Open in new window / Try shogun cloud
--- Log opened Thu Apr 25 00:00:29 2013
@sonney2kwiking, then only disable them for osx00:08
@sonney2kwiking, or anything !linux00:08
@sonney2kthe integration tests we have cover much more than the few tiny tests we currently have so they are invaluable and turning them of completely is no option00:09
-!- makokal [] has quit [Quit: makokal]00:44
-!- iglesiasg [] has quit [Quit: leaving]01:23
-!- shogun-notifier- [] has quit [Quit: transmission timeout]01:26
-!- abinash [75ef5e6e@gateway/web/freenode/ip.] has joined #shogun01:34
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun05:34
deerishiwhois @lisitsyn05:47
-!- abinash [75ef5e6e@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]06:26
-!- iglesiasg [] has joined #shogun08:05
-!- mode/#shogun [+o iglesiasg] by ChanServ08:06
-!- sids_aquarius_ [] has quit [Ping timeout: 264 seconds]08:47
-!- sids_aquarius_ [] has joined #shogun08:49
-!- sids_aquarius [] has joined #shogun09:02
-!- sids_aquarius_ [] has quit [Ping timeout: 246 seconds]09:02
-!- lisitsyn [~blackburn@] has quit [Quit: Leaving.]09:07
-!- makokal [] has joined #shogun09:37
-!- makokal [] has quit [Quit: makokal]09:49
@iglesiasgwiking: around?09:50
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:04
-!- lisitsyn [] has joined #shogun10:17
-!- mode/#shogun [+o lisitsyn] by ChanServ10:17
@lisitsynwiking: :D10:17
@lisitsynwould you like to visit that city I live in? ;)10:18
-!- hoijui [] has joined #shogun10:19
-!- sonne|work [] has joined #shogun10:24
@lisitsynsonne|work: welcome back hehe10:24
sonne|worklisitsyn: hey hey10:25
sonne|workman what a neural network overoverhype!10:25
sonne|workI felt like in the early days of kernel machines!10:25
@lisitsynsonne|work: haha because they work10:25
@lisitsynfor images10:25
sonne|workthey work under very specific conditions10:26
sonne|workwith lots and lots of data10:26
sonne|workand lots and lots of computing power10:26
sonne|workand they are one of the few methods that drastically benefit from GPUs10:26
sonne|workso I guess that is why10:26
@lisitsynsonne|work: yes they work nice on gpus10:27
sonne|workbut as usual no free lunch :D10:27
@lisitsynof course10:27
@lisitsynbut it appears to be better than older stuff10:27
@lisitsynfor some tasks10:27
@lisitsyniglesiasg: may be you want to visit samara in september? :)10:28
sonne|worklisitsyn: no it is the older stuff10:32
sonne|workit is the same stuff from the 80's or so but done with more data & computing power10:33
@lisitsynsonne|work: true10:33
sonne|workand they still have local minima and need clever data normalization10:33
sonne|worksuitable number of hidden layers10:33
sonne|workstructure etc10:33
sonne|workI would say it didn't get easier but hey I guess one can say that about any method10:34
@lisitsynsonne|work: so no more svms? :)10:35
sonne|workI saw cascades of decision / regression trees10:35
sonne|workand the usual stuff like logistic regression etc10:36
sonne|workand NN's10:36
@iglesiasglisitsyn: oh nice a conference in your city :)10:38
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun10:38
@lisitsyniglesiasg: yeah - there is some time to come up with something10:40
@iglesiasglisitsyn, sonne|work : btw could you guys please this PR?10:40
@iglesiasgit doesn't break anything new according to travis10:40
-!- shogun-notifier- [] has joined #shogun10:41
shogun-notifier-shogun: Fernando Iglesias :develop * 5a265c2 / src/shogun/machine/ (4 files):
shogun-notifier-shogun: Setters/getters in structured machines.10:41
shogun-notifier-shogun: They are required to implement SO-SVMs from interfaces.10:41
shogun-notifier-shogun: Sergey Lisitsyn :develop * 7b4c255 / src/shogun/machine/ (4 files):
shogun-notifier-shogun: Merge pull request #1012 from iglesias/feature/improve_structure10:41
shogun-notifier-shogun: Setters/getters in structured machines.10:41
@iglesiasgmy python bindings for structured learning uses it and I think Patrick wanted to test something10:41
@iglesiasglisitsyn: thanks!10:41
vgorbatilisitsyn: hi there10:48
@lisitsynvgorbati: hey10:48
-!- van51 [] has joined #shogun10:50
vgorbatilisitsyn: it looks like you are in good mood for merging PR's - do you have time to review my PR too?:)
@lisitsynvgorbati: could you please rebase it?10:53
@lisitsynyou have 4 merge commits here10:53
vgorbatilisitsyn: basically I need to call 'git rebase develop'?10:55
@lisitsynno it won't help there10:56
vgorbatilisitsyn: and what will help? Previously I just called 'git pull upstream develop' from time-to-time to stay in touch with develop branch.10:59
-!- travis-ci [] has joined #shogun10:59
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 []10:59
-!- makokal [] has joined #shogun11:01
@iglesiasglisitsyn: have you seen the cool perceptron we have now?
@lisitsyniglesiasg: is he drunk?11:10
@lisitsynrosenblatt gets drunk11:11
@iglesiasglisitsyn: haha not, just randomly initialized11:11
@lisitsynI see11:11
@iglesiasgI used that figure for my report11:11
@lisitsynvgorbati: always do 'git pull --rebase upststream develop' :)11:11
@lisitsynokay lets not mess with that now11:11
vgorbatilisitsyn: so, just to make it clear for me, in future, when I want to get changes from upstream/develop I should run 'git pull --rebase upstream develop' instead of 'git pull upstream develop11:12
vgorbati', right?11:13
@lisitsynvgorbati: yes11:16
@lisitsynit won't produce these merge commits11:16
@lisitsynjust rebases your commits on top of upstream branch11:16
vgorbatilisitsyn: as a result there wont be any 'merge ...' commits when I do PR?11:17
vgorbatilisitsyn: sorry, I should read before writing:)11:17
@lisitsynantihistaminic stuff makes me sleeeep11:18
-!- HeikoS [] has joined #shogun11:20
-!- mode/#shogun [+o HeikoS] by ChanServ11:20
van51from what i read between you two, i am beginning to understand that I should have done something differently11:22
van51because my PR is on the master branch11:23
van51am I right?11:25
vgorbatilisitsyn: :) caught a cold?11:25
vgorbativan51: as far as I understand, all the PR's should be send to develop branch11:25
@lisitsynvgorbati: allergy11:26
sonne|workvan, vgorbati: exactly just do it against develop11:27
vgorbatilisitsyn: I guess the allergy is not to cats, because then your git-photo is strange:)11:28
@lisitsynvgorbati: spring stuff11:29
@lisitsynvgorbati: that's because I am a cat11:29
vgorbatilisitsyn: :)11:29
vgorbatilisitsyn: during spring you go crazy?)11:30
vgorbatilisitsyn: like a cat?11:30
@lisitsynvgorbati: hah probably no11:31
van51ok so I re-did a PR, can I delete the previous one? :S11:34
@lisitsynvan51: just clost it11:35
van51lisitsyn: ok, done :)11:35
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Quit: Page closed]11:36
-!- lisitsyn [] has quit [Ping timeout: 240 seconds]12:00
-!- lisitsyn [] has joined #shogun12:05
-!- iglesiasg [] has quit [Quit: leaving]12:13
-!- van51 [] has quit [Ping timeout: 268 seconds]12:20
-!- HeikoS [] has quit [Quit: Leaving.]12:43
@wikingsonne|work: ping?13:13
@wikingsonney2k: pong :)13:14
-!- hoijui_ [] has joined #shogun13:22
sonne|workwiking: pong13:24
-!- sijin_ [~smuxi@] has joined #shogun13:26
@wikingsonne|work: sent u an email and to the other core developers13:29
-!- Netsplit *.net <-> *.split quits: hoijui, sijin13:29
sonne|workok so I wait 30mins13:29
@wikingi dont think so13:29
@wikingit should be already there13:29
@wikingsent it to Soeren Sonnenburg <>13:30
@wikingsonne|work: it's all about the intergration tests...13:30
@wikingand that we should basically disable it temporarily13:30
@wikingor at least return always 013:30
@wikingso that it does not fail even if some of it fails :P13:30
sonne|workwiking: greylisting13:31
@wikingsonne|work: oh motherfucker13:31
@wikingsonne|work: evne from gmail? :)13:31
@wikingsonne|work: *eve13:31
@wikingsonne|work: u shoudl have it in /msg :P13:34
@wikingunless your irc client put me on flood :P13:34
sonne|workwiking: yeah but I already said - why disable it?13:34
sonne|workit is very valuable for linux13:34
sonne|workif you want to disable it then just for !linux systems13:34
sonne|workAdmittedly I don't feel very comfortable of having our own rng in shogun13:35
sonne|workit is a cruzial thing - does eigen ship one btw?13:35
@wikingok but they are destined to fail of !linux13:36
@wiking*on !linux13:36
@wikingeither we same the models for other OS as well in data repo13:36
@wikingor what u suggest is that we don't run them on non-linux systems13:37
@wikingbut then in my opinion is a badly designed test :(13:37
@wikingit's kind of like kill it's purpose :P13:37
@wikingif the test can only be run on a given system....13:37
-!- shogun-notifier- [] has quit [Quit: transmission timeout]13:41
sonne|workwiking: I didn't say we cannot fix it13:41
sonne|workbut I am against disabling it13:41
sonne|workthese integration tests are much more helpful then our few unit tests13:41
sonne|workand it takes long ages to get the whole suite back to work if we ignore them for 6 months (==this gsoc)13:42
@wikingok but imo we should do something about it, becuase now one really has to check the travis builds in detail to see whether a PR or a commit does indeed break the shogun build, instead of really just checking a small icon which is either red or green...;P13:45
sonne|workwiking: wait why?13:47
sonne|workeverything should work now right?!13:47
@wikingsonne|work: no now it fails13:47
@wikingsonne|work: everything13:47
sonne|workwiking: why?13:47
@wikingsonne|work: check out this for example13:48
@wikingaccording to travis this PR is not good... it fails the build13:48
@wikingbut it's not because of that given PR13:48
sonne| setting 1/213:49
@wikingit's because the python_modular integration tests fails13:49
sonne|workonly that fails13:49
sonne|worknothing more13:49
@wikingsonne|work: indeed13:49
@wikingsonne|work: but still you don't know this13:49
sonne|workso we just have to fix that13:49
@wikingunless you check yourself13:49
@wikingthe travis builds13:49
@wikingwell on github you only know that the travis build failed13:49
sonne|workwhen something breaks you have to check the stuff and fix it13:49
sonne|workyes sure13:49
@wikingyes but now it's always broken13:49
sonne|workbut how else would you do it?13:49
@wikingno matter what u do13:49
sonne|workwiking: well then just fix the example / the bug13:50
@wikingsonne|work: ok... let me put it another way13:50
@wikingi myself have no possible means to find out why that one little integration test fails on travis13:50
@wikingand this is like this for the last 1-2 weeks13:51
sonne|workwiking: yeah me neither13:51
@wikinghence nobody has any time/motivation etc13:51
sonne|workbecause it works here13:51
@wikingto fix this13:51
sonne|workso I just don't know how to fix it13:51
@wikingand i suggested a way to overcome this problem13:51
sonne|workwe could just blacklist this particular test13:51
sonne|work(like we do with lots of the others already)13:52
@wikingsonne|work: either remove the integration test, or just don't return 1 if one of the the integration tests fails13:52
sonne|workwe have to fail13:52
sonne|workotherwise no one will check test13:52
sonne|workthen better blacklist it13:52
@wikingok then let's blacklist that one serialization test13:52
sonne|workwiking: just open up generator.py13:53
sonne|workand add the test to the blacklist13:53
@wikingsonne|work: btw it's really funny enough that *that* given test does not fail on osx...13:53
sonne|workit will kill both cases though13:53
@wikingsonne|work: but a lot of others does....13:53
sonne|workI don't understand how it can fail13:53
sonne|workI mean it works for one and not for the other13:54
sonne|workmaybe we should not blacklist but print the differences13:54
@wikingwell that's an option as well13:54
@wikingor you have a more sophisticated one?13:58
sonne|workwiking: maybe better add a failed_string14:01
sonne|workthen add a tuple to failed14:01
sonne|workfailed.append((setting_str, failed_str))14:01
sonne|workso that we see it as summary at the end14:01
sonne|workotherwise it gets lots14:01
sonne|workbut otherwise yes - that is whay I would do14:02
@wikingand failed_str is the 2 tuples?14:02
sonne|workand also mention if the length doesn't fit etc14:02
sonne|workwiking: there is a failed=[] array at the top14:02
sonne|workfailed_str should be some string in the end14:02
@wikingand i see that setting_str is appeneded14:02
sonne|workit probably should also list if len(a) != len(b) etc14:03
sonne|workbut hey it is a start14:03
-!- pranjaldaga1 [~pranjalda@] has joined #shogun14:07
@wikingok from this one why does it fail14:07
@wikingi dont see it14:07
@wiking(' setting 1/2', [(CombinedKernel, [True, True, True, True, True], [True, True, True, True, True]), (CombinedKernel, [True, True, True, True, True], [True, True, True, True, True])])14:07
sonne|workwiking: argh14:08
-!- iglesiasg [] has joined #shogun14:08
-!- mode/#shogun [+o iglesiasg] by ChanServ14:08
@wikingbut it fails on my machine :S14:08
sonne|workso it is the guts of combined kernel14:09
sonne|workwiking: for SGObjects we have some function to print parameters IIRC14:09
sonne|workbut I don't know whether we can print the actual current values14:10
@wikingok anyhow14:10
@wikingi did this14:10
@wiking-                                                       failed.append(setting_str)14:10
@wiking+                                                       failed_string = []14:10
@wikingwant this in or not?14:10
@wiking+                                                       failed_string.append(a)14:10
@wiking+                                                       failed_string.append(b)14:10
@wikingjust to see what travis says14:10
@wiking+                                                       failed.append((setting_str, failed_string)) print "%-60s ERROR" % setting_str14:10
sonne|workgood enough14:12
sonne|workjust push it14:12
sonne|workwiking: let me check if sgobject can print values14:12
@wikinglet's see what travis does with it14:13
-!- shogun-notifier- [] has joined #shogun14:13
shogun-notifier-shogun: Viktor Gal :develop * f05af6d / tests/integration/python_modular/
shogun-notifier-shogun: More debug information for python_modular integration tests14:13
sonne|workwiking: the only alternative idea I have is to serialize this sgobject into an asciifile14:13
sonne|workand then show the actual diff -u14:13
@wikinganything is good just commit or tell me what to do14:15
@wikingi just want to fix this shit with travis14:16
@wikingas it's really a lot of help for all of us14:16
@wikingsonne|work: btw: have u figured out what the heck is with buildbot?14:16
-!- sijin_ [~smuxi@] has left #shogun []14:17
-!- sijin_ [~smuxi@] has joined #shogun14:18
-!- sijin_ [~smuxi@] has quit []14:18
-!- pranjaldaga1 [~pranjalda@] has quit [Ping timeout: 260 seconds]14:19
-!- Netsplit over, joins: sijin14:20
sonne|workwiking: no - I didn't get it to work14:20
sonne|workwiking: in principle version 0.8.7 supports multiple branches14:21
sonne|workbut gitpoller always says - no changes14:21
sonne|workI simply don't understand why14:21
sijinSorry to interupt, I just found Ckernel got segmentation fault when using it twice.14:23
sonne|worksijin: what are you exactly doing?14:25
sijinI just use it like this: Ckernel*p = new CGaussianKernel(); p->init(feature1,feature2);p->cleanup(); p->init(feature1, feature2); p->cleanup()14:26
sijinthen i got segmentation fault14:27
shogun-notifier-shogun: iglesias :develop * 34056f3 / src/shogun/classifier/Perceptron.cpp,src/shogun/classifier/Perceptron.h:
shogun-notifier-shogun: Add boolean member to Perceptron to allow model initialization.14:28
shogun-notifier-shogun: This is useful to show that the perceptron converges to different14:28
shogun-notifier-shogun: solutions depending on initial values of w and bias.14:28
shogun-notifier-shogun: iglesias :develop * 6fd3a5a / examples/undocumented/python_modular/graphical/
shogun-notifier-shogun: Perceptron graphical example.14:28
shogun-notifier-shogun: For the same data, the solution is different depending on initial values14:28
shogun-notifier-shogun: of w and bias14:28
shogun-notifier-shogun: Soeren Sonnenburg :develop * e10f78b / / (3 files):
shogun-notifier-shogun: Merge pull request #1007 from iglesias/feature/custom_init_perceptron14:28
shogun-notifier-shogun: Add small functionality to perceptron and graphical example14:28
sijinfeature1, and feature1 are pointers of generated data.14:29
sonne|worksijin: please post the exact code (some small example) for us to reproduce14:30
sonne|workit can be anything currently (also a bug on your side)14:30
-!- foulwall [] has joined #shogun14:30
sonne|workfoulwall: hey there14:31
foulwallHello everybody, sorry for the absence  these days.14:31
foulwallsonne|work: I prepared for the mid term exams.14:32
sonne|workI guessed so14:32
sonne|workfoulwall: just don't forget to submit you application in time14:33
sonne|workfoulwall: btw I showed you demos to some professors14:33
sijin  here is the code14:33
sonne|workthey liked it14:33
sonne|workfoulwall: and they asked whether they can use this in teaching14:34
sonne|workfoulwall: so if we do it really well lots of  students will see your demos :D14:34
-!- travis-ci [] has joined #shogun14:35
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:35
foulwallsonne|work: I'll fix the GP regression demo tonight(add the region as you said)14:36
foulwallsonne|work: of course they can.14:36
sonne|workfoulwall: yeah but it would be great wouldn't it? I mean people really learning / seeing what you've done14:37
foulwallsonne|work: I'll try to make the demo show the immediate step14:37
sonne|workfoulwall: that also needs some changes inside shogun14:37
foulwallsonne|work: that'll be great14:37
sonne|worklets first do the demos14:38
sonne|workthen add more14:38
foulwallsonne|work: and for the 3 choice, I choice the easiest one.14:38
foulwallthe first one. I'll make the demo more beautiful14:38
foulwallsonne|work: for the 3 choices, I choose the first one.14:40
foulwallsonne|work: I'll do some work later today. I'll go for a class.14:41
foulwallsonne|work: cu.14:41
-!- foulwall [] has left #shogun []14:41
sijinsonne|work: did you see the code? any problems?14:42
sonne|worknot clear yet14:46
-!- travis-ci [] has joined #shogun14:51
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:51
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun14:54
sijinsonne|work: can you try that code on your computer?14:58
sonne|worksijin: no time yet15:02
sijinok, thx all the same15:02
-!- abinash [75ef5e6e@gateway/web/freenode/ip.] has joined #shogun15:13
-!- gsomix [~Miranda@] has joined #shogun15:18
gsomixsonne|work: are you back?15:20
sonne|workgsomix: yes berlin15:21
-!- van51 [] has joined #shogun15:21
gsomixsonne|work: great!15:22
-!- iglesiasg [] has quit [Quit: Lost terminal]15:37
@wikingsonne|work: ok so the error is atm meaningless: ' setting 1/2', [(GMNPSVM, GMNPSVM), (GMNPSVM, GMNPSVM)])15:40
@wikingsonne|work: any ideas what we should do next?15:40
sonne|workwiking: yeah we need a method to print the parameters and *values* of shogun objects16:02
sonne|workI think we already have some print_parameters or sth16:02
sonne|workso it should be very doable to also print the actual value in there16:02
sonne|workwiking: on the other hand this is more or less the same as serializing a shogun object into an ASCII file16:03
sonne|workand maybe the latter is better16:03
sonne|workand then just do a diff on top of it16:03
sonne|workwiking: that woudl work like this:fstream = SerializableAsciiFile("foo.asc", "w")16:04
sonne|workand then just outputting diff -u16:05
@wikinglisitsyn: ping?16:16
@wikingsonne|work: would we consider a library like this for integration:
@wikingor it's too strict?16:17
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:17
@wikingok no worries i've found this one which is totally gpl 2:
-!- abinash [75ef5e6e@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:21
lisitsynwiking: yes?16:23
@wikinglisitsyn: can we just put this into shogun/lib: //
@wikingand add a wrapper around it16:23
@wikingfor shogun?16:23
lisitsynlooks like it is not free16:24
@wikingeven weka is using this16:24
lisitsynI mean it is just non-commercial16:24
lisitsynah alright16:25
lisitsynif it s gpl then ye16:25
@wikinglisitsyn: ah u r checking the other16:25
@wikinglisitsyn: ok question lib/external/librf16:25
@wikingor lib/librf?16:25
lisitsynwiking: external probably16:25
@wikingok let's add a new branch finally to git16:29
@wikinghuh this import will be rather funny16:35
@wiking./lib/external/librf/tree.cpp:15:25: error: librf/utils.h: No such file or directory16:35
-!- hoijui_ [] has quit [Quit: Leaving]16:38
van51can someone help me on how to make two commits into one? :)16:52
sonne|workvan51: git reset <oldgitid>17:04
sonne|workgit commit --amend17:04
-!- makokal [] has quit [Quit: makokal]17:05
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]17:09
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun17:11
van51sonne|work: thanks, but what if I don't have the code locally at the moment?17:11
@wikinghow do i publish a git flow feature on shogun-toolbox/shogun.git? :S17:20
-!- shogun-notifier- [] has quit [Quit: transmission timeout]17:28
-!- gsomix [~Miranda@] has joined #shogun17:42
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]17:46
-!- lisitsyn [] has quit [Quit: Leaving.]18:01
@wikingi have something, but not the right place18:07
-!- sijin [~smuxi@] has quit []18:09
-!- theo [802818b4@gateway/web/freenode/ip.] has joined #shogun18:10
-!- theo [802818b4@gateway/web/freenode/ip.] has quit [Client Quit]18:14
@wikingsonne|work: ping18:22
-!- iglesiasg [] has joined #shogun18:26
-!- mode/#shogun [+o iglesiasg] by ChanServ18:26
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun18:31
@wikingwtf is this18:38
@wikingmodshogun_wrap.cxx:5476:3: error: "SWIG wrapped code invalid in 32 bit architecture, regenarete code using -DSWIGWORDSIZE32"18:38
-!- pranjaldaga [~pranjalda@] has joined #shogun18:40
@iglesiasgwiking: the reason why travis fails in python_modular?18:40
-!- HeikoS [] has joined #shogun18:42
-!- mode/#shogun [+o HeikoS] by ChanServ18:42
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]18:47
@wikingiglesiasg: no it's the reason why python modular fails on freebsd18:51
-!- lisitsyn [~blackburn@] has joined #shogun18:54
-!- mode/#shogun [+o lisitsyn] by ChanServ18:54
-!- iglesiasg [] has quit [Quit: Reconnecting]18:58
-!- iglesiasg [] has joined #shogun18:58
-!- mode/#shogun [+o iglesiasg] by ChanServ18:59
gsomixsonne|work: sonney2k: lisitsyn: iglesiasg: HeikoS: wiking: good evening. can you take a look at it? :)18:59
gsomixI'm opened for any suggestions18:59
@iglesiasgit doesn't seem to load here :S19:01
@HeikoSsame here :)19:02
gsomixoh gosh :(19:02
@HeikoSgsomix: like all the ideas!19:04
@iglesiasggsomix: yeah, nice ones19:04
@HeikoSespecially R and matlab bindings :) but they might be challenging19:05
@iglesiasggsomix: my favourite are the 4th and the 6th19:05
gsomixiglesiasg: thanks19:07
gsomixHeikoS: yep, most challenging task is matlab interface, because swig doesn't have officially support (only for octave)19:08
@iglesiasggsomix: and one more suggesting that I would love seeing is cmake support19:08
@iglesiasgsuggesting -> suggestion19:08
@HeikoSgsomix: did you have a look at that project by the guy who wanted to improve swig for matlab?19:08
@HeikoSgsomix: but this might be quite hard in fact19:08
@iglesiasgHeikoS: gsomix: IMHO matlab interface doesn't pay off the effort19:09
@iglesiasgHeikoS: I know you are a matlab supporter though :)19:09
@HeikoSiglesiasg: I think I recently changed the sides :)19:11
@HeikoSbut still, we need to have it19:11
@HeikoSso many people use it19:11
gsomixHeikoS: a little, just looked at code. not have time for test right now. I plan to test it after proposal submiting.19:11
gsomixHeikoS: so, we can lay aside this idea for near future.19:17
@HeikoSI would be very keen in having automatic interfaces for matlab19:17
@HeikoSalmost everyone I know uses it19:17
@HeikoSI tell them to use python, but it doesnt help :)19:18
gsomixiglesiasg: cmake support. added to list19:18
@HeikoSsame for R btw19:18
@HeikoSboth bio and stats people mostly use R19:18
@iglesiasggsomix: it would be pretty awesome, but it can be tought. I am not sure19:18
@HeikoS(and matlab)19:18
gsomixHeikoS: yep, I use matlab for my research, and my supervisor, and all in my lab. it is very popular.19:19
* gsomix afk19:21
-!- Vikalp [~AndChat49@] has joined #shogun19:49
VikalpHi everybody20:06
VikalpM getting linking error while using shogun sgmatrix20:07
VikalpCan anybody please help me out?20:07
@lisitsynVikalp: hey20:15
@lisitsynare you using -lshogun key?20:15
@lisitsynwhat is the error?20:16
@wikingwhy is this:20:16
@wiking-LINKFLAGS_PYTHON="-ldl -lutil"20:16
@wiking-POSTLINKFLAGS_PYTHON="-ldl -lutil"20:16
@wikingin ./configure20:16
@lisitsynwiking: they should not be here?20:17
@wikinglisitsyn: well it's fine as long as it's only intended for linux ;)20:18
@lisitsynwiking: ahh20:18
@wikingthere's no libdl on *bsd20:18
@lisitsynwiking: I should buy a mac20:18
@wikinglisitsyn: any idea how i can move the feature/RandomForest branch to shogun-toolbox/shogun repo?20:19
@lisitsynwiking: I actually had a chance but bought that shitty samsung20:19
@wikingthe gitflow feature publish put it to my own repo20:19
@lisitsynwiking: but why do you want to have it in main repo20:19
@wikinglisitsyn: so that you can contribute as well?20:20
@lisitsynwiking: yeah makes sense20:20
@lisitsynwiking: okay gitflow just names branches20:20
@wikingi mean in any feature if you publish it20:20
@lisitsynso git pull it20:20
@wikingit would be great to go to the upstream repo20:20
@wikingsince u want to publish it20:21
@lisitsynwiking: just pull and push it20:21
@wikingin order that others can contribute as well20:21
@wikingyou mean git push upstream feature/RandomForest20:22
@wikingthere :)20:22
-!- shogun-notifier- [] has joined #shogun20:22
shogun-notifier-shogun: Viktor Gal :feature/RandomForest * 8b17964 / src/shogun/lib/external/librf/ (17 files):
shogun-notifier-shogun: Initial import of librf20:22
shogun-notifier-shogun: Random Forest implementation20:22
shogun-notifier-shogun: Viktor Gal :feature/RandomForest * e9928f8 / src/shogun/lib/external/librf/ (12 files):
shogun-notifier-shogun: Fix header include for librf20:22
shogun-notifier-shogun: Viktor Gal :feature/RandomForest * 68b8c99 / src/shogun/lib/external/librf/test.cpp,src/shogun/lib/external/librf/weights.cpp:
shogun-notifier-shogun: librf: remove unnecessary files20:22
@lisitsynwiking: yeah if you can20:22
@wikingok this one compiles fine20:23
@wikingwe need wrappers and unit tests20:23
@wikingthe original code does have unit tests20:23
@wikingso i'll integrate them20:23
@wikingonly thing that it's not written using gtest :P20:24
@wikingbut of course i don't know if it's a good idea to add unit tests that are testing none-shogun code... but i guess it wouldn't hurt20:24
@lisitsynwiking: we should have some strict policy20:25
@lisitsynhow to handle external code20:25
@wikinganyways this needs now a shogun wrapper20:29
@wikingbtw: is this a linearmachine? :D20:30
@lisitsynwiking: no!20:30
-!- Vikalp [~AndChat49@] has quit [Quit: Bye]20:31
-!- saket [0e8b5206@gateway/web/freenode/ip.] has joined #shogun20:31
@lisitsynwiking: just a machine20:31
@lisitsynwiking: what is linear in random forest? ;)20:31
@wikingi'm just asking what should be the head20:32
@wikingbtw <shogun/classifier/RandomForest.h> ?20:32
@lisitsynwiking: good question20:32
@lisitsynwiking: it handles multiclass right?20:33
@wikinglisitsyn: afaik it should20:33
@lisitsynthen makes sense to put it to multiclass20:33
@wikingah ok20:33
@wikingknn is there as well20:33
-!- gsomix [~Miranda@] has quit [Read error: Connection reset by peer]20:46
-!- travis-ci [] has joined #shogun20:54
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []20:54
@wikingwoah cool20:55
@wikingactually travis is useful \o/20:55
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun21:01
-!- gsomix [~Miranda@] has joined #shogun21:04
-!- pranjaldaga [~pranjalda@] has quit [Ping timeout: 245 seconds]21:08
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun21:12
shogun-notifier-shogun: :develop * 5e03c62 / .travis.yml,src/Makefile,src/shogun/labels/DenseLabels.cpp:
shogun-notifier-shogun: Merge branch 'develop' of into vgorbati-develop21:14
shogun-notifier-shogun: :develop * f2874ed / :
shogun-notifier-shogun: Merge branch 'develop' of into vgorbati-develop21:14
shogun-notifier-shogun: :develop * 2d84d67 / / (60 files):
shogun-notifier-shogun: Merge branch 'develop' of into vgorbati-develop21:14
shogun-notifier-shogun: :develop * b15aec8 / src/NEWS,src/shogun/lib/tapkee/tapkee_methods.hpp:
shogun-notifier-shogun: Merge branch 'develop' of into vgorbati-develop21:14
shogun-notifier-shogun: :develop * 9d40286 / / (5 files):
shogun-notifier-shogun: Added a wrapper (CTDistributedStochasticNeighborEmbedding class) for t-SNE dimensionality reduction method, implemented in tapkee (according to
shogun-notifier-shogun: Also added a basic unit test for it, that embeds the test data from 3D to 2D using implemented wrapper, and checks that resulting data is actually in 2D, and that number of features stay the same.21:14
shogun-notifier-shogun: Sergey Lisitsyn :develop * c88afbf / / (5 files):
shogun-notifier-shogun: Merge pull request #1015 from vladislav-horbatiuk/vgorbati-develop21:14
shogun-notifier-shogun: Added a wrapper for t-SNE method implemented in tapkee library and a basic unit test for it.21:14
@lisitsynwiking: so we have t-sne in shogun now too21:14
@HeikoShey guys, currently running a distributed program on 160 computers :D21:15
@lisitsynHeikoS: superman!21:15
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]21:17
@HeikoSlisitsyn: this is very cool stuff21:27
@HeikoSoh and btw21:28
@HeikoSa stats postdoc here will implement a simple mala sampler in shogun21:28
@HeikoSand with that hopefully create the foundation for some sampling action21:28
@lisitsynHeikoS: what is it?21:28
@HeikoSMetropolis adjusted Langevin algorithm21:29
@lisitsynno idea!21:29
@HeikoSa method to sample probability distributions that are complicated21:29
@HeikoSyou can use that to solve integrals that are too hard otherwise21:30
@HeikoSthink of it as brute force breaking things when you cannot do anything elegant ;)21:32
@HeikoSbut in an elegant way :D21:32
@lisitsynHeikoS: sounds like monte-carlo21:32
@lisitsynHeikoS: I am still in 60s :D21:33
@HeikoShah :)21:33
-!- gsomix [~Miranda@] has quit [Read error: Connection reset by peer]21:33
-!- makokal [] has joined #shogun21:33
@HeikoSbut very nice that the guy here does it21:33
@HeikoSthey use it for a project and I convinced them to do it in shogun21:33
@lisitsynHeikoS: yeah sure21:34
@HeikoSthis is nice since if we have many of those, we might attract a few more stats people21:34
@lisitsynHeikoS: we should think how can we speed up development21:34
@HeikoSnext year's gsoc will have a sampling project for sure! :)21:35
@lisitsynHeikoS: for example we could add a script21:35
@lisitsynthat creates a class21:35
@lisitsynputs it to modular interface21:35
@lisitsynHeikoS: see what I mean?21:35
@lisitsynHeikoS: I mean for your guy it is not obvious how to put a class to modular interface21:42
@lisitsynHeikoS: and it would be nice to have a tool that does that21:42
@lisitsynsay I write 'new_class statistics DaCoolSampling'21:43
@lisitsynand here we go21:43
-!- saket [0e8b5206@gateway/web/freenode/ip.] has quit [Quit: Page closed]21:45
-!- gsomix [~Miranda@] has joined #shogun21:46
@HeikoSlisitsyn: sorry gotta go.... bye :)21:46
@lisitsynHeikoS: alrighto21:46
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun21:56
@sonney2kwiking, any progress on printing out stuff?22:24
@wikingsonney2k: not really22:24
@wikingsonney2k: i bascially gave up22:25
@wikingsomebody should take over22:25
@wikingi'm doing now random forest rather22:25
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]22:30
gsomixsonney2k: hey :)22:37
-!- sids_aquarius [] has quit [Ping timeout: 246 seconds]22:52
-!- sids_aquarius [] has joined #shogun22:53
-!- travis-ci [] has joined #shogun23:09
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 []23:09
@lisitsynmy turn to suffer23:09
-!- naywhaya1e [] has joined #shogun23:10
-!- Netsplit *.net <-> *.split quits: naywhayare23:15
--- Log closed Fri Apr 26 00:00:29 2013