Open in new window / Try shogun cloud
--- Log opened Fri May 17 00:00:05 2013
@sonney2kI never think hard00:00
shogun-buildbotbuild #1125 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Sergey Lisitsyn <>00:00
@lisitsynI hardly think hard00:00
@sonney2kI am just too soft00:00
@lisitsynthen I fall asleep00:00
@lisitsynsonney2k: okay one solution00:01
@lisitsyna) add a base class like DenseFeaturesDataKeeper00:02
@lisitsynb) inherit a template class from it00:02
@lisitsync) add constructors for densefeatures for all supported types00:02
@lisitsynwhere we create appropriate data keeper00:02
@sonney2kyeah void* would be a solution00:02
van51sonney2k: I think the problem is when you call DynArray::insert_element and the array is empty00:02
@sonney2kvan51, very well possible00:03
@lisitsynsonney2k: not really voide00:03
van51after it has at least one element it's ok00:03
@sonney2kvan51, plesae write a unit test and fix it00:03
@lisitsynsonney2k: let me try that this weekend and we will see00:04
@sonney2kvan51, I think it is also troublesome to call insert_* with an index >= number_of_elements00:04
van51sonney2k: ok, but can it wait until tomorrow?00:04
@lisitsynsonney2k: do you expect really significant speedup with that?00:04
@sonney2klisitsyn, with what?00:05
@lisitsynsonney2k: with no templates exposed00:05
@sonney2ktemplate removal?00:05
@sonney2klisitsyn, just try it00:05
@sonney2kremove all %template from .i00:05
van51sonney2k: I was thinking about that, but I didn't see any index checks on the array classes00:06
van51so I thought it was up to the user to do it correctly00:06
@sonney2kvan51, yeah it was intended like this00:06
van51sonney2k: I'll try and fix it tomorrow00:08
van51and will provide some unit tests00:08
@lisitsynsonney2k: what do we lost this way is compile time typechecking00:09
@lisitsynwhich is relevant for C++00:09
@lisitsynI mean if I add that thing we get runtime typechecking00:12
@lisitsynbut if you do sth like set_feature_vector(some sgvector of wrong type)00:12
@lisitsynyou get no error in C++00:12
@iglesiasggood night guys, I am falling asleep on the keyboard :D00:13
@lisitsyniglesiasg: same here actually00:13
@lisitsyngood night00:13
-!- iglesiasg [d58f3231@gateway/web/freenode/ip.] has quit [Quit: Page closed]00:13
@sonney2kvan51, thanks00:15
van51sonney2k: no need.. I introduced it, I should fix it :p00:16
-!- travis-ci [] has joined #shogun00:16
travis-ci[travis-ci] it's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []00:16
-!- zxtx [] has joined #shogun00:20
shogun-buildbotbuild #1126 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: <>00:37
-!- goldbug [] has joined #shogun00:51
-!- van51 [] has quit [Ping timeout: 252 seconds]00:52
-!- foulwall [~foulwall@2001:da8:215:503:b56d:503:d77b:effa] has joined #shogun01:33
-!- FSCV [~FSCV@] has quit [Quit: Leaving]01:55
-!- foulwall [~foulwall@2001:da8:215:503:b56d:503:d77b:effa] has quit [Remote host closed the connection]02:00
-!- shogun-notifier- [] has quit [Quit: transmission timeout]02:10
-!- Netsplit *.net <-> *.split quits: @sonney2k, flxb, naywhayare, @wiking, sijin02:33
-!- goldbug [] has quit [Ping timeout: 253 seconds]02:36
-!- Netsplit over, joins: sijin, flxb02:38
-!- sijin [~smuxi@] has quit [Excess Flood]02:38
-!- naywhayare [] has joined #shogun02:52
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun02:52
-!- ServerMode/#shogun [+o wiking] by hobana.freenode.net02:52
-!- sonney2k [] has joined #shogun02:52
-!- ServerMode/#shogun [+o sonney2k] by hobana.freenode.net02:52
-!- foulwall [~foulwall@2001:da8:215:503:1485:6985:a389:96a2] has joined #shogun03:24
-!- nube [~rho@] has quit [Quit: Leaving.]03:34
shogun-buildbotbuild #397 of nightly_default is complete: Failure [failed test]  Build details are at
-!- zxtx [] has quit [Ping timeout: 256 seconds]04:24
-!- votjakovr [] has joined #shogun05:02
-!- nube [~rho@] has joined #shogun05:28
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]06:06
-!- nube [~rho@] has quit [Quit: Leaving.]06:29
-!- nube [~rho@] has joined #shogun06:29
-!- foulwall [~foulwall@2001:da8:215:503:1485:6985:a389:96a2] has quit [Ping timeout: 245 seconds]06:40
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun09:06
-!- iglesiasg [] has joined #shogun09:06
-!- mode/#shogun [+o iglesiasg] by ChanServ09:06
-!- zxtx [] has joined #shogun09:26
-!- hushell [] has joined #shogun09:31
-!- foulwall [~foulwall@2001:da8:215:6100:30d1:8741:6f9a:88b5] has joined #shogun09:46
foulwallmorning sonney2k.09:46
-!- hushell [] has quit [Quit: leaving]10:17
-!- nube [~rho@] has quit [Ping timeout: 264 seconds]10:18
-!- iglesiasg_ [c1934d18@gateway/web/freenode/ip.] has joined #shogun10:33
-!- hushell [] has joined #shogun10:36
sonne|workfoulwall: morning!10:41
sonne|workfoulwall: please extract all the demos from current shogun-web git and put it in a separate django one10:42
sonne|workwe can then create shogun-demos10:42
sonne|work(the repository)10:42
foulwallsonne|work: ok!10:42
-!- iglesiasg [] has quit [Quit: Lost terminal]10:44
-!- hushell [] has quit [Ping timeout: 256 seconds]10:51
-!- iglesiasg_ [c1934d18@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]10:53
-!- nube [~rho@] has joined #shogun10:55
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]10:57
@wikingsonne|work: ping? shall we have a c4.5 and c5.0 implementation in shogun... as i'm now starting from scratch the random forest implementation so i realised that it's good to have decision trees implemented10:58
@wikingsonne|work: moreover: do anybody is planning to do this in gsoc? so just that i dont take somebody's project...10:58
sonne|workwiking: well there is no OSS implementation of C4.5 and C5.0 code is only slightly better den cat /dev/unrandom10:59
sonne|workwiking: but yes feel free10:59
sonne|workwiking: you could port weka's j45 over to shogun11:00
sonne|workmight be the most readable...11:00
@wikingsonne|work: yeah i want first random forest ASAP11:00
@wikingbut then i'll do some other tree implementation as well11:00
sonne|workwell random forest == some tree impl but just bagged on top11:01
sonne|workI think the correct approach would be11:01
sonne|workto do a BaggedClassifier11:01
sonne|workthat just trains the same classifier a couple of times11:01
sonne|workusing some sampling scheme11:02
sonne|workof the training data points11:02
sonne|workthen you can fit in *any* other classifier11:02
sonne|worklike svm / tree / knn11:02
sonne|workand you have your random forest as subset11:02
@wikingsounds reasonable11:02
@wikingmmm my battery is running low11:03
-!- hushell [] has joined #shogun11:13
-!- ozan [] has joined #shogun11:22
-!- ozan [] has quit [Remote host closed the connection]11:23
sonne|worklisitsyn: heya11:27
@lisitsynsonne|work: my docs for visa are ready11:27
sonne|worklisitsyn: time to get your hands dirty you vodka addict :D11:27
@lisitsynlets see how your embassy works11:27
* sonne|work is in some kind of strange moooooooooood today11:27
sonne|workpush push11:27
sonne|workwhy must it be that difficult...11:28
@lisitsynsonne|work: because we are barbarians11:29
sonne|workvodka drinking babarians11:29
sonne|workriding polar bears11:29
@lisitsynsonne|work: at least I don't have to proceed through interview11:31
@lisitsynsonne|work: like for usa11:31
@lisitsynsonne|work: if I am going to visit that mentor summit I should take care of visa right in august I think11:31
-!- foulwall [~foulwall@2001:da8:215:6100:30d1:8741:6f9a:88b5] has quit [Remote host closed the connection]11:31
@lisitsynbecause it is even worse :D11:32
@lisitsynsonne|work: what about my hands and dirt? any suggestion?11:33
@lisitsynI am not available anyway though :D11:33
sonne|worknot available?11:34
@lisitsyntrying to work right now11:34
@lisitsynsonne|work: trying to make multiagent planning system work :D11:35
sonne|workkeep trying :D11:36
@lisitsynsonne|work: that's the definition of hype11:37
@lisitsynnot the deep learning11:37
@lisitsynMAS is pure hype11:37
@lisitsynfrom 90s :D11:37
@lisitsynbut we live 20 years back from the other world so it is here11:37
sonne|worklet me add the 'big data', 'data science' crappy do to this list11:38
@lisitsynsonne|work: haha big data yes11:39
sonne|workwhat is big11:40
sonne|workbut most certainly my data is bigger than your data :P11:40
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun11:40
@lisitsynsonne|work: everyone thinks he is a facebook :D11:41
sonne|worklisitsyn: hey
sonne|workbabababbbiiiig daaaaaaaaaaaaataaaa!11:53
@lisitsynsonne|work: huge data is next12:02
-!- foulwall [~foulwall@2001:da8:215:503:e934:af25:452:c32c] has joined #shogun12:13
-!- nube [~rho@] has quit [Quit: Leaving.]12:14
-!- iglesiasg [] has joined #shogun12:22
-!- mode/#shogun [+o iglesiasg] by ChanServ12:22
-!- foulwall [~foulwall@2001:da8:215:503:e934:af25:452:c32c] has quit [Remote host closed the connection]12:26
-!- lambday [3ba2174d@gateway/web/cgi-irc/] has joined #shogun12:27
-!- foulwall [] has joined #shogun12:27
-!- mdragan [8d55007a@gateway/web/freenode/ip.] has joined #shogun12:53
-!- iglesiasg [] has quit [Quit: Lost terminal]12:59
lambdaylisitsyn: hi13:09
@lisitsynlambday: hey13:09
lambdayI did a rebase... and after that doing make-unittests gives an error.... apparently CMath:init_random has been changed..13:10
lambdaywait let me paste13:10
@lisitsynyeah that would be better13:10
lambdayIn function `shogun::CMath::init_random(unsigned int)':13:10
lambdayIn function `shogun::CMath::init_random(unsigned int)':13:10
lambdayshit... wait!13:10
lambdayMath.h:458: undefined reference to `shogun::CMath::rand_state13:11
lambdaybut now its been replaced with something else, I just checked13:11
foulwallsonne|work: sonney2k, could I start a django app for each demo, or make a classification such as supervised, unsupervised etc and start apps for each of it , or just start a 'demo' app and include all the demos in it?13:11
@lisitsynlambday: did you try to clean up that thing? sounds like something related to recent changes in random13:12
lambdaywhy it would give this error then? I freshly did a make and my src shows new CMath13:12
lambdaylisitsyn: I did git pull --rebase upstream develop, and the make clean and make13:13
lambdaylisitsyn: yes... seems so13:13
@lisitsynI think not all things are properly updated in tests13:13
lambdaylisitsyn: I think the unit-tests just makes a init_random() call... whose definition in CMath has been changed... should be fine, right?13:14
@lisitsynlambday: so did clean help13:15
lambdaylisitsyn: nope! :(13:15
lambdayclueless :-/13:15
lambdayshall I try cloning somewhere else and then rebase and test again?13:16
@wikinglambday: actually init_random did not change13:16
@lisitsynlambday: no makes no sense for me13:17
@wikingit's just PRNG behind it is changed13:17
@wikingbut the API is the same13:17
lambdaywiking: hmm... the error shows undefined reference to CMath::rand_state, I checked and it was there previously, now changed13:18
@wikingbut then you have a problem13:18
@wikingthat you dont have the latest CMath.h/cpp13:18
lambdaywiking: I do :( I checked and it shows the new init_random()13:20
lambdaytotally weird :-/13:20
lambdaymay be I'm doing something wrong - let me check again13:20
sonne|workfoulwall: I single demo app is totally sufficient13:22
sonne|workfoulwall: you just need to 'mirror' the urls we have currently in your stuff13:22
foulwallsonne|work: ok...13:22
sonne|worklambday: maybe better do a git clean -dfx (will erase anything not in git)13:23
sonne|workrecompile / install / run tests13:23
lambdaysonney2k: okay.. thanks... trying13:24
lambdaysonney2k: worked :D yippie13:28
-!- Netsplit *.net <-> *.split quits: mdragan, vgorbati13:31
-!- Netsplit over, joins: mdragan, vgorbati13:31
lambdaysonne|work: thanks a lot :)13:33
-!- lambday [3ba2174d@gateway/web/cgi-irc/] has quit [Quit: - A hand crafted IRC client]13:45
-!- nube [~rho@] has joined #shogun13:47
-!- nube [~rho@] has quit [Quit: Leaving.]13:59
-!- iglesiasg [] has joined #shogun13:59
-!- mode/#shogun [+o iglesiasg] by ChanServ14:00
-!- hao_ [45775f6b@gateway/web/freenode/ip.] has joined #shogun14:06
hao_Good morning everyone.14:07
sonne|workhey there14:07
hao_well I am an applicant of google SoC.14:08
hao_And I contacted with my mentor, some days ago. He told me to make some contribution on the issues on git14:09
hao_But only from the name of the issues, I am not sure which one will fit me and I can really make some help.14:10
hao_And I choosed one, but found there is already some one working on it.14:11
hao_So I find it may be better choose one issue after asking around.14:12
hao_Would anyone here please recommend me some issues or works to work with?14:12
@iglesiasghao_: Is there any of them which interests you?14:13
hao_Import eliptic curve functions, but lambday is working on it.14:15
hao_This is my proposal. May help you to know about me.14:15
@lisitsynyour competitor iglesiasg!14:16
-!- nube [~rho@] has joined #shogun14:16
hao_Well, what's that..competitor?14:21
@lisitsynhao_: you are applying for the same task both14:22
@iglesiasghao_: we are both applying for the same project14:22
-!- nube [~rho@] has quit [Quit: Leaving.]14:22
hao_Implement metric learning algorithms with applications to metagenomics ?14:23
@iglesiasghao_: yes14:24
-!- iglesiasg_ [c1934d18@gateway/web/freenode/ip.] has joined #shogun14:24
hao_what I can say.14:24
hao_So each project can only have one student?14:25
@iglesiasgnot in general, but I believe that it is the case of the one we are concerned now14:28
-!- iglesiasg_ [c1934d18@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]14:28
hao_Ok, I see. Thank you.14:30
vgorbatilisitsyn: I see you have merged a PR so fast:) Should I take now a task from HeikoS as we discussed?14:31
@lisitsynvgorbati: yeah please ask them for some task14:31
foulwallsonne|work:, I've finished and created a repo.14:32
@lisitsynwhat about it?14:32
vgorbatilisitsyn: I guess I need a bit more info about what needs to be done:)14:33
sonne|workfoulwall: ok nice I will move that then to the official shogun toolbox repository. now I only need a patch that removes the demos from shogun-web14:33
@lisitsynvgorbati: ask sonne|work but in general it should be something like just remove the method and resolve all errors14:33
foulwallsonne|work:  ok I'll send a pr on shogun-web14:34
sonne|workfoulwall: thanks14:34
vgorbatisonne|work: so I will need to find all places where this method may be used and replace it by appropriate method calls?14:35
sonne|workfoulwall: I think we should start to use some of shogun's data sets for illustration14:35
sonne|workvgorbati: drop die this method14:35
sonne|workno one should use it14:35
foulwallsonne|work: ok14:35
sonne|workvgorbati: shoot, remove!14:35
vgorbatisonne|work: :)ok, understood:)14:35
vgorbatiI guess it won't be that hard:)14:36
sonne|workfoulwall: you could add support to load the data sets under toy/ from there14:37
sonne|workthen just plot the first 2dims of that data set14:37
sonne|workand then any method could use them14:37
sonne|workfoulwall: like you have now 'artificial' genatore14:37
sonne|workmanually clicked14:38
sonne|worknow in addition data loaded14:38
sonne|workfoulwall: so you need some kind of re-usable component that you can display on any website for this14:38
foulwallsonne|work: you mean data in
sonne|workfoulwall: I like your -1, +1 plot the most so all could just use this14:38
sonne|workfoulwall: yes14:38
foulwallsonne|work: Ok! I'll do have a look at this repo tonight and do it tomorrow14:39
sonne|workformat of that is not yet standardized but will happen this gsoc :)14:39
sonne|workfoulwall: thanks a lot14:39
sonne|workfoulwall: but you got the idea yes? some say drop down list - to let the user choose to manually do input / select a data file / randomly generate data14:40
foulwallsonne|work: I'll send some little patchs on the existing demo tonight. like when user move the mouse on the curve and show the coordinate14:41
foulwallsonne|work: I see.14:41
hao_Well. Is that possible to change the gsoc project? Because I guess the project I choosed has been filled.14:43
@lisitsynhao_: there is no project with no student at all :)14:44
hao_the result has almost been decided or not?14:45
@lisitsynhao_: well we definitely have some list of top candidates14:47
@lisitsynwe have to come up with the final decision on 22th I think14:48
hao_I see. kind of late to begin.14:48
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Quit: Page closed]14:50
hao_But I still really want to try my best to make different14:50
hao_any advice?14:50
@lisitsynis there any issue you'd like to resolve?14:51
sonne|workfoulwall: cool!14:52
sonne|workfoulwall: regarding the components stuff - what I have in mind is some minimal set of 'widgets' that we can use for almost any example - we simple want DRY'ness to high14:53
foulwallsonne|work: Ok, let me think how to do it tonight.14:53
sonne|workhao_: or any algorithm you want to do or so14:54
sonne|workfoulwall: we can always discuss about this - you should not be alone on that14:54
foulwallsonne|work: Ok, always online, never disappear again:)14:55
foulwallsonne|work: demo wiper PR are sent, have a look14:58
hao_Well I am reading the issues lists, but not sure which one to go. Actually I'd like to implement an algorithm.14:58
@iglesiasglunch time in Spain :D, see you later people15:00
-!- iglesiasg [] has quit [Quit: leaving]15:00
sonne|workhao_: then just chose $ALGORITHM and contribute it to shogun15:02
sonne|workthere is more ML out there than we could ever implement.15:02
-!- hao_ [45775f6b@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]15:09
-!- gsomix [] has joined #shogun15:24
sonne|workgsomix: hey there!15:25
gsomixsonne|work: too hot in Samara... how are you? :)15:27
sonne|workI plan to go swimming this evening15:27
sonne|workwater should be above 16 C15:27
sonne|workand it is hot here too (27 C)15:28
sonne|workgsomix: how is it going any progress on the line reader?15:29
sonne|workgsomix: might be that foulwall is our first customer :D15:29
gsomixsonne|work: oh, ok. at this evening, as planned.15:30
sonne|workfoulwall: gsomix intends to do data readers for shogun15:30
gsomixworking at ray-tracing now15:30
gsomixfoulwall: hi ^^15:31
foulwallgsomix: :)15:31
sonne|workgsomix: real world ray tracing or CG stuff?15:31
gsomixsonne|work: not CG, for optical problems15:32
sonne|workgsomix: I was expecting that :D15:32
gsomixbtw, I know how to design simple reflectors.15:34
gsomixI'm cool optician, lol. :]15:34
sonne|workgsomix: when I grow old I will order glasses from you!15:35
@lisitsynsonne|work: still having good vision? lucky15:37
@lisitsynmy eyesight is going blurry day to day :D15:38
sonne|worklisitsyn: I think I am down from 200% -> 100%15:40
@lisitsynsonne|work: were you an eagle before?15:40
sonne|workyes lost lots of feathers too15:40
sonne|workso I am on a no-fly list now15:41
-!- hao [d8a55f4b@gateway/web/freenode/ip.] has joined #shogun16:04
-!- hao [d8a55f4b@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:17
-!- foulwall [] has quit [Remote host closed the connection]16:36
-!- gsomix [] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]16:45
-!- sonne|work [] has quit [Quit: Leaving.]16:49
-!- foulwall [~foulwall@2001:da8:215:c252:e9b3:9bb0:e7cb:be49] has joined #shogun17:21
-!- nube [~rho@] has joined #shogun17:22
-!- nube [~rho@] has quit [Read error: No route to host]17:29
-!- nube [~rho@] has joined #shogun17:31
-!- FSCV [~FSCV@] has joined #shogun17:37
-!- votjakovr [] has quit [Quit: Leaving]17:47
-!- FSCV [~FSCV@] has quit [Ping timeout: 248 seconds]17:50
foulwallnight, it's too late in the night, I'll work more time tomorrow. happy weekend!18:14
-!- foulwall [~foulwall@2001:da8:215:c252:e9b3:9bb0:e7cb:be49] has quit [Remote host closed the connection]18:14
-!- FSCV [~FSCV@] has joined #shogun18:51
@sonney2kfoulwall good night!19:24
-!- pickle27 [] has quit [Quit: Leaving]19:27
-!- hushell [] has quit [Quit: leaving]19:42
-!- FSCV [~FSCV@] has quit [Ping timeout: 256 seconds]19:46
-!- votjakovr [] has joined #shogun20:30
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun20:37
* sonney2k yay that was excellent20:47
@lisitsynsonney2k: what was20:49
-!- van51 [] has joined #shogun20:53
@sonney2klisitsyn, the water21:02
@sonney2klisitsyn, lake mueggelsee :D21:03
@lisitsynsonney2k: ahh the mueggelsee21:03
@sonney2klisitsyn, the famous one21:03
@lisitsynsonney2k: never heard about it21:03
@sonney2klisitsyn, there is a german movie now about it 'shark hunt in lake mueggelsee'21:03
@sonney2kcool trash21:03
@sonney2ksnake mueller the shark hunter21:04
@lisitsynsonney2k: I mean we could go for breaking the things to stuff like shogun-liblinear etc21:20
@lisitsynI have no opinion on that21:22
@sonney2klisitsyn, there are a couple of issues21:22
@sonney2klots of stuff in shogun is very much connected21:22
@sonney2kvan51, good evening!21:22
@sonney2kfor example features & preprocessors21:22
@sonney2kkernels <-> features21:22
-!- van511 [] has joined #shogun21:22
@sonney2kkernels <-> distances21:22
@sonney2kvan511, good evening you too :D21:22
van511so what were you saying? :)21:23
@sonney2kvan511, we are waiting for van511121:23
@sonney2kvan511, just hi21:23
van511i'll pass the info to him, don't worry21:23
@sonney2klisitsyn, so I don't really know how to split this stuff out21:23
@lisitsynsonney2k: me neither21:23
@sonney2kvan511, heh21:23
@lisitsynjust generating ideas as usual :D21:24
@sonney2kso lets say we woudl want to extract kernels into a separate lib21:24
@sonney2kwhat would we do?21:24
@lisitsynno no21:24
@lisitsynmakes no sesne21:24
@sonney2kwhy not?21:24
@sonney2kkernels are a submodule of shogun21:25
@lisitsynsonney2k: what we can extract are real algorithms21:25
@lisitsynthat's core21:25
@lisitsynthe mighty trio21:25
@sonney2kwe can only extract these algorithms if they are designed to be extracted21:25
@sonney2kconsider liblinear / libsvm21:25
@sonney2kwe cannot extract them21:25
@sonney2kwe have things in there like ctrl+c -> stop or continue21:26
@sonney2ktrainign time limit21:26
@lisitsynyeah true21:26
@sonney2kown kernels21:26
@sonney2kor even dot features21:26
-!- van51 [] has quit [Ping timeout: 264 seconds]21:26
-!- van511 is now known as van5121:26
@sonney2klisitsyn, e.g. ocas was designed with that in mind21:26
@lisitsynyeah same with tapkee21:26
@sonney2klisitsyn, because vojtech and me were working on it together21:26
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]21:27
@sonney2kand vojtech wanted ocas to be around also stand-a-lone21:27
@sonney2klisitsyn, about extracting kernels21:28
@sonney2kone could move the SG* datatypes into a separate lib that everyone then would use21:28
@sonney2kthen a kernel could just interface with these21:28
@lisitsynsonney2k: I seenothing special in SG* types21:29
@sonney2klisitsyn, we don't want features in kernels21:29
@lisitsynsonney2k: why?21:29
@sonney2kand kernels currently work with features21:29
@sonney2klisitsyn, if we split out kernels21:29
@lisitsynI don't really get that point21:29
@lisitsynwhy to separate them?21:30
@sonney2kkernels should be just kernels not kernels & features21:30
@lisitsynsounds pretty painful21:32
@sonney2kit is all painful21:33
@sonney2kthink about some library like say libsvm using other shogun classifiers as dynamically generated input21:34
@sonney2kthen we cannot extract21:34
@sonney2kthis whole module separation thing only works when you limit functionality21:34
@lisitsynsonney2k: yes we can't extract that21:34
@lisitsynyes sure21:34
@sonney2kmaybe not bad but one has to keep that in mind21:35
@lisitsynsonney2k: what is most important thing I see21:35
@sonney2kif we would really do it I would say each directoy in shogun should be a module21:35
@lisitsynvojtech maintains ocas21:36
@lisitsynI maintain tapkee21:36
@sonney2kyeah but if there is 0 interaction between the two it doesn't help us21:36
@sonney2ke.g. libsvm/liblinear21:36
@lisitsynyes sure21:36
@sonney2kwe take their code21:36
@sonney2kmodify it21:36
@lisitsynwe have to modify their code yes21:36
@sonney2kbut that's it21:37
@sonney2kwe don't update and are not pushed to update21:37
@sonney2kand they don't care about our extensions21:37
@sonney2kit works fine with tapkee because you keep it up-to-date also wrt shogun interface21:37
@lisitsynyes I understand21:38
@sonney2kbut in the end it gives shogun not more - you work for tapkee and not directly on shogun21:38
@sonney2kif you were a non-core shogun author21:38
@sonney2kshogun would benefit21:38
@sonney2kbut yes that would be the way to go21:38
@lisitsynsonney2k: there are a lot of drawbacks in that for sure21:39
@sonney2ksomeone like you taking care of his own lib but who also wants to get exposure with shogun21:39
@sonney2kin addition ML people are like grass hoppers21:40
@sonney2kthey move from topic to topic in no time21:40
@sonney2kand rarely want to maintain any code21:40
@sonney2kI recall the ML-OSS discussions we had in the workshops21:41
@sonney2kpeople said I give out my code sure but I don't want to support it21:41
@sonney2kand it is also different somehow - once an algorithm works it is finished21:42
@sonney2kfinal nothing more needed21:42
@sonney2kthere are only new algorithms improving that algorithm21:42
@lisitsynyou make me feel depressed :D21:42
@lisitsynusually i think there is no sense in all that stuff21:43
@sonney2kvotjakovr, please cheer lisitsyn up hand him a vodka or something21:43
@sonney2kvan51, wiking you too21:43
@lisitsynwe lose him jaja21:43
-!- mdragan [8d55007a@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]21:44
@sonney2klisitsyn, I recall your last years phase when we did the SG* transition and stuff was like boom kaboom for weeks21:44
@lisitsynsonney2k: what do you mean?21:44
@sonney2kit had the same effect on you :D21:44
@sonney2kno I think we have lots of cool algorithms already now but we really just need to expose them better21:45
@sonney2kand sure there always is algoirthm X doing somethign better for whatever use case21:45
@sonney2kor framework Y for some other usecase21:45
@sonney2kwe just have to find (and be satisfied with) our niche21:46
@lisitsynmotivation is very hard to find21:46
@sonney2kand to me it is currently the hardware I have access to: single machine but multicore21:47
@lisitsynI feel bad with doing something that is not going to be used21:47
@sonney2kyou always want to change the world in a day21:47
@sonney2kshogun is being used21:47
@lisitsynyes of course21:47
@lisitsynI am not satisfied with regular day-to-day things21:48
@lisitsynwe need a breakthrough21:48
@sonney2kthe mailinglist even has >300 subscribers21:48
@lisitsynsonney2k: that's soviet-style I want a breakthrough  :D21:51
@lisitsynsonney2k: but actually it is the way software evolves21:52
@lisitsynsoftware dies and borns and old software takes advantage of things developed in new software21:53
votjakovrGood night guys :) What's happened? Did I miss something? Now about 3 a.m. in siberia, so my mind is very slow.21:55
-!- shogun-notifier- [] has joined #shogun21:56
shogun-notifier-shogun: cameron :develop * fc8e6cb / examples/undocumented/python_modular/graphical/
shogun-notifier-shogun: Added new interactive kmm demo21:56
shogun-notifier-shogun: Sergey Lisitsyn :develop * 12f7a85 / examples/undocumented/python_modular/graphical/
shogun-notifier-shogun: Merge pull request #1099 from goldbug/develop-kmm-example21:56
shogun-notifier-shogun: Added new interactive kmm demo (issue #885)21:56
@lisitsynvotjakovr: what's the weather in barnaul?21:56
@lisitsynvotjakovr: that's the longest message I've ever seen from you ;) you usually fix GPs ninja style21:56
votjakovrlisitsyn: It's pretty cool: +4C21:57
@lisitsyndamn that's cold21:58
@sonney2kvan51, btw how are the tests going?21:59
votjakovrlisitsyn: About a week ago we had snow :)21:59
@lisitsynvotjakovr: oh come on21:59
@sonney2kvotjakovr, I was just swimming in a lake here21:59
@sonney2kno heating!21:59
@lisitsynimpossibru to live there :D21:59
@sonney2kkidn of north pople22:00
@lisitsynno its pretty southern22:00
@lisitsynbut very 'continental'22:00
@lisitsynmongolia is worse I guess :D22:01
@sonney2klisitsyn, any clue where gsomix is?22:02
@sonney2khe wanted to send a PR tonight...22:02
@lisitsynsonney2k: no, no idea22:03
@lisitsynI am in the other city22:03
votjakovrlisitsyn: Yeah, there are mountains and nothing else))22:03
@lisitsynsonney2k: you should visit votjakovr's place22:04
@lisitsynsonney2k: it is basically as far from you as say nigeria :D22:07
@lisitsynvotjakovr: I am half of siberian, my mother is from kemerovo22:09
@sonney2kI would need to practice22:09
@sonney2klisitsyn, look
@sonney2kwe have a new repository22:09
@sonney2klisitsyn, yes has some nice gp demo now in there22:11
shogun-buildbotbuild #997 of bsd1 - libshogun is complete: Success [build successful]  Build details are at
votjakovrlisitsyn: Awesome :) I was there once, there's a lot of mines22:12
@sonney2kit totally is kernel build time today :)22:15
@sonney2kmy uptime is just to high22:15
@sonney2kthat is something that needs to be dealt with22:15
shogun-buildbotbuild #989 of deb2 - static_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Sergey Lisitsyn <>22:18
shogun-buildbotbuild #998 of bsd1 - libshogun is complete: Failure [failed test_1]  Build details are at  blamelist: Sergey Lisitsyn <>22:18
votjakovrBtw I'd like to ask: Is there a good way to get log of each  element of eigen3 vector? I mean something like vector.cwiseSqrt().22:21
-!- hushell [] has joined #shogun22:24
-!- dvalcarce [5b757556@gateway/web/freenode/ip.] has joined #shogun22:26
@lisitsynvotjakovr: vector.array().log()22:26
dvalcarcehi all22:26
-!- travis-ci [] has joined #shogun22:26
travis-ci[travis-ci] it's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []22:26
-!- van51 [] has quit [Ping timeout: 246 seconds]22:28
votjakovrlisitsyn: Oh, thank :)22:29
@sonney2kit is good to have a eigen3 expert around22:31
@sonney2kdvalcarce, hey there22:31
dvalcarcesonney2k, hi22:31
-!- van51 [] has joined #shogun22:31
@sonney2kdvalcarce, we have moved everything to shogun-demo now22:31
@sonney2kvan51, bringing good news?22:31
dvalcarcesonney2k, I have just realize22:31
dvalcarcesonney2k, should I move the PR to the new repo?22:32
@sonney2kdvalcarce, yes totally!22:32
@sonney2kthe new contour stuff looks soooooo much better22:32
dvalcarcesonney2k, ok! btw, what do you mean by using more values at the colorbar?22:32
dvalcarcethx :)22:33
dvalcarcesonney2k, I think I have a basic upload system at bakufu, too22:33
van51my router is really struggling with the wifi these days :S22:34
van51it's in the next room and I keep getting dc'd22:34
van51sonney2k: I haven't started them yet22:34
van51I started looking at the issue with some debug messages to see what it was about but I didn't get anywhere22:35
van51so I am going to begin them now22:35
@sonney2kvan51, yeah I had these issues with linux kernel >=3.322:35
@sonney2kso I am now stuck in 3.2.45 or whatever it is now22:35
-!- van511 [] has joined #shogun22:36
@sonney2kvan511, man22:37
van511were any of my messages delivered? :)22:37
@sonney2kyes 622:37
@sonney2kvan511, please do a test - we have them in tests/unit/*22:38
van511ok, I'm on it :)22:38
@sonney2kdvalcarce, cool!22:38
-!- van51 [] has quit [Ping timeout: 256 seconds]22:40
@sonney2kdvalcarce, once you have a bit more I should create shogun-bakufu also in shogun-toolbox/ and move it over22:42
@sonney2klisitsyn, do you know whether one can give access rights to only one of the repositories on github?22:43
@sonney2knot all at once?22:43
@lisitsynsonney2k: I think so22:43
@sonney2kit would make sense for both foulwall and dvalcarce22:43
shogun-buildbotbuild #1128 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Sergey Lisitsyn <>22:44
-!- van511 is now known as van5122:45
shogun-buildbotbuild #1127 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: cameron <>22:48
@sonney2kI have to reboot to the new kernel22:49
@sonney2kbrb (hopefully)22:49
@sonney2klisitsyn, did you see cat's mail?22:56
@lisitsynsonney2k: yes22:56
@sonney2kso it is pretty obvious22:56
@sonney2kno chance for us to get even more slots22:57
@lisitsynyeah looks like22:57
@sonney2kman I am hungry again22:58
@sonney2kis it normal to have dinner twice?22:58
@lisitsynsonney2k: depends22:58
@sonney2kyeah I    manage to gain some weight :D23:00
@sonney2kyou know whales & water -> no problems23:00
@lisitsynsonney2k: if you are not having beer belly yet it must be ok to have a dinner twice :)23:01
-!- votjakovr [] has quit [Quit: Leaving]23:01
@lisitsynI mean you are the first world war veteran you must have a good metabolistic stuff23:02
@lisitsynto be in shape still23:02
dvalcarcesonney2k, I like to focus on bakufu, what should I do next?23:07
dvalcarce(I don't have much time, my finals begin next week, but I'll try to do something)23:07
@sonney2kdvalcarce, anything you can do for bakufu  like file upload some minimal user management23:09
@sonney2kas in each user should have his own space / directory to store stuff in - the standard django-registration struff23:09
@sonney2kdvalcarce, I know you don't have much time but in case you still have  more  you could train e.g. visualize 2 dims of the data or train some simple linear model   on it (if we focus on classification for now) etc etc23:11
@sonney2kbut certainly your finals are more important23:12
@sonney2k..than whole gsoc probably23:12
dvalcarcesonney2k, yeah, finals are important... but gsoc is cooler :P23:13
dvalcarcei will check user management23:13
-!- hushell [] has quit [Quit: leaving]23:14
@sonney2kdvalcarce, once you have sth I should create shogun-bakufu and we shoudl start merge things over23:15
dvalcarcesonney2k, right now ( you can upload and delete documents23:16
van51sonney2k: so, the insert_element(obj, 0) when the array is empty is idd causing a seg fault23:16
van51but what I can't understand is how did the octave example run smoothly then23:16
van51because I used the same logic in CombinedKernel as well23:17
@sonney2kvan51, no idea - write a unit test and fix this insert_element to work in all reasonable cases23:20
@sonney2kdvalcarce, I will move it over then23:29
van51sonney2k: actually, valgind reports it as an invalid memory read23:29
van51but there is no seg fault23:29
@sonney2kvan51, lucky then23:29
@sonney2kdvalcarce, but not  now (too tired)23:29
van51so maybe octave is more tolerant in those situations than python?23:29
@sonney2kvan51, no we were just lucky23:30
@sonney2kwe should always check if valgrind23:30
@sonney2kwell some corruption in not too important memory23:30
@sonney2kbut no idea really23:30
@sonney2kjust relax and fix it23:31
van51i just find it weird!23:31
dvalcarcesonney2k, ok, there is no hurry :)23:31
@sonney2kI gave up trying to understand all bugs23:31
@sonney2kalright that's it for me for today cya!23:39
--- Log closed Sat May 18 00:00:06 2013