Open in new window / Try shogun cloud
--- Log opened Sun Mar 10 00:00:10 2013
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]00:19
-!- FSCV_ [~FSCV@65.19.131.247] has joined #shogun00:28
-!- heiko1 [~heiko@b019b9ef.bb.sky.com] has joined #shogun00:29
-!- heiko [~heiko@b019b9ef.bb.sky.com] has quit [Read error: Connection reset by peer]00:29
-!- FSCV [~FSCV@76.73.3.220] has quit [Ping timeout: 248 seconds]00:30
-!- SCV__ [~FSCV@216-230-229-167-colo.oplink.net] has joined #shogun00:33
-!- heiko1 [~heiko@b019b9ef.bb.sky.com] has quit [Read error: Connection reset by peer]00:33
-!- heiko [~heiko@b019b9ef.bb.sky.com] has joined #shogun00:33
-!- FSCV_ [~FSCV@65.19.131.247] has quit [Ping timeout: 245 seconds]00:34
-!- heiko [~heiko@b019b9ef.bb.sky.com] has quit [Client Quit]00:35
-!- FSCV_ [~FSCV@173.254.212.46] has joined #shogun00:38
-!- SCV__ [~FSCV@216-230-229-167-colo.oplink.net] has quit [Ping timeout: 248 seconds]00:39
-!- SCV__ [~FSCV@108.60.131.13] has joined #shogun00:41
-!- FSCV_ [~FSCV@173.254.212.46] has quit [Ping timeout: 260 seconds]00:44
-!- SCV__ [~FSCV@108.60.131.13] has quit [Ping timeout: 264 seconds]00:55
-!- SCV__ [~FSCV@189.139.249.183] has joined #shogun01:08
-!- n4ndo [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]01:36
-!- SCV__ [~FSCV@189.139.249.183] has quit [Ping timeout: 260 seconds]02:50
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun02:54
n4nd0sonney2k: it looks from the mails that the webpage is crashing badly02:54
n4nd0I will take a look ASAP (by Monday I think)02:54
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Client Quit]02:55
shogun-buildbotbuild #317 of nightly_default is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/31703:55
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun05:37
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun07:06
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]07:26
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has joined #shogun08:33
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun10:23
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 250 seconds]10:46
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun10:46
-!- heiko [~heiko@b019b9ef.bb.sky.com] has joined #shogun12:24
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has quit [Read error: Connection reset by peer]12:33
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 260 seconds]12:54
-!- heiko [~heiko@b019b9ef.bb.sky.com] has quit [Quit: Leaving.]14:15
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun15:51
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: Lost terminal]16:07
blackburnwiking: ping16:15
-!- ishan [~ishan@14.139.122.114] has joined #shogun17:07
-!- heiko [~heiko@b019b9ef.bb.sky.com] has joined #shogun17:19
-!- ishan_ [~ishan@14.139.122.114] has joined #shogun17:24
-!- ishan [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]17:24
-!- ishan_ [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]17:54
-!- ishan_ [~ishan@14.139.122.114] has joined #shogun17:55
-!- heiko [~heiko@b019b9ef.bb.sky.com] has quit [Quit: Leaving.]17:59
-!- ishan__ [~ishan@14.139.122.114] has joined #shogun18:24
-!- ishan_ [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]18:24
-!- ishan__ [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]18:54
-!- ishan__ [~ishan@14.139.122.114] has joined #shogun18:55
-!- heiko1 [~heiko@b019b9ef.bb.sky.com] has joined #shogun19:18
-!- ishan__ [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]19:24
-!- ishan__ [~ishan@14.139.122.114] has joined #shogun19:24
heiko1 blackburn, around=?19:29
blackburnheiko1: yes19:29
heiko1blackburn: what about your unstable methods?19:30
heiko1multiclass logistic regression?19:30
blackburnheiko1: MC LR?19:30
heiko1do we need those?19:30
heiko1should I leave them?19:30
heiko1why did you add them?19:30
blackburnfeel free to remove19:30
heiko1what about a test instead? :D19:30
blackburnheiko1: I was unsure it is working properly19:30
blackburnheiko1: I am in limited time to write a test.. it had low score on my dataset IIRC19:30
heiko1blackburn:  well did you test it?19:30
heiko1or just hacked it and then left the room? :)19:31
blackburnheiko1: no I used it for something19:31
heiko1well then,  I will remove. Feel free to attach a test :) I am a fan of tests nowadays :)19:31
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun19:41
shogun-notifier-shogun: Sergey Lisitsyn :master * cfb5bf7 / src/shogun/lib/tapkee/ (35 files): https://github.com/shogun-toolbox/shogun/commit/cfb5bf76fe67654d01a80895b71226a50b729f0b19:41
shogun-notifier-shogun: Big update for tapkee19:41
shogun-notifier-shogun:19:41
shogun-notifier-shogun: - Relicensed code to BSD19:41
shogun-notifier-shogun: - Better includes arrangement19:41
shogun-notifier-shogun: - Doc improvements19:41
blackburnheiko1: I went away from gpl :)19:41
heiko1blackburn: to?19:41
heiko1bsd?19:41
blackburnheiko1: bsd19:41
heiko1wow19:41
heiko1reasons?19:41
blackburnheiko1: simpler, want to allow commercial use19:42
heiko1thats possible with gpl19:42
blackburnheiko1: lgpl then19:42
heiko1gpl also allows that19:42
heiko1as far as I know?19:42
blackburnheiko1: you'd need to publish your sources with gpl19:42
heiko1well yes19:43
blackburnwith lgpl it is better19:43
heiko1but that doesnt block commercial use19:43
heiko1and it ensures that tapkee gets better19:43
blackburnheiko1: no it doesn't19:43
blackburnheiko1: well gpl tries to do that but from my point it doesn't :)19:44
heiko1well, its your project :)19:44
heiko1many people use bsd in fact19:44
heiko1the google libs are also bsd19:44
blackburnI mean yes people somewhat should publish changes that they have done19:44
blackburn(with gpl)19:44
heiko1yes19:45
blackburnbut I'd say gpl just scary away people and nothing more :D19:45
heiko1I dont really think so, but lets dont start a discussion :)19:45
blackburnI'd rather let people do whatever they want19:45
heiko1blackburn: do you know how I can paste code to ipython?19:45
heiko1intended code19:45
blackburnheiko1: nope19:45
heiko1it tells me something about %dpaste magic19:45
heiko1but I dont know that19:45
blackburnheiko1: we didn't receive any improvement of shogun :)19:45
blackburnfrom closed-source side19:46
blackburnheiko1: I am actually still tied to LGPL as covertree is LGPL19:46
heiko1I doubt that this would have been the case if we had bsd19:46
blackburnheiko1: no it won't be the case19:47
blackburnso if there is no difference..19:47
blackburnheiko1: so you started to use ipython?19:49
heiko1blackburn: yes19:49
heiko1the python terminal sucks so much I couldnt take it any longer :)19:49
blackburnheiko1: I don't use python that much but should get to it too19:49
heiko1ipython is very nice19:50
heiko1tab autocompletition19:50
heiko1etc19:50
-!- ishan__ [~ishan@14.139.122.114] has quit [Ping timeout: 257 seconds]19:53
blackburnheiko1: and moreover I do not like what FSF declare :)19:54
heiko1I tend to ignore these things :)19:54
blackburnheiko1: 'android is not free! they must share anything!'19:54
heiko1but I like the fact that people are forced to publish source code,19:54
-!- ishan [~ishan@14.139.122.114] has joined #shogun19:54
heiko1but I might be wrong with this19:54
blackburnheiko1: well imagine the case people do paid research19:55
blackburnand they are disallowed to share the code19:55
heiko1then dont use gpl19:55
blackburnthen they wouldn't use shogun, right?19:55
heiko1they could19:55
heiko1they only would have to release the shogun source19:55
blackburnheiko1: well they could definitely get trouble with gpl software19:57
blackburnthat's why I think research code should rather be bsd/mit/anything19:57
heiko1no, you only have to show your changes19:57
heiko1but not if you use shogun for something else19:57
heiko1just shogun has to be released19:57
heiko1I dont agree, I want my research/code to be available for everyone19:58
blackburnheiko1: that's the case19:58
heiko1and I want people to force to keep this19:58
blackburnthat's not my case :)19:58
blackburnI do not like to force people to do that, just if they want19:59
shogun-buildbotbuild #627 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/627  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>19:59
heiko1then use bsd :)19:59
blackburnI do not expect someone could really make something significantly better from my code19:59
heiko1blackburn: thats a good point indeed20:00
heiko1but look for example at wine20:00
heiko1codeweavers is like doing half of the development20:00
heiko1and everything *has* to go back to wine20:00
blackburnheiko1: that's totally different thing - I like that big stuff is GPL20:00
heiko1which is good, because its getting better20:00
blackburnheiko1: I actually like that shogun is gpl too (but it could restrict some users too)20:03
blackburnheiko1: but small tools, I'd rather let them free, my name here is more important for me than any restrictions - I would be just happy if somebody used it20:04
heiko1blackburn: yeah, you are right, I think it is a matter of taste20:04
heiko1however, companies can easily develop libraries that are much more sohpisticated than shogun/tapkee etc20:04
blackburnheiko1: no, I wouldn't say that's that easy20:05
heiko1as a company you want very specific things, not general frameworks like shogun20:05
heiko1so if I need kernel PCA, I just implement it for my specific application20:05
heiko1much easier to controll20:05
blackburnheiko1: yeah sure, for simple algorithms20:06
heiko1well, what is a complicated algorithm?20:06
heiko1any complicated algorithm is a horror to maintain20:06
heiko1so I would rather implement it from scratch20:07
heiko1and make sure that it works20:07
heiko1that using somebody else's code, espicially if I was a company that depends on it20:07
blackburnheiko1: I think liblinear is used though20:07
heiko1I want to have 99.999% reliability and you can never expect that from open-source software20:07
heiko1these things are different again20:07
heiko1since very small pieces of software like you said20:08
heiko1so these should be bsd style20:08
heiko1anyway, I really want to avoid these discussions :) currently fixing warnings20:08
blackburnheiko1: heh alright20:08
shogun-notifier-shogun: Sergey Lisitsyn :master * da237a4 / src/LICENSE.tapkee: https://github.com/shogun-toolbox/shogun/commit/da237a41d4932068ae37ec5121b263a03b7fcb4820:10
shogun-notifier-shogun: Added tapkee license file20:10
shogun-notifier-shogun: Sergey Lisitsyn :master * 6a89ce1 / src/shogun/multiclass/MulticlassLogisticRegression.cpp: https://github.com/shogun-toolbox/shogun/commit/6a89ce13a1961ccab9863c395461a3aed49bc83620:10
shogun-notifier-shogun: Removed unstable from MC LR20:10
blackburnheiko1: what are warnings you fix?20:15
heiko1perceptron20:15
heiko1changing example to use linearly seperable data20:15
blackburnI see20:16
heiko1blackburn:  do you know how I change the testsuite?20:17
heiko1since I changed the xamples20:17
blackburnheiko1: guys from scikit-learn redirect gsoc-newcomers directly to writing unittests20:17
blackburnheiko1: you run generator.py probably20:17
heiko1blackburn: that is a very good idea!20:17
heiko1blackburn:  what do I commit then?20:17
blackburnheiko1: it stores the output to data20:17
blackburnI think there was a way to do that for concrete example20:18
heiko1blackburn: cool so then I just commit the data?20:18
blackburnyeah20:18
heiko1but only my files right?20:19
heiko1how do I do this? :D20:19
heiko1(sorry for being so dumb with those things)20:19
blackburnheiko1: what do you need to update?20:19
heiko1perceptron classifier ex20:19
blackburnheiko1: python generator.py ../../../examples/undocumented/python_modular/classifier_averaged_perceptron_modular.py20:21
heiko1so only this example file or what?20:21
heiko1Ah20:22
heiko1I see20:22
heiko1thanks20:22
blackburnheiko1: it should update the file I think20:22
heiko1my git does not show me any changes to a data file?20:23
blackburnbad20:23
blackburnlets reverse engineer that then20:24
heiko1maybe I will just ask sonney2k, this will be faster :)20:24
heiko1I will commit examples for now and then update testsuite tomorrow or something20:24
blackburnheiko1: no it should update the file20:24
-!- ishan [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]20:25
heiko1blackburn: I will try to solve some of the integration test issues this week20:25
heiko1blackburn: maybe you could focus on the reported bugs?20:25
-!- ishan [~ishan@14.139.122.114] has joined #shogun20:25
heiko1would be great to have the testsuite in the buildbot20:25
blackburnheiko1: do we have any unclosed bugs?20:25
heiko1blackburn: vw20:26
heiko1mkl modelselection20:26
heiko1with vw I mean the streaming stuff20:26
blackburnha, what a PITA20:26
heiko1I have the feeling if we leave that now, we will never fix it20:26
heiko1and then it is lost20:26
blackburnheiko1: I will try to find some time for that this week20:26
blackburnheiko1: lets play a game20:27
heiko1Russian roulette? :D20:27
blackburnheiko1: we should commit one unit-test each day20:27
blackburnand if one of us didn't that20:27
blackburnother one cuts a finger of this unlucky boy20:28
heiko1haha20:28
heiko1let me think ...20:28
heiko1...20:28
heiko1...20:28
heiko1...20:28
heiko1NO!20:28
shogun-buildbotbuild #628 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/62820:28
heiko1 :D20:28
blackburnhaha20:28
blackburnI though so20:28
heiko1blackburn:20:29
heiko1so I changed the examples20:29
heiko1I ran generator20:29
heiko1the tester sais that examples are fine20:29
heiko1(so I assume the data has changed)20:29
heiko1my git shows that data is different20:29
heiko1but doesnt show anything20:29
blackburnheiko1: are you in ../data dir?20:29
heiko1so I dont know what has changed / whether the right stuff has changed20:29
heiko1using graphical git tool from shogun root20:30
blackburnno, enter data dir first20:30
blackburnas it is a submodule20:30
blackburndifferent repo inside a repo20:30
blackburnI heard you like lisp so I put a lisp to your lisp20:30
heiko1I like lisp???20:30
blackburnheiko1: http://static1.blip.pl/user_generated/update_pictures/1582299.jpg20:31
heiko1blackburn: haha :)20:31
blackburnheiko1: so were any changes be detected?20:32
heiko1yes :)20:32
heiko1but my gitg fails to show them to me20:32
blackburnuse console luke20:32
blackburnI don't trust GUI :D20:32
heiko1gitg is so buggy :)20:33
heiko1and then? normal push?20:34
blackburnheiko1: yes push that file20:34
heiko1how to send it back?20:34
blackburnheiko1: back to?20:34
heiko1I committed locally20:34
heiko1how do I send it to shogun git20:34
blackburnthere is a command20:34
blackburnlet me remember20:34
blackburngit push !20:34
blackburn:D20:34
blackburnjk, just git push20:34
heiko1fatal: remote error:20:34
heiko1  You can't push to git://github.com/shogun-toolbox/shogun-data.git20:34
heiko1  Use git@github.com:shogun-toolbox/shogun-data.git20:34
blackburnoh okay20:35
heiko1should I do that?20:35
blackburnsurgery20:35
blackburnheiko1: wait a minute20:35
blackburnheiko1: git remote set-url origin git@github.com:shogun-toolbox/shogun-data.git20:35
heiko1To git@github.com:shogun-toolbox/shogun-data.git20:36
heiko1 ! [rejected]        master -> master (non-fast-forward)20:36
heiko1error: failed to push some refs to 'git@github.com:shogun-toolbox/shogun-data.git'20:36
heiko1hint: Updates were rejected because a pushed branch tip is behind its remote20:36
heiko1hint: counterpart. If you did not intend to push that branch, you may want to20:36
heiko1hint: specify branches to push or set the 'push.default' configuration20:36
heiko1hint: variable to 'current' or 'upstream' to push only the current branch.20:36
heiko1uh20:36
blackburngit push origin master20:36
heiko1blackburn:  would you mind updating the data?20:36
heiko1I will push the new examples20:37
blackburnheiko1: just specify a branch and you will push it20:37
heiko1can I safely do git pull in the xamples?20:37
heiko1sorry in the data/ dir20:37
blackburnpull? what for?20:38
heiko1heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/data$ git push origin master20:38
heiko1To git@github.com:shogun-toolbox/shogun-data.git20:38
heiko1 ! [rejected]        master -> master (non-fast-forward)20:38
heiko1error: failed to push some refs to 'git@github.com:shogun-toolbox/shogun-data.git'20:38
heiko1hint: Updates were rejected because a pushed branch tip is behind its remote20:38
heiko1hint: counterpart. Check out this branch and merge the remote changes20:38
heiko1hint: (e.g. 'git pull') before pushing again.20:38
heiko1hint: See the 'Note about fast-forwards' in 'git push --help' for details.20:38
heiko1heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/data$20:38
blackburnjust git pull --rebase then20:38
heiko1unstaged changes argh20:38
blackburngit stash20:38
blackburngit pull --rebase20:38
blackburngit stash pop20:38
blackburngit push origin master20:38
heiko1same error20:39
heiko1I give up now20:39
blackburnheiko1: git status?20:39
heiko1(sorry my tolerance with this is low, I know :)20:39
blackburnheiko1: how do you use linux? I feel I spent half a lifetime being stucked20:39
blackburn:D20:39
heiko1blackburn: I know, I did this, but now I just want to update the example you know, not spend hours fighting with git20:40
blackburnheiko1: git status!20:40
heiko1not currently in any branch20:40
blackburnheiko1: the more you fight the more precise your kung-fu is!20:40
blackburnheiko1: so we have a winner!20:41
blackburnheiko1: git reset HEAD~120:41
heiko1just forced this20:41
heiko1yes20:41
blackburngit checkout master20:41
heiko1this always happens  when I try to use git :)20:41
blackburncommit, push20:41
blackburnheiko1: git don't cry ;)20:42
blackburnheiko1: git it happens20:42
heiko1blackburn:20:42
heiko1same error20:42
heiko1heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/data$ git push20:42
heiko1To git@github.com:shogun-toolbox/shogun-data.git20:42
heiko1 ! [rejected]        master -> master (non-fast-forward)20:42
heiko1error: failed to push some refs to 'git@github.com:shogun-toolbox/shogun-data.git'20:42
heiko1hint: Updates were rejected because the tip of your current branch is behind20:42
heiko1hint: its remote counterpart. Merge the remote changes (e.g. 'git pull')20:42
heiko1hint: before pushing again.20:42
heiko1hint: See the 'Note about fast-forwards' in 'git push --help' for details.20:42
heiko1heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/data$20:42
blackburnheiko1: git stash && git pull --rebase && git stash pop && git push origin master?20:42
heiko1that worked, lets see :)20:43
blackburnheiko1: lesson #1 : don't use git20:43
blackburn:D20:43
heiko1how can I see whether I was sucessful?20:43
blackburngit status20:43
blackburngit pull20:43
heiko1blackburn: I seriously hate complicated things :D20:44
heiko1but how do I know that I just did not delete my new data?20:44
heiko1where can I see whether its in data now20:44
blackburnheiko1: https://github.com/shogun-toolbox/shogun-data20:44
blackburnwe have nothing out there20:44
blackburnheiko1: run generator again I think20:44
shogun-notifier-shogun: Heiko Strathmann :master * 92439e3 / src/shogun/multiclass/MulticlassLogisticRegression.cpp: https://github.com/shogun-toolbox/shogun/commit/92439e34ed87bda59a58be24ee52169bb877f38f20:44
shogun-notifier-shogun: removed unstable warning and added some proper assertion messages20:44
shogun-notifier-shogun: Heiko Strathmann :master * 83bc042 / examples/undocumented/python_modular/classifier_perceptron_modular.py: https://github.com/shogun-toolbox/shogun/commit/83bc0426c9801ae745ac509843068a019652a84320:44
shogun-notifier-shogun: changed training data to be (probably) linearly seperable to avoid warnings20:44
shogun-notifier-shogun: Heiko Strathmann :master * ac2044b / examples/undocumented/python_modular/classifier_averaged_perceptron_modular.py: https://github.com/shogun-toolbox/shogun/commit/ac2044bd27f3c7162de362d6bef2b0621c66dfa320:44
shogun-notifier-shogun: changed training data to be (probably) linearly seperable to avoid warnings20:44
shogun-notifier-shogun: Heiko Strathmann :master * 56ce595 / examples/undocumented/python_modular/classifier_averaged_perceptron_modular.py: https://github.com/shogun-toolbox/shogun/commit/56ce595b10cdd742414f99b95552b6738d76d38a20:44
shogun-notifier-shogun: fixed copy/paste error20:45
shogun-notifier-shogun: Heiko Strathmann :master * 2f9db76 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/2f9db76ec1bef4e598a765ffd50cfd67c6ce32df20:45
shogun-notifier-shogun: Merge pull request #914 from karlnapf/master20:45
shogun-notifier-shogun:20:45
shogun-notifier-shogun: example remove warning20:45
heiko1blackburn: tells me it did not work20:45
blackburnheiko1: you weren't at any branch.. just run generator again20:45
blackburnand commit as usual20:45
heiko1blackburn: sucess!!20:47
heiko1thanks :)20:47
heiko1thanks for looking after my stupidity, I think I know how it works now20:47
heiko1blackburn: will do some cooking now, see you later!20:47
heiko1blackburn: we should be at zero waninngs with that now20:48
blackburnheiko1: see you!20:51
shogun-buildbotbuild #630 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/630  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>20:52
-!- ishan_ [~ishan@14.139.122.114] has joined #shogun20:54
-!- ishan [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]20:54
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun21:07
heiko1blackburn: larank is totally broken, ouputs nonsense21:10
heiko1when was it last edited?21:10
heiko1or maybe its the multiclass stuff?21:10
blackburnheiko1: that's pretty nice21:10
heiko1I will write a unit test on linearly seperable data ...21:11
blackburnheiko1: I used larank year ago may be21:11
heiko1ok21:12
heiko1so first step is unit test :)21:12
heiko1I am so glad we have those!21:12
blackburnheiko1: yes ideally the software should be made of unit-tests first21:12
heiko1blackburn: agreed21:12
heiko1but we can force this from now on21:13
heiko1blackburn:  lol21:18
heiko1larank example asserts that batch output and output is the same21:18
heiko1but doesnt check whether training labels are similar to training predictions :)21:18
heiko1(which is what is wrong=)21:18
blackburnheiko1: hehe21:18
blackburnheiko1: C++ is actually a bad lang for proper testing unfortunately21:22
heiko1I guess for us its different due to swig stuff etc21:22
blackburnheiko1: in say java it is much easier to analyze stuff21:22
blackburnheiko1: yes 2x21:22
heiko1indeed21:22
heiko1I would go for python everything, but its so cool with all the interfaces21:22
-!- ishan__ [~ishan@14.139.122.114] has joined #shogun21:25
-!- ishan_ [~ishan@14.139.122.114] has quit [Read error: Connection reset by peer]21:25
-!- heiko1 [~heiko@b019b9ef.bb.sky.com] has quit [Quit: Leaving.]21:49
-!- ishan__ [~ishan@14.139.122.114] has quit [Remote host closed the connection]21:49
-!- heiko [~heiko@b019b9ef.bb.sky.com] has joined #shogun21:49
@sonney2kheiko, so you got it to work?22:03
@sonney2kpython generator.py example.py22:03
@sonney2kand then cd ../data22:03
@sonney2kand on master branch commit to your github remote (not the standard origin)22:04
blackburnsonney2k: yes he did22:05
@sonney2kok then well done :)22:05
blackburnsonney2k: the only issue that arises with new submodule init - the submodule repo is not at any branch22:05
blackburnwhich means you lose your commits22:06
@sonney2kwhy would one do submodule init again?22:06
blackburnsonney2k: no reason but I guess heiko didn't commit anything yet22:06
@sonney2kblackburn, heiko are you making progress wrt release?22:10
blackburnsonney2k: I think so22:10
@sonney2kblackburn, why did you add a separate license file for tapkee?22:11
@sonney2kblackburn, ok22:11
@sonney2knice22:11
blackburnsonney2k: BSD :)22:11
@sonney2kyeah but don't you have that in each file anyways?22:12
blackburnsonney2k: I don't22:12
@sonney2kI mean inside of shogun it is all GPL anyways22:12
blackburnsonney2k: we actually should include liblinear license too22:12
@sonney2kno why?22:12
blackburnthey want it22:12
@sonney2kliblinear / tapkee inside shogun are GPL ...22:12
blackburnClearly indicate that LIBSVM is used.22:13
blackburnRetain the LIBSVM COPYRIGHT file in your software.22:13
blackburnwell it is just what they say in faq22:13
blackburnsonney2k: I didn't know if i should place a license file but I am referring it in sources so decided to put it in too22:13
blackburnsonney2k: I hope you are ok with bsd ;)22:13
@sonney2kblackburn, I don't care under which license you put your software. And as I said inside of shogun it is GPL anywyas22:14
blackburnsonney2k: yeah22:14
@sonney2kbut for me BSD would not be an option :)22:16
blackburnsonney2k: why?22:16
@sonney2kI like the GPL idea22:16
blackburnsonney2k: we talked with heiko - I like GPL idea for serious stuff22:16
@sonney2keverything is serious22:17
blackburnbut for useless crap like my software something more restrictive is better (for me)22:17
@sonney2kless restrictive you mean?22:19
blackburnsonney2k: yeah sure22:20
blackburnsonney2k: we are measuring DR stuff now22:23
blackburnsonney2k: waffles uses super mega method for computing eigenvectors22:23
blackburn100 power iterations22:23
blackburnthat should be patented..22:23
blackburnno more no less, 10022:23
blackburnsonney2k: n4nd0 ran it yesterday and it didn't converge after 12 hours (that takes a few seconds for normal impls)22:24
blackburnwe don't know what to write into the table in the paper..22:24
heikosonney2k: I think I got it how to do the testing, currently fixing la rank, thought there was a bug, but it was just nonsense input data (do not except classifier to do good stuff on noise :), added unit test and will change the data soon22:46
n4nd0sonney2k: are you also getting those million mails about the webpage crashing?22:54
n4nd0sonney2k: I have not looked into it but I think it might be caused because some page has been modified added from the admin interface22:54
@sonney2kn4nd0, no then it should not give errror msg's22:59
@sonney2kheiko, cool thanks!22:59
blackburnsonney2k: n4nd0: where from do you get mails?23:00
@sonney2kblackburn, the website23:01
blackburnsonney2k: it sends a message on crash?23:01
@sonney2ksure23:01
blackburncool functionality23:01
@sonney2kbacktraces23:01
shogun-notifier-shogun: Sergey Lisitsyn :master * 13ba0c7 / src/shogun/lib/tapkee/ (3 files): https://github.com/shogun-toolbox/shogun/commit/13ba0c7f37e58c23b0fc0a70f54933d428167f8123:02
shogun-notifier-shogun: A few patches for tapkee23:02
blackburnheiko: I started a oneday-onetest game!23:05
heikoblackburn: I already updated two examples and fixed the integration tests!23:05
blackburnI think I'll lose my finger23:06
blackburn:D23:06
shogun-notifier-shogun: Sergey Lisitsyn :master * 84e4b13 / tests/unit/converter/Isomap_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/84e4b13917bd2e0d86e2640df00db768fcf4062023:06
shogun-notifier-shogun: Added isomap unittest23:06
blackburnuh I will have finger today23:07
heikohehe :)23:07
blackburnsonney2k: did you manage to send a mail to quoc and olivier?23:11
n4nd0sonney2k: why not?23:11
n4nd0sonney2k: if some page is added and the attributes are not correctly set then it may crash23:11
n4nd0I find no other reason to why it has been without problems for a couple of weeks and they started yesterday23:12
shogun-buildbotbuild #631 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/63123:21
n4nd0good night guys!23:25
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]23:26
heikoblackburn: what is the multiclass data in the shogun data dir like?23:43
heikodoes it make any sense?23:43
blackburnheiko: I don't know actually23:43
heikoare the class pairs linearly seperable?23:43
blackburnit wasn't done by me I am afraid23:44
heikook23:44
blackburnit was here before I came :)23:44
heikoblackburn: I think it is nonsense data23:46
blackburnheiko: might be..23:46
heikowhats the point in measuring if a method does the same if this was nonsense?23:46
blackburnheiko: that's a secret chinese method23:46
heikoblackburn:  another integration test now works!23:57
heikoblackburn:  multiclass ocas?23:57
heikofails on a toy problem23:57
blackburnheiko: nice23:57
blackburnheiko: fails?23:57
blackburnheiko: it worked too back in a day23:57
heiko[ 0.  1.  2.  0.  1.  2.  0.  1.  2.  0.]23:57
heiko[ 2.  1.  1.  2.  1.  1.  2.  2.  2.  2.]23:57
heikotop is test label23:57
heikobelow is predictions23:57
heikoeasy multiclass data23:58
heikohas as many dimensions as classes and the dimension of the class is shifted23:58
--- Log closed Mon Mar 11 00:00:10 2013