Open in new window / Try shogun cloud
--- Log opened Thu Apr 17 00:00:09 2014
-!- PirosB3 [] has joined #shogun01:05
shogun-buildbotbuild #254 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at  blamelist: tklein23 <>, Thoralf Klein <>01:07
thoralfHeikoS1: I've got something for you.01:08
thoralfHave a look on github.01:09
HeikoS1thoralf: wow that sounds nasty01:10
HeikoS1thoralf: I have to look into it tomorrow, currently doing something else, sorry01:10
thoralfHeikoS1: No worries.  Just that you know there's something funny waiting for you. :)01:11
HeikoS1thoralf: you always dig out interesting stuff! :)01:13
thoralfHeikoS1: Most of the time it happens when I try to be productive...01:14
thoralf... today I was just trying to fix buildbots memory issues ;)01:14
lisitsynI made d-ptrs 3x faster01:15
lisitsynHeikoS1: thoralf: ^01:16
lisitsynmay be we can employ this later in shogun01:16
thoralflisitsyn: Make it another 2x and we're impressed. ;)01:16
lisitsynokay basic test is01:16
lisitsyna class and d-ptr with an int01:17
lisitsynmethod call just does i++;01:17
lisitsynhere's what I got01:17
lisitsynmy thing is01:17
lisitsynAverage performance: 6316.20049 runs/s01:17
lisitsynand usual d-ptr is01:17
lisitsynAverage performance: 2009.62692 runs/s01:17
lisitsynthoralf: you won't make me give up ;)01:18
thoralflisitsyn: I didn't even try. ;)01:19
-!- newlido [~quassel@] has quit [Quit: No Ping reply in 180 seconds.]01:42
-!- newlido [~quassel@] has joined #shogun01:43
-!- lisitsyn [~qdrgsm@] has quit [Ping timeout: 276 seconds]01:51
-!- PirosB3 [] has quit [Quit: PirosB3]01:56
-!- newlido [~quassel@] has quit [Ping timeout: 258 seconds]02:03
HeikoS1thoralf: solved, no need for the PR02:35
thoralfHeikoS1: Hmm?02:37
HeikoS1see my comment02:38
thoralfYou fixed it?02:39
thoralfHeikoS1: I tried what you just did... this will fail. ;)02:40
HeikoS1thoralf: whats the problem?02:41
thoralfDon't remember.  I think it was double-freeing in other classes.02:42
thoralfBut we'll see tomorrow.02:42
HeikoS1but this removes the double ref-increase02:43
HeikoS1if others fail, then they are doing things wrong.02:43
HeikoS1haha, might be more of them than classes that do it correctly ;)02:44
HeikoS1thoralf: did you see the subset stack stuff=?02:44
-!- thoralf [] has quit [Ping timeout: 250 seconds]02:48
-!- abinash_panda [~androirc@] has joined #shogun03:00
shogun-buildbotbuild #678 of nightly_all is complete: Failure [failed compile test]  Build details are at  blamelist: Thoralf Klein <>, Wu Lin <>, tklein23 <>, pl8787 <>, Viktor Gal <>, Heiko Strathmann03:16
shogun-buildbot<>, liqiang <>03:16
-!- HeikoS1 [] has quit [Ping timeout: 240 seconds]03:18
-!- abinash_panda [~androirc@] has quit [Quit: AndroIRC - Android IRC Client ( )]03:18
-!- pfm___ [] has joined #shogun03:21
-!- pfm_ [] has quit [Ping timeout: 240 seconds]03:22
-!- pl8787 [9fe22b60@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]03:53
-!- pl8787_ [9fe22b60@gateway/web/freenode/ip.] has joined #shogun04:23
-!- soumyaC [uid15286@gateway/web/] has quit [Quit: Connection closed for inactivity]05:26
shogun-buildbotbuild #442 of osx1 - libshogun is complete: Failure [failed git]  Build details are at  blamelist: tklein23 <>, Thoralf Klein <>07:41
-!- andrew__ [32b86e8d@gateway/web/freenode/ip.] has joined #shogun08:01
andrew__hey all08:02
andrew__I've been MIA for some weeks- extremely busy with school.08:03
andrew__I would like to move forward ASAP with entrance tasks for GSOC if it isn't too late08:03
andrew__really it would make things much easier on me if it could be taken on faith that I understand version control/github workflows and will know c++ when the time comes, but I understand the need to assess potential students08:07
-!- thoralf [] has joined #shogun08:22
andrew__What is the status of issue #1889, which I had intended to work on before nikolis stepped in...08:24
andrew__presumably there is benchmarking and testing to do?08:24
andrew__anyway in lieu of leveraging the reputation of berkeley cs, the next 24 hours is when I'll have time to be doing entrance tasks so please let me know if anyone has ideas08:36
thoralfandrew__: 24 hours from now?08:37
thoralfThis seems very spontaneous... ;)08:37
andrew__yeah sorry... i've just been swamped and lo and behold the deadline is tomorrow...08:39
thoralfandrew__: Did you already start any entrance task?09:10
thoralfIs there pending an open one or have you already finished one?09:11
andrew__not really. I decided a few weeks ago to take on #1889 and #1873.09:18
andrew__Subsequently nikolis has apparently taken over 1889, which was a bit frustrating but so it goes I suppose09:19
andrew__bedtime for me - I'll check the channel in the morning09:20
-!- andrew__ [32b86e8d@gateway/web/freenode/ip.] has quit [Quit: Page closed]09:21
-!- thoralf [] has quit [Quit: Konversation terminated!]09:35
-!- newlido [~quassel@] has joined #shogun10:14
-!- PirosB3 [] has joined #shogun10:19
-!- dhruv13j [dhruv13j@gateway/shell/bnc4free/x-rsngwrlwzijytwjz] has joined #shogun10:23
-!- lambday [67157f51@gateway/web/freenode/ip.] has joined #shogun10:46
-!- newlido [~quassel@] has quit [Quit: No Ping reply in 180 seconds.]11:32
-!- newlido [~quassel@] has joined #shogun11:32
-!- lisitsyn [] has joined #shogun11:36
-!- newlido [~quassel@] has quit [Quit: No Ping reply in 180 seconds.]11:48
-!- newlido [~quassel@] has joined #shogun11:49
-!- PirosB3 [] has quit [Quit: PirosB3]11:49
-!- soumyaC [uid15286@gateway/web/] has joined #shogun12:04
-!- HeikoS1 [] has joined #shogun12:27
-!- HeikoS1 [] has quit [Ping timeout: 265 seconds]12:51
-!- HeikoS1 [] has joined #shogun13:07
-!- HeikoS1 [] has quit [Read error: Connection reset by peer]13:22
-!- HeikoS1 [] has joined #shogun13:45
lisitsynHeikoS1: !14:00
HeikoS1lisitsyn:  hi!14:00
lisitsynHeikoS1: we need to push kaggle thing ;)14:01
HeikoS1lisitsyn: yep I know14:02
lisitsynit seems that we basically stopped once started14:02
HeikoS1but first the other things that were on my list before :)14:02
HeikoS1I still think about it14:02
lisitsynHeikoS1: yeah sure14:02
HeikoS1just some delay ;)14:02
lisitsynHeikoS1: yeah that's how I work as well14:02
lisitsynI usually do things14:02
lisitsynbut sometimes year after :D14:02
lisitsynyou are better in this sense14:02
lisitsynHeikoS1: well anyway lets just be conscious :)14:03
lisitsynI mean kaggle is continuous stuff14:03
HeikoS1I am excited about this14:04
lisitsynso if we can't make it14:04
HeikoS1want to do it!14:04
HeikoS1currently fixing a serious bug in CList14:04
HeikoS1reference counting is done wrong14:04
HeikoS1and people made workarounds against this14:04
lisitsynHeikoS1: ok we need shared_ptr to stop this madness :D14:04
HeikoS1so now lots of segfaults14:04
HeikoS1lisitsyn: write a design draft!14:04
HeikoS1I would love this to stop14:04
lisitsynHeikoS1: I am trying to write up my thesis :D14:05
lisitsynone of the reasons why I am slow14:05
HeikoS1ah I see14:08
HeikoS1what is it about?14:08
lisitsynHeikoS1: well I wanted to make inductive transfer14:08
lisitsynwith dim reduction14:08
lisitsynjust find some latent space of both source and target domains14:08
lisitsynand transfer a model somehow14:08
lisitsynI am already late to conduct any serious research on that so probably would be a shitty thing :D14:09
-!- lambday [67157f51@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]14:20
-!- lambday [67157d51@gateway/web/freenode/ip.] has joined #shogun14:35
lambdayHeikoS1: hi!14:36
lambdayHeikoS1: I have been trying for a while but I think the matrix ^2 thing can't be done if we can't store the matrix1!14:36
lambdayHeikoS1: in that case, if we plan to choose within block direct estimation, we're stuck with blocksize max 100k14:38
-!- newlido [~quassel@] has quit [Ping timeout: 276 seconds]14:48
HeikoS1lambday: yes I think you are right14:54
HeikoS1ill task to dino about that later today14:54
HeikoS1but thats fine in fact14:54
HeikoS1one can just use more blocks14:54
lambdayHeikoS1: I think we should give a warning if the blocksize is way too huge and the var estimation method is within block direct estimation one14:54
HeikoS1lambday: yeah good idea, maybe even check for memory allocation errors14:55
HeikoS1but anything over say 8GB should be warned about :)14:55
HeikoS1lambday: I am currently trying to fix this bug that thoralf found14:56
HeikoS1thats very tricky14:56
HeikoS1touched something and now lots of things fail14:56
HeikoS1but CList did wrong reference counting14:56
HeikoS1so we have to do this :(14:56
HeikoS1anf found some other errors on the fly14:57
-!- lambday_ [67157f51@gateway/web/freenode/ip.] has joined #shogun14:58
lambday_HeikoS1: sorry I got dc...14:58
lambday_HeikoS1: extensive unit-testing has to be done14:58
lambday_HeikoS1: please create issues for unit-testing then we can also help14:59
HeikoS1lambday_:  yeah14:59
HeikoS1Yeah I will do once I understand whats going on14:59
-!- lambday [67157d51@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]14:59
-!- lambday_ is now known as lambday15:00
-!- pfm___ is now known as pfm_15:22
-!- lambday [67157f51@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]15:37
-!- prams [~pratheek@] has joined #shogun15:37
-!- newlido_ [~quassel@] has joined #shogun16:15
-!- khalednasr [~k.nasr92@] has joined #shogun16:18
-!- khalednasr [~k.nasr92@] has quit [Quit: Leaving]16:25
-!- PirosB3 [] has joined #shogun17:45
-!- darshanime [6f5d05c2@gateway/web/freenode/ip.] has joined #shogun17:59
-!- PirosB3 [] has quit [Quit: PirosB3]18:00
-!- HeikoS1 [] has quit [Quit: Leaving.]18:01
darshanimehello guys.18:01
darshanimei am new here.18:01
darshanimei am learning Machine Learning18:02
darshanimeI just wanted to know, when can I be ready to contribute ?18:03
darshanimehow do I know when i can do things ?18:03
darshanimealso what is the first step if I wish to do so?18:03
darshanimeshould I fork the source code and try to decipher it from head to toe18:03
-!- iglesiasg [] has joined #shogun18:28
-!- mode/#shogun [+o iglesiasg] by ChanServ18:28
pl8787_darshanime: you may use shogun lib first .18:33
darshanimestudy that lib ?18:33
pl8787_darshanime: just use the ML model in it18:34
pl8787_such as svm or gp18:34
darshanimeright. then ?18:34
-!- abinash_panda [01261095@gateway/web/freenode/ip.] has joined #shogun18:35
pl8787_If you want to make contribution on shogun, build the source code and add some code in it.18:35
-!- iglesiasg [] has quit [Quit: Leaving]18:36
darshanimealso suggest some IRC client18:36
pl8787_I think write some document is always a start for a newer.18:37
pl8787_I use web page.18:37
-!- HeikoS [] has joined #shogun18:54
-!- mode/#shogun [+o HeikoS] by ChanServ18:54
-!- lisitsyn1 [~qdrgsm@] has joined #shogun18:55
-!- lisitsyn [] has quit [Disconnected by services]18:56
-!- lisitsyn1 is now known as lisitsyn18:56
-!- mode/#shogun [+o lisitsyn] by ChanServ18:56
-!- lisitsyn1 [] has joined #shogun18:57
-!- newlido_ [~quassel@] has quit [Remote host closed the connection]19:08
-!- kislay|exams [~kislay@] has joined #shogun19:43
-!- sunil1337 [cb7c9722@gateway/web/freenode/ip.] has joined #shogun19:46
-!- PirosB3 [] has joined #shogun19:52
-!- lambday [67157f51@gateway/web/freenode/ip.] has joined #shogun20:01
lambdayHeikoS: by today I am trying to take care of the sum methods... is it okay it I take a look at this when I work on streaming MMD issue?20:01
-!- lisitsyn1 was kicked from #shogun by lisitsyn [lisitsyn1]20:21
@lisitsynHeikoS: hey20:24
@lisitsynI'll let you know if I die/lose connection/anything else ;)20:24
-!- kislay|exams [~kislay@] has quit [Quit: Leaving]20:28
lambdaylisitsyn: you kicked yourself out?20:32
@lisitsynlambday: yeah20:32
lambdaylisitsyn: that's.... something :D20:32
@lisitsynlambday: kicked *other me * out20:32
lambdaylisitsyn: multiple "me".. parallel universe :D20:32
lambdaymultithreaded universe :-/20:32
@lisitsynwell forgot to turn off irc at job20:33
@lisitsynI am lazy to set up any kind of remote desktop20:34
@lisitsynlambday: while working on blog post as you ordered20:35
@lisitsynI came with something else20:35
lambdaylisitsyn: not ordered, my lord! humbly requested :D20:35
lambdaylisitsyn: sounds intriguing :D20:36
@lisitsynlambday: okay imagine d pointer20:36
@lisitsynbut allocated in some buffer20:36
@lisitsynso it is embedded to the class20:36
@lisitsynbuffer is not in heap but exactly in the class20:36
@lisitsynlike char storage[10]20:37
@lisitsynlambday: guess the speedup ;)20:37
lambdaybut limitations?20:37
@lisitsynlambday: yes20:37
lambdayI mean, you're planning to do it on stack?20:37
@lisitsynyou provide exact size20:37
@lisitsynmain limitation - you provide exact size in declaration20:37
@lisitsynso if you add new member and your 'implementation' class doesn't fit there anymore20:38
@lisitsynyou are doomed to alter declaration20:38
@lisitsynit is not on stack but 'inside' of the class20:38
@lisitsynwhere we provide d-ptr20:38
lambdayoh got it!20:38
@lisitsynlambday: on some simple benchmark it was 3x faster20:39
@lisitsynthe benchmark is counter20:39
@lisitsyneach method call increases it20:39
lambdaycounter of what?20:40
@lisitsynlambday: well just int20:40
@lisitsynwe put an int to 'implementation'20:40
@lisitsynand increase it20:40
lambdaylisitsyn: nono I mean it counts what?20:41
@lisitsynnah nothing20:41
@lisitsynjust increases it :D20:41
lambdayoh you mean number of times a method is called in some fixed duration?20:41
@lisitsynit is called a lot of times20:41
@lisitsynso we have some estimate20:41
@lisitsynhow fast it is20:41
@lisitsynlambday: basically it is 50ns vs 12ns20:42
@lisitsynper call20:42
@lisitsynlambday: that's a matter of being cache-friendly20:43
lambdayI had no idea that it fasten things up to this large factor!20:43
@lisitsynlambday: let me find you another example20:44
@lisitsynyou'd like it20:44
@lisitsynjust a sec20:44
lambdaylisitsyn: checking..20:47
@lisitsynlambday: it is pretty big so if tl;dr I can tell you about some speedup he mentioned there20:47
lambdayhmm... yeah I get the idea! but didn't think that the different can be this big!20:50
@lisitsynlambday: yeah so basically he made things that he iterate anyway local20:51
@lisitsyncaching again20:51
lambdaylisitsyn: I am working on something - this encourages me to check/try out something here as well!20:52
@HeikoSlambday: yeah do that. Currently every<thing works20:56
@lisitsynlambday: okie I need a hour to finish my post on these revisited d-ptrs ;)20:56
lambdayHeikoS: alrighty! hope to send PR for this sum thing by tonight!20:57
lambdaylisitsyn: yeah! please let us know when you finish!20:57
@HeikoSlambday: nice!20:57
lambdaylisitsyn: I am trying to check on some benchmark with some cache-friendly implementation for my task as well :D20:58
@lisitsynlambday: you can actually measure cache misses20:58
sunil1337lambday: had worked on it this semester :)21:10
-!- andrew__ [6cdd1305@gateway/web/freenode/ip.] has joined #shogun21:12
lambdaysunil1337: wow this sounds interesting :D let me try this out on my code :D21:12
sunil1337lambday: you'll be amazed! :)21:14
andrew__is it too late to do gsoc entrance tasks? I realize the deadline is in a few hours21:27
-!- thoralf [] has joined #shogun21:27
@HeikoSlambday: good proper design going on there!21:28
lambdayHeikoS: in the dot product interface you mean?21:29
@HeikoSlambday: that too but meant the kernel sums21:29
lambdayHeikoS: yeah I was trying to manage everything using one vec but now trying out different things21:29
-!- prams [~pratheek@] has quit [Remote host closed the connection]21:31
andrew__hey thoralf, so...21:32
-!- sunil1337 [cb7c9722@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]21:38
andrew__so I'm looking at the current state of #188921:39
andrew__it looks this never get merged, and is missing benchmarks?21:39
andrew__but hey if it's too late, that's ok21:44
-!- khalednasr [~k.nasr92@] has joined #shogun21:51
-!- andrew__ [6cdd1305@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]22:03
-!- thoralf [] has quit [Ping timeout: 276 seconds]22:09
-!- PirosB3 [] has quit [Quit: PirosB3]22:10
-!- andrew [32b86e8d@gateway/web/freenode/ip.] has joined #shogun22:17
-!- thoralf [] has joined #shogun22:29
-!- thoralf [] has quit [Client Quit]22:31
-!- kislay|exams [~kislay@] has joined #shogun22:32
-!- HeikoS1 [] has joined #shogun22:34
-!- HeikoS [] has quit [Ping timeout: 276 seconds]22:36
@lisitsynlambday: still here?22:40
lambdaylisitsyn: yeah!22:41
lambdaylisitsyn: checking :D22:43
-!- shogun-notifier- [] has joined #shogun22:43
shogun-notifier-shogun: pl8787 :develop * af9bfb7 / tests/unit/kernel/
shogun-notifier-shogun: Fix CustomKernel-unittest bug.22:43
shogun-notifier-shogun: pl8787 :develop * 3227d9e / src/shogun/kernel/Kernel.cpp:
shogun-notifier-shogun: Fixed Kernel.cpp Warning.22:43
shogun-notifier-shogun: Heiko Strathmann :develop * 1cbe21a / src/shogun/kernel/Kernel.cpp,tests/unit/kernel/
shogun-notifier-shogun: Merge pull request #2164 from pl8787/indexfeatures22:43
shogun-notifier-shogun: Fix CustomKernel-unittest bug.22:43
shogun-notifier-shogun: Wu Lin :develop * ff25e97 / examples/undocumented/libshogun/variational_approx_example.cpp:
shogun-notifier-shogun: fixed the file path issue22:44
shogun-notifier-shogun: Wu Lin :develop * d5c943f / / (31 files):
shogun-notifier-shogun: Merge remote-tracking branch 'upstream/develop' into develop22:44
shogun-notifier-shogun: Wu Lin :develop * 3f10d2e / examples/undocumented/libshogun/CMakeLists.txt:
shogun-notifier-shogun: update the upstream22:44
shogun-notifier-shogun: Wu Lin :develop * 98e5ba6 / examples/ (3 files):
shogun-notifier-shogun: add the description22:44
shogun-notifier-shogun: Heiko Strathmann :develop * 7c74f4a / examples/ (4 files):
shogun-notifier-shogun: Merge pull request #2161 from yorkerlin/develop22:44
shogun-notifier-shogun: fixed the file path issue for the variational example22:44
@lisitsynlambday: I wanted to finish it faster so if you notice anything wrong let me know ;)22:45
-!- lambday_ [67157f51@gateway/web/freenode/ip.] has joined #shogun22:46
lambday_lisitsyn: absolutely... reading :)22:46
-!- lambday [67157f51@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]22:47
shogun-buildbotbuild #2216 of bsd1 - libshogun is complete: Success [build successful]  Build details are at
-!- andrew [32b86e8d@gateway/web/freenode/ip.] has quit [Ping timeout: 240 seconds]22:51
-!- HeikoS1 [] has quit [Quit: Leaving.]22:52
lambday_lisitsyn: this is great post man!!22:53
@lisitsynlambday_: cool thanks22:53
lambday_yeah really liked it...22:53
@lisitsynlambday_: I mostly care about the idea :D22:53
@lisitsynwant to find a way to make it more usable22:53
lambday_by great I meant that mostly :D22:53
@lisitsynas efficiency is cool22:53
@lisitsynlambday_: mumbai! :D22:58
@lisitsyngoogle analytics22:58
lambday_lisitsyn: huh?22:58
@lisitsynlambday_: probably detected you22:58
@lisitsynlambday_: well I have google analytics attached to the website ;)22:59
lambday_hehehehe :D22:59
lambday_you'll have quite regular visits now22:59
lambday_I've been following you ;)22:59
-!- kislay|exams [~kislay@] has quit [Quit: Leaving]23:25
-!- darshanime [6f5d05c2@gateway/web/freenode/ip.] has quit [Quit: Page closed]23:32
lambday_lisitsyn: there?23:41
@lisitsynlambday_: yes23:55
lambday_lisitsyn: this code causes memory leak on my machine23:56
lambday_any clue?23:56
@lisitsynlambday_: which code?23:56
lambday_int main()23:56
lambday_#pragma omp parallel23:56
lambday_return 0;23:57
-!- kislay [~Abhijeet@] has joined #shogun23:57
lambday_lisitsyn: this ^23:57
@lisitsynthat's ok23:57
@lisitsyndon't care about that23:57
lambday_lisitsyn: why not?23:57
lambday_1,008 bytes in 3 blocks are possibly lost in loss record 3 of 4 ==26711==    at 0x4A081D4: calloc (in /usr/lib64/valgrind/ ==26711==    by 0x3049811C44: _dl_allocate_tls (dl-tls.c:296) ==26711==    by 0x304A408862: pthread_create@@GLIBC_2.2.5 (allocatestack.c:580) ==26711==    by 0x3BFAE08DD5: ??? (in /usr/lib64/ ==26711==    by 0x4008B7: main (opemnp_test.cpp:79)23:57
@lisitsynlambday_: that's openmp thing23:57
@lisitsynbasically false alarm23:58
@lisitsynit is known issue23:58
@lisitsynlambday_: it allocates some pool23:58
@lisitsynwhich is being freed much much later23:58
@lisitsynshared across apps or so23:58
lambday_not cool man!23:58
@lisitsynnah that's ok23:58
@lisitsyndon't bother ;)23:59
lambday_lisitsyn: thanks :D23:59
--- Log closed Fri Apr 18 00:00:10 2014