GsoC 2013 Ideas List - and was focussing on Implementing RandNLA algorithms,,I think there's much more to it , for SHOGUN TOOLBOX,,as in many algorithms yet to be implemented in Shogun,isn't it?09:55
sumitI thence just wanted to know if I was missing out on some already implemented ones..09:55
n4nd0aham ok09:56
n4nd0I completely misunderstood, sorry09:56
sumitits ok :)09:57
n4nd0I don't know much about RandNLA, let's see what blackburn can tell you about it09:58
sumitWell,,so I was about to contact the concerned mentors,,but was trying to understand Shogun Structure,,And I think this is a promising area to work in,isn't it?I am sending the wiki link,,I wish to contribute by implementing some arrays of algorithms enlisted in there09:58
blackburnsumit: nothing is implemented yet09:58
n4nd0sumit: sure, feel free to contribute. That would be great09:58
sumit ,, Yes thanks :),,,So blackburn , I will send you an email wherein I shall discuss the approach and other details , thanks :)09:59
blackburnsumit: we are waiting for some response from Igor10:06
blackburnsumit: feel free to email him and CC me10:06
blackburnthat's cool there is a list of topics but we should understand what to implement and what for :)10:07
sumitblackburn: Sure Sir10:07
sumitYes Yes, I was just about to send the mail to you both10:07
shogun-buildbotbuild #335 of nightly_default is complete: Failure [failed test]  Build details are at
-!- sumit [75e1a693@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:55
-!- heiko [] has joined #shogun11:19
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun11:26
-!- n4nd0 [] has quit [Quit: leaving]11:41
-!- n4nd0 [] has joined #shogun11:44
-!- shogun-notifier- [] has quit [Quit: transmission timeout]11:47
-!- ZeThomas [~thomaspr@2a02:2c40:100:a000:8036:9367:7b72:6a98] has joined #shogun11:54
ZeThomashey, quick question11:54
n4nd0ZeThomas: shoot11:54
ZeThomasis there a way to do leave-one-out testing/validation efficiently (with SVMLight, or others)?11:55
n4nd0ZeThomas: mm no sure, have you taken a look to the evaluation module?11:57
n4nd0I'm not* sure11:57
ZeThomasyeah, I've been browsing the docs for some time now, can't really find something11:58
ZeThomasbut maybe I'm overlooking11:58
sonne|workZeThomas: no11:58
ZeThomasand this kind of machine, described here:
ZeThomashow could I implement this, is there a way to alter some of the elements in a kernel matrix?12:00
ZeThomas(because I just need to put diagonals to 0)12:00
n4nd0ZeThomas: I guess there must be a way to do it, but it probably depends on the machine you want to use12:02
ZeThomassay I want a SVMLight with GaussianRBF, but I want to set all diagonal elements of the kernel matrix to zero before I train12:05
ZeThomashow would I best do this? tinker with the values of a GaussianKernel object (if possible), or subclass it maybe?12:05
n4nd0ZeThomas: let me check a second12:07
n4nd0ZeThomas: maybe CustomKernel can help you12:11
-!- blackburn [] has quit [Quit: Leaving.]12:16
lambdayn4nd0: hi12:40
lambdayheiko: hello12:40
lambdayI have removed the SSK String kernel12:41
heikolambday:  hi, have you pushed to your PR?12:42
heikolambday: nice, I just saw it12:43
lambdayone question - does doxygen support amsmath specific latex equation? say, \begin{align*}?12:43
heikolambday: I dont know, you can try it out :)12:44
heikobut usually we just use the std \[ \] stuff12:44
heikolambday: okay, looks good, now unit test and12:45
heikothen Ill merge12:46
lambdayheiko: okay... I am doing make doc.. I thought it would be nice to describe it the way you did :)12:46
heikothe way I did?12:46
lambdayon the issues page - step by step..12:47
lambdaylog of sum over diagonal evements times 212:47
heikoah I see, yes you can do that12:48
heikowould be good!12:48
lambdayheiko: will let you know if it works.. will start with the unit test then12:48
-!- blackburn [] has joined #shogun12:56
-!- n4nd0 [] has quit [Ping timeout: 264 seconds]13:03
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]13:05
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun13:10
-!- ZeThomas [~thomaspr@2a02:2c40:100:a000:8036:9367:7b72:6a98] has quit [Quit: Leaving]13:16
-!- n4nd0 [] has joined #shogun13:17
n4nd0sonne|work: I vote for you for shogun mentor of the structured learning project :)14:23
-!- sumit [75e1d426@gateway/web/freenode/ip.] has joined #shogun14:28
lambdayheiko: doxygen info is updated. please check..14:44
lambdayand how do I run the existing unit tests? I've installed valgrind.. and then was trying to run "make" from shogun/tests/unit...14:44
-!- sumit [75e1d426@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]15:09
-!- n4nd0 [] has quit [Quit: leaving]15:31
heikolambday: sorry was out for lunch15:48
heikolambday: make unit-tests does it for you15:48
heikoor make in the tests/unit folder15:48
heikoand make valgrind in the unit/folder runs valgrind for the tests (some of them leak)(15:49
heikomake valgrind-per-module runs them seperately to hunt for memory leaks easier15:50
heikolambday: a few things on committing15:51
heikowhen you do a pull, always use pull --rebase15:51
heikootherwise our commit history is messed up15:51
heikoalso, never do commit -a, you dont want to commit anything else than source files15:51
heikosorry, you will have to remove all the documentation commits (see PR)15:51
-!- sonne|work [] has quit [Quit: Leaving.]16:03
lambdayheiko: okay.. I was wondering why the other files got changed, because I didn't change anything in them..16:07
heikoyour doxygen changed them16:07
heikolambday: just commit your code changes seperately, then you will have no problems16:07
lambdayheiko: okay.. so, I have to do a rollback on my forked repo?16:08
lambdayand add the documentation later on and commit again?16:08
heikono, never commit the documentation16:12
heikoonly the doxygen comments16:12
heikowe generate automatically16:12
heikoonly commit the interface and the implementation of your method16:13
heiko(and then unit-test)16:13
heikonothing else16:13
heikonever do commit -a16:13
heikolambday: sorry for all this mess, but this is exactly why we have these intro tasks :)16:14
lambdayheiko: I am sorry for making this mess :( totally new to github..16:16
heikolambday:  dont worry, I had exactly the same problems in the beginning :D16:17
lambdayheiko: hope to get a grip on git asap :)16:17
heikoyou will16:17
lambdayand you mean I should only commit the source files, right? I did git commit -a and that seems to mess it all up16:18
lambdayheiko: if I want to reset the state of the repo right after a previous commit, should "git reset <commit-code>" do it?16:21
heikoyes, but please make a copy of your changes to not accidentally loose anything16:21
lambdayheiko: okay..16:22
-!- blackburn [] has quit [Quit: Leaving.]16:45
-!- FSCV [~FSCV@] has joined #shogun17:15
@sonney2kso in 48 hours its gsoc deadline17:18
-!- shogun-notifier- [] has joined #shogun17:28
shogun-notifier-shogun: Soeren Sonnenburg :master * 6050e08 / tests/integration/python_modular/
shogun-notifier-shogun: blacklist failing/flawky integration tests17:28
-!- blackburn [~blackburn@] has joined #shogun17:38
@sonney2kblackburn, 48 hours!17:39
blackburnsonney2k: that's ok17:41
blackburnsonney2k: they won't take a look at ideas for a few days I think17:41
blackburnsonney2k: April 1 - 5:Google program administrators review organization proposals.17:42
blackburnsonney2k: so we've got a weekend to polish before they take a look17:42
blackburnsonney2k: I am able to do that at least17:44
blackburnso no worries we are ok17:44
lambdayheiko: I have reverted to the point before it got messy. and then updated Statistics.h.. please check..17:53
lambdayheiko: writing unit test would mean, I have to write a file with TEST(Statistics, log_det) {...} with a small matrix and test whether log_det returns the value as expected, using EXPECT_EQ??18:11
lambdayand then run unit-tests and check for memory leaks using valgrind?18:14
shogun-buildbotbuild #966 of deb3 - modular_interfaces is complete: Failure [failed test ruby_modular]  Build details are at  blamelist: Soeren Sonnenburg <>18:24
heikolambday: yes, just like the existing tests18:41
lambdayheiko: okay.. there is no test for mathematics, so should I create a directory "mathematics" in tests/unit and put the cc file there?18:43
heikolambday: there should be one for statistics18:43
lambdayyes there is..18:43
heikoyou are right18:44
heikocreate a new one for mathematics/Statistics18:44
heikoso yes, new folder and new file18:44
lambdayheiko: okay.. :)18:45
shogun-notifier-shogun: Heiko Strathmann :master * a7eb2ec / src/shogun/kernel/CustomKernel.h:
shogun-notifier-shogun: corrected typo18:56
shogun-notifier-shogun: Heiko Strathmann :master * 4f89569 / src/shogun/statistics/KernelTwoSampleTestStatistic.cpp:
shogun-notifier-shogun: added some assertions and prepared custom kernels for quadratic time mmd18:56
shogun-notifier-shogun: Heiko Strathmann :master * dfa3fec / src/shogun/statistics/ (2 files):
shogun-notifier-shogun: added constructor for using a custom kernel without features18:56
shogun-notifier-shogun: Heiko Strathmann :master * 978b5b2 / tests/unit/statistics/
shogun-notifier-shogun: updated unit test to use new constructor for custom kernel if no data is available18:56
shogun-notifier-shogun: Heiko Strathmann :master * ab7bbed / / (5 files):
shogun-notifier-shogun: Merge pull request #957 from karlnapf/master18:56
shogun-notifier-shogun: better handling for custom kernels for quadratic time mmd18:56
shogun-notifier-shogun: Heiko Strathmann :master * cb523e3 / tests/unit/statistics/
shogun-notifier-shogun: added all three methods for null approximation to precomputed kernel test19:07
shogun-notifier-shogun: Heiko Strathmann :master * 0eff35b / tests/unit/statistics/
shogun-notifier-shogun: Merge pull request #958 from karlnapf/master19:07
shogun-notifier-shogun: extended test19:07
-!- zhuoran [80eef759@gateway/web/freenode/ip.] has joined #shogun19:45
-!- zhuoran [80eef759@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:49
shogun-buildbotbuild #967 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at
-!- heiko [] has left #shogun []19:52
@sonney2kblackburn, well what 'simple' tasks do we have?20:37
@sonney2kIMHO some file i/o stuff20:37
blackburnsonney2k: we don't have simple tasks at all I am afraid20:37
@sonney2ksome write examples20:37
@sonney2kwith python / django underneath and d3js20:38
blackburnsonney2k: no news neither from igor nor cheng20:41
@sonney2kblackburn, what else would you imagine?20:41
@sonney2kblackburn, no cheng sent his stuff20:41
blackburnsonney2k: to you?20:41
blackburnsonney2k: do you mean ideas names as stuff?20:41
@sonney2kand you20:41
@sonney2khis ideah20:41
blackburnbut no descriptions right?20:42
@sonney2kw/ description20:42
@sonney2klet me resend it20:42
blackburnlet me check20:43
blackburnsonney2k: ohhhh gmail did a nice job20:44
blackburnsonney2k: it thinks cheng is going to steal my credit card20:44
blackburnsonney2k: gmail is playing bad jokes on me, he placed some chinese journal invitation to spam as well :D20:47
blackburnI disabled spam filter on second mail, hope that will help20:48
-!- sumit [ca4eaca2@gateway/web/freenode/ip.] has joined #shogun20:49
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]21:03
blackburnsonney2k: I am still scared with heiko requirements :D21:05
shogun-notifier-shogun: Soeren Sonnenburg :master * ef114dd / tests/integration/ (8 files):
shogun-notifier-shogun: make test scripts fail when a test fails21:11
@sonney2kblackburn, hehe21:12
blackburnsonney2k: he needs a little vapnik21:13
@sonney2kor he needs to clone himself21:14
blackburnsonney2k: have you seen lcov output?21:14
blackburnI mean how it looks like21:14
@sonney2knever tried21:17
@sonney2kblackburn, we need that for test coverage...21:17
blackburnsonney2k: let me show you, a minute21:18
@sonney2kbut I guess for now it is more important to get the build back to green21:18
blackburnsonney2k: I have learned how to work with it in tapkee so I'll put it to shogun21:18
@sonney2kwe still have flaky tests in basically all static interfaces21:18
blackburnoh right21:18
@sonney2kand then octave_modular and some in python_modular21:18
blackburnsonney2k: we need more modularity21:19
@sonney2kI hope I managed to get the buildbot working as I wanted21:19
@sonney2kblackburn, what do you mean?21:19
blackburnsonney2k: may be it would be better to make things less coupled somehow21:20
blackburnsonney2k: I do not like implementations right in classes21:20
@sonney2kblackburn, you are talking about multiple shogun libs?21:20
@sonney2kor what?21:20
blackburnsonney2k: yeah may be that21:20
blackburnsonney2k: I feel classes should be adapters for real core21:20
blackburnit feels improper to have god classes21:21
blackburnsonney2k: a good rule is <20 lines per method right?21:23
blackburnsonney2k: ah yes and I still dream about client-server shogun21:24
shogun-buildbotbuild #842 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at  blamelist: Soeren Sonnenburg <>21:24
blackburnsonney2k: do you know anything about remote method invocation in java? that is used in EJB21:25
blackburnso class you own could be just a stub that calls a method remotely21:26
blackburnI can't get rid it could be a crazy feature21:27
@sonney2kblackburn, heh with static interfaces this would even be possible :D21:29
@sonney2kjust send over the command and it will do21:30
blackburnsonney2k: it is possible with modular as well21:30
blackburnsonney2k: I think I have an idea how can we speed up compilation and improve code21:30
blackburnsonney2k: we could put all class fields into some say 'context'21:31
blackburnthat stores all the data as may be map?21:31
@sonney2kblackburn, you are having some epiphany ?21:31
blackburnsonney2k: yes jesus and allah are here tell me what to do21:32
@sonney2kblackburn, thats what we have with parameters21:32
@sonney2kand then?21:32
blackburnsonney2k: I mean private fields are not exposed in headers21:32
blackburnsonney2k: won't that help swig to be faster?21:33
@sonney2kblackburn, I don't know whether I understood you correctly but your suggestion is to put all member variables into a single map (one per class?)21:34
blackburnsonney2k: yeah somewhat21:35
blackburnit could be generic though21:35
@sonney2kthat won't help swig though, IIRC swig is not exposing access to non-public variables21:36
blackburnsonney2k: hmm alright21:36
@sonney2kblackburn, what helps with swig is really reducing the number of exposed functions21:41
@sonney2kand stuff like we have sgvector/matrix/dense/spars/stringfeatures of all kinds of basic types causes an explosing in wrapper code size21:41
blackburnsonney2k: I have to admit : I hate getters and setters21:43
-!- sumit [ca4eaca2@gateway/web/freenode/ip.] has quit [Quit: Page closed]21:43
blackburnit feels java 1.221:43
@sonney2kblackburn, what do you prefer? builder patterns :D21:45
blackburnsonney2k: builder is nice21:45
@sonney2kblackburn, look
@sonney2konly few examples fail21:45
blackburnsonney2k: yeah I just checked that21:45
@sonney2ktests actaully21:46
blackburnsonney2k: oh HMM21:46
@sonney2kblackburn, any idea what we should do to get things green rather quick?21:46
blackburnsonney2k: not really21:47
blackburnsonney2k: HMM is 5580 lines21:47
blackburnthat is terrible21:47
blackburnI don't know if HMM code can be shorter but that's a dead code unfortunately21:48
@sonney2kmaybe we should blacklist these 5 tests for now21:48
blackburnsonney2k: HMM for sure21:48
@sonney2kand then add the blacklist stuff to the list of quick fixes21:48
@sonney2kHMM is not quickly fixed though21:48
@sonney2k(I guess)21:48
blackburnsonney2k: same with kernels21:49
blackburnI approached these kernels one day21:49
blackburnbut that's scary as well21:49
@sonney2kfixing kernels shoudl be much easier I think21:49
@sonney2kso lets summarize our todo21:50
@sonney2k1) get proposal ready (most important)21:50
@sonney2k2) get ideas list ready21:50
@sonney2k3) get shogun back to green21:50
@sonney2k4) switch to new dev model and let the games begin21:50
blackburnsonney2k: 1) is more or less done - some fixes21:50
@sonney2kadd website changes somewhere in there21:51
@sonney2kblackburn, so lets finish 1) & 2) tomorrow21:52
blackburnsonney2k: alright21:52
@sonney2kfor the rest we likely need help21:52
blackburnsonney2k: I will have to write descriptions for igor stuff and vowpal wabbit21:52
@sonney2kI am progressing very slowly with 3)21:52
blackburnsonney2k: will you have time to write some descriptions for static and d3?21:53
@sonney2kand I a couple more too for the optimization stuff from last year, the fast kernel expansion stuff etc21:53
@sonney2klets hope so21:53
blackburnI am going to have a busy day at job tomorrow21:54
@sonney2kbut bedtime now21:54
blackburnhope will manage to do all the stuff21:54
blackburnyeah same for me21:54
blackburngood night21:54
-!- blackburn [~blackburn@] has left #shogun []21:55
shogun-buildbotbuild #970 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Soeren Sonnenburg <>22:00
-!- zhuoran [80eef759@gateway/web/freenode/ip.] has joined #shogun23:03
-!- FSCV [~FSCV@] has quit [Quit: This computer has gone to sleep]23:49
-!- FSCV [~FSCV@] has joined #shogun23:54
-!- zhuoran [80eef759@gateway/web/freenode/ip.] has quit [Quit: Page closed]23:56
-!- FSCV_ [~FSCV@] has joined #shogun23:58
--- Log closed Thu Mar 28 00:00:17 2013