Open in new window / Try shogun cloud
--- Log opened Thu Oct 18 00:00:05 2012
--- Day changed Thu Oct 18 2012
blackburnhow much?00:00
wikingi still want to see what mosek does though00:00
wikingone loss function for my case: 34.5900:00
wikingthe published one best case is 35.600:00
blackburnwhat's the difference between your method and previous?00:01
wikingwell the underlying so-solver00:01
blackburnhmm so just a solver issue?00:01
blackburnare you using BMRM?00:01
blackburnproximal point?00:02
wikingp3bmrm for that matter00:02
wikingnot ppbmrm00:02
wikingbut the p3bmrm00:02
wikingwith multiple cps00:02
blackburnwell you should publish it then :D00:02
wikingi need to get the other benchmarking data set00:03
wikingbut it not only gives me better results00:03
wikingbut much less training time is needed as well00:03
wikingit's like 2/300:03
wikingso pretty good00:03
blackburnhere comes a title "Efficient latent structured support vector machines training via multiple cutting plane models"00:04
blackburnlets write an abstract :D00:05
blackburnNowadays latent models of visual concept recognition are widespread and efficient. However blablabla00:06
wikingbtw freak is already included in opencv 2.4.200:08
wikingafaik we have a timer function in shogun right? ?:)00:10
blackburnwiking: CTime right00:14
blackburnwiking: yes freak is in for a while already00:14
blackburnwiking: I tried it as a dense descriptor00:16
blackburnworks ok but in the end I started to believe it is just wrong00:17
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]01:05
shogun-buildbotbuild #119 of nightly_none is complete: Failure [failed compile]  Build details are at
shogun-buildbotbuild #136 of nightly_default is complete: Failure [failed test]  Build details are at
-!- in3xes [~in3xes@] has joined #shogun07:36
-!- in3xes [~in3xes@] has quit [Remote host closed the connection]08:34
wikingsonne|work: online?09:15
wikingsonne|work: can u give me an access to a buildslave where there is clang?09:16
sonne|work that machine is doing all mailinglist, webhosting, buildslave09:17
sonne|workso no09:17
wikingi will not vandalize it09:17
sonne|workbut I should have a bit of time once on the mentor summit to do these changes09:17
wikingjust want to fix the static analyzer09:17
wikingit should be a 2liner :)09:18
wikingor is there sense to add another buildslave?09:18
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun09:18
wikingsince i have a spare ubuntu server running around09:19
wikingso there we could have static analyzer09:19
blackburnwiking da servers master09:19
wikingsonne|work: maybe that sounds better? and then that would be the static analyzer bot?09:19
sonne|workwiking: we don't have an ubuntu buildslave yet09:20
sonne|workso it would make sense09:20
wikingsonne|work: oh ok09:20
sonne|workand yes give me the 2 liner :)09:20
wikingsonne|work: i'll do the thingy09:20
wikingdont go away!09:20
sonne|workwiking: I don't have access to the bot from work but can maybe do it form the airport tomorrow09:21
wikingsonne|work: buildslave create-slave shogun-buildbot shogun-ubuntu1 <passwd>09:31
wikingthat sounds good for you?09:31
sonne|workprobably good - w/o access to the bots I have no clue (brain == empty)09:33
wikingsonne|work: okok i've tried to follow what's in the freebsd bot config09:33
wikingi've sent you in /msg the details, or should i send it via email?09:34
wikingi'll run now a test with static analyzer on that host on shogun09:37
wikingblackburn: here?09:38
wikingblackburn: what packages should i have on ubuntu for a decent shougn build?09:39
sonne|workwiking: apt-get build-dep shogun09:40
sonne|workand also the newer ones09:40
wikingmmm something is weird with atlas/lapack09:42
wikinginstalled libatlas-dev09:42
wikingbut doesn't recongise it09:42
blackburnwiking: libatlas3gf-base09:42
sonne|workwiking, well you need arpack, nlopt, superlu, snappy etc too09:45
wikingChecking for Atlas support ... no09:45
sonne|workwiking: do you want to just build libshogun or the whole thing with octave, java, lua etc...09:45
wikinglibshogun at first09:46
blackburnwiking: install libatlas* :D09:58
wikingok i've done with it09:58
blackburneverything that has atlas word :)09:58
wikingnow it's doing the static-analyzer09:58
wikingwhich package is NLOPT?10:02
wikingi couldn't find anything with apt-cache search nlopt10:02
blackburnwiking: bingo, it is nonexistent10:03
wikingi see10:03
wikingok then10:03
wikingu install it by hand?10:03
sonne|workwiking: there is one in debian unstable - so probably newer ubuntus will have it10:18
wikingblackburn: yesterday i've tried to do a SGSparseVector -> Eigen::SparseVector mapping but it seems it's more complicated than i thought... i.e. it would take an awful lot of copy elements... and i think this way we'd loose the performance gain of eigen10:19
wikingsonne|work: noup :( i have 12.04.110:19
wikingi.e. latest stable ubuntu and doesn't have it :(10:19
blackburn12.10 is out :)10:19
wikingoh woah10:19
blackburnwiking: well yeah it is a problem10:19
sonne|workwiking: then grab the .deb from debian10:20
wikingblackburn: 12.10 release date today :DDD10:21
wikingblackburn: 0day you are10:21
wikinghas n4nd0 been around lately?10:21
wikingfinally i've managed to progress with his mosek solver as well10:21
blackburnwiking: no, he is rather in gmail than in irc10:22
wikingblackburn: yeah i see him on g+10:23
wikingi've found the bug in his code10:24
wikingbut it's rather strange10:24
wikingi think there's something around CList10:24
wikingwith referencing...10:24
wikingyey it worked:
blackburnstatic thing rocks heh10:30
wikingno mem-leak  8X10:36
wikingalmost cannot believe :D10:36
wikingok here's a little PR for mosek :)10:40
wikingmerge plz
sonne|workwiking: write him an email10:57
sonne|workwiking: whats wrong with CList?10:57
sonne|workdid he use the thread safe functions or the non-safe ones?10:57
wikingsonne|work: already discussed with him.... he's studying for an exam he has tomorrow10:58
wikingso we'll talk after10:58
wikingstarting from there10:59
wikingtill line 17710:59
wikingas i see it10:59
wikingi'd expect that result has m_refcount =110:59
wikingon line 177711:00
wikingbut that's not the case11:00
sonne|workthe non thread safe ones11:00
wikinghow can i have non-safe one?11:01
sonne|workwiking: it doesn't matter here - there are no threads11:03
wikingyeah :)11:03
wikingbut anyhow11:03
sonne|workwiking: shouldn' thte refcount be 2?11:03
wikingan insert_element should ++ the refcount11:03
wikingexpecially that the list is delete_element=true11:04
sonne|workbut he also uses argmax to get it11:04
sonne|workso it might be 1 already there11:04
sonne|workwiking: what is the actual count?11:05
wikingnot that doesn't add a ref11:05
sonne|workso it has 0 there?11:05
wikingok then i'm stupid :D11:07
wikingbecuase the insert -> 111:08
wikingand l177 -> 011:08
shogun-buildbotbuild #449 of deb2 - static_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Viktor Gal <>11:10
-!- adoniscik [] has quit [Ping timeout: 240 seconds]11:28
sonne|workblackburn: you?11:36
sonne|workor clue?11:36
blackburnsonne|work: heiko11:36
blackburnor me11:36
blackburnrather heiko11:36
shogun-buildbotbuild #558 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Viktor Gal <>11:37
sonne|workwiking: could you please ask heiko then? otherwise you will be drowning in emails11:38
blackburnheh I bet wiking already set up his spam filter :D just as I did11:39
blackburnwell joking gmail puts all these mails to the special folder I never read11:39
blackburnbut it is not called 'spam'11:39
shogun-buildbotbuild #559 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Soeren Sonnenburg <>11:43
shogun-buildbotbuild #450 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at
wikingi think i've solved the mosek thing... but it performs really bad :S12:11
wikingso maybe i didn't12:11
-!- naywhayare [] has quit [Ping timeout: 246 seconds]15:32
-!- naywhayare [] has joined #shogun15:32
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]15:37
-!- sonne|work [~sonnenbu@] has left #shogun []16:45
-!- adoniscik [] has joined #shogun19:28
-!- blackburn [~blackburn@] has joined #shogun20:22
wikingblackburn: lol man20:27
wikingblackburn: have u heard about this stuff:20:27
wikingand btw: best one of the best PRs
wikingnote the bottom of the PR description:  Good to merge -- The Travis build passed (Details)20:28
wikinghere's a nice blog post about it20:29
wikingbut back to the travis thing20:29
wikingi'll try it now on my own repo20:30
wikingof shogun to set it up20:30
blackburnI do not understand CI :D20:30
wikingbut this PR testing sounds good20:30
wikingwell this travis shit is basically another type of buildbot20:30
wikingbut it's well integrated with github20:30
wikingso you have things like20:30
wikingonce you have a PR20:31
wikingit'll automatically try to build it20:31
wikingand test it for that matter20:31
wikinglooks nice20:31
blackburnokay I see20:32
wikingbut i want to test this PR for mpl :D20:34
wikinglooks hilarious20:34
wikingeven the dependencies are installing20:45
wikingbut i think i'll have to add more packages there as that's for an old shogun20:45
wikingwoah this one even takes mo20:46
wikingoooh noooo20:46
wikingi forgot a cd src20:46
wiking2nd try:!/vigsterkr/shogun/jobs/284351620:52
blackburn wiking it is totally awesome :D20:55
wikingblackburn: ahahahha20:55
wikingi want this for SHOGUN!20:57
wikingit's basically an extra buildbot20:58
wikingintegrated in github20:58
wikingheheh python_modular is building20:59
blackburnwiking: where is it run and why it is free? :)21:01
wikingok clang droped dead21:02
wikingblackburn: dunno :D21:02
wikingblackburn: let's be happy that it's still free21:02
wikingThis test series was run on a worker box sponsored by Engine Yard21:02
wikingok wtf is this:21:03
wikingIn file included from modshogun_wrap.cxx:9527:21:03
wiking2853/usr/lib/gcc/i686-linux-gnu/4.6/../../../../include/c++/4.6/cstring:101:3: error:21:03
wiking2854      declaration conflicts with target of using declaration already in scope21:03
wiking2855  memchr(void* __s, int __c, size_t __n)21:03
wiking2856  ^21:03
wiking2857/usr/include/string.h:95:14: note: target of using declaration21:03
wiking2858extern void *memchr (__const void *__s, int __c, size_t __n)21:03
wiking2859             ^!/vigsterkr/shogun/jobs/2843516/L285221:03
wikingthat's with clang21:05
wikinggcc 4.6 is still running21:06
wikingthat seems to be working till now21:11
wiking1943Executing your script (./configure && make && make test) took longer than 25 minutes and was terminated.21:30
blackburnwe are pretty big to compile in 25 mins21:35
wikingi'm gonna tweak around a bit (some says that make -j 4 helps to speed up compilation, even though there's only one core in the VM)21:36
wikingbut the other option is to only use certain interfaces in travis21:37 :Partial service outage22:11
wikingbattlestation doooown22:20
wikingMajor service outage22:20
wikingsomebody brokez z internet22:20
wikingWe are experiencing issues due to a DDOS attack, working hard to restore service...22:35
wikingand doubleROTFL for not using ipython till now22:35
blackburnis it cool?22:36
wikingyeah i like it22:36
wikingat least it has completion when i start hitting the <tab>22:37
wikingsf is hyper slow as well for me22:41
blackburnyoutube is down :)22:41
wikingwoah motherfucker22:44
wikingi've just realised that stupid macports22:44
wikingis keeping the distfiles if i do a port clean22:44
wikingmy distfiles dir is currently 6 gigs22:44
wikingwhat a waste...22:45
blackburnwhy not overwritten?22:45
wikingno i mean it keeps the tar.gz for all the installed ports22:46
blackburnI see22:46
blackburnokay I'm out of power22:47
blackburnsee you22:47
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]22:49
--- Log closed Fri Oct 19 00:00:17 2012