Open in new window / Try shogun cloud
--- Log opened Wed Jun 05 00:00:03 2013
--- Day changed Wed Jun 05 2013
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun00:00
-!- gsomix [~gsomix@188.168.3.7] has quit [Ping timeout: 245 seconds]00:05
-!- gsomix [~gsomix@188.168.3.7] has joined #shogun00:06
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has joined #shogun01:46
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has left #shogun ["QUIT :Leaving."]01:53
-!- gsomix [~gsomix@188.168.3.7] has quit [Ping timeout: 256 seconds]02:58
-!- gsomix [~gsomix@188.168.3.7] has joined #shogun02:58
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Read error: No route to host]03:00
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun03:03
-!- nube [~rho@49.244.20.247] has quit [Ping timeout: 256 seconds]03:05
-!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has joined #shogun04:31
-!- gsomix [~gsomix@188.168.3.7] has quit [Ping timeout: 245 seconds]04:54
-!- nube [~rho@36.253.65.127] has joined #shogun04:54
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 248 seconds]05:04
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]05:15
-!- nube1 [~rho@36.253.65.127] has joined #shogun05:26
-!- nube [~rho@36.253.65.127] has quit [Ping timeout: 248 seconds]05:27
-!- nube [~rho@116.90.228.137] has joined #shogun05:33
-!- nube1 [~rho@36.253.65.127] has quit [Ping timeout: 248 seconds]05:36
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has joined #shogun05:52
lambdaysonney2k: good morning :)05:52
@sonney2klambday, hey!05:54
@sonney2klambday, if you have time please fix the build :) I mean the rng init might help...06:07
lambdaysonney2k: there is another problem.. in the Complex unittest (which just checks for abs and do not have any permutation before that) also fails for 1e-15..06:07
lambdaywhat could be the reason for that one? :(06:07
-!- gsomix [~gsomix@83.234.169.64] has joined #shogun06:09
@sonney2klambday, that should not happen06:11
lambdaysonney2k: I checked buildbot logs and it shows failure for that test too :(06:13
-!- gsomix [~gsomix@83.234.169.64] has quit [Ping timeout: 248 seconds]06:16
-!- gsomix [~gsomix@83.234.169.64] has joined #shogun06:16
@sonney2klambday, who is computing abs(a)06:18
@sonney2kfor a being a complex type?06:18
gsomixgood morning06:18
@sonney2kis this in std:: ?06:19
lambdayoh.. its the std::abs... I should have used CMath::abs() for complex, but that also currently calls std::abs() only06:19
@sonney2klambday, then you found in a bug in std::06:19
@sonney2kthe result is just wrong on bsd06:20
@sonney2klambday, using an arbitrary precision calculator the result should be06:20
@sonney2k7.8102496759066543941306:20
@sonney2k; (5**2+6**2)^0.5 - 7.8102496759066539766506:21
@sonney2k0.0000000000000004174806:21
@sonney2k; (5**2+6**2)^0.5 - 7.810249675906655806:21
@sonney2k-0.0000000000000014058706:21
@sonney2kis what we see otherwise06:21
@sonney2kso <1e-15 is what you have with the number you gave as reference06:21
@sonney2kbut the other is really off06:21
@sonney2kgsomix, morning! today is test day right?06:22
lambdaysonney2k: okay.. I should change the reference too then..06:22
lambdaybut what to do with this bsd? I am not sure :(06:22
@sonney2kthere is no fix06:22
@sonney2kwe can only go to 1e-1406:23
@sonney2kprecision06:23
lambday:(06:23
lambdayokay06:23
gsomixsonney2k, yep.06:23
@sonney2kand then ask wiking to install newer std stuff06:23
@sonney2kgtg, cu06:23
lambdaysonney2k: okies06:23
lambdayciao :)06:23
lambdayI'll submit the PR soon06:23
-!- nube [~rho@116.90.228.137] has quit [Ping timeout: 248 seconds]06:25
-!- foulwall_ [~foulwall@2001:da8:215:503:8c13:264:5b04:37a7] has joined #shogun06:26
-!- nube [~rho@36.253.65.127] has joined #shogun06:41
-!- nube1 [~rho@116.90.228.137] has joined #shogun06:53
-!- nube [~rho@36.253.65.127] has quit [Ping timeout: 248 seconds]06:54
-!- gsomix [~gsomix@83.234.169.64] has quit [Ping timeout: 248 seconds]07:37
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]07:47
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has joined #shogun08:17
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun08:18
shogun-notifier-shogun: van51 :develop * d4115d5 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/d4115d5512d5899abb983b3b428a4346d708785b08:18
shogun-notifier-shogun: Changed CList to CDynamicObjectArray in CCombinedDotFeatures (WIP)08:18
shogun-notifier-shogun: Soeren Sonnenburg :develop * 2eb8ee7 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/2eb8ee78c046490ec87ce2b64d92c3f66743f63708:18
shogun-notifier-shogun: Merge pull request #1138 from van51/feature/dotfeats08:18
shogun-notifier-shogun:08:18
shogun-notifier-shogun: Changed CList to CDynamicObjectArray in CCombinedDotFeatures08:18
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has quit [Client Quit]08:21
shogun-buildbotbuild #1071 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1071  blamelist: Soeren Sonnenburg <sonne@debian.org>08:35
shogun-buildbotbuild #903 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/903  blamelist: Soeren Sonnenburg <sonne@debian.org>08:40
shogun-buildbotbuild #904 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/904  blamelist: van51 <vangelis_51@hotmail.com>08:40
shogun-buildbotbuild #1072 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/107208:42
-!- gsomix [~Miranda@r206-10.smr.ru] has joined #shogun08:43
-!- lisitsyn [~blackburn@37.61.181.124] has joined #shogun08:44
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Quit: Leaving.]08:48
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun08:48
gsomixlisitsyn: hey, listing. do you want me to do fourier transformation of your cat? do you have a cat?08:49
lisitsynno I don't08:50
lisitsyn:(08:50
gsomixsonne|work: and you? have a cat?08:50
-!- foulwall_ [~foulwall@2001:da8:215:503:8c13:264:5b04:37a7] has quit [Remote host closed the connection]08:54
-!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has quit [Ping timeout: 260 seconds]08:59
-!- lisitsyn [~blackburn@37.61.181.124] has left #shogun []09:09
sonne|workgsomix: no but feel free to transform my fish :D09:10
-!- iglesiasg [~iglesiasg@193.147.77.22] has joined #shogun09:15
-!- mode/#shogun [+o iglesiasg] by ChanServ09:15
shogun-buildbotbuild #1196 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/119609:18
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun09:36
-!- mode/#shogun [+o lisitsyn] by ChanServ09:36
@lisitsynsonne|work: I've started on model selection syntax09:36
sonne|worknice09:37
gsomixok, signal processing...09:39
-!- gsomix [~Miranda@r206-10.smr.ru] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]09:39
-!- nube1 [~rho@116.90.228.137] has quit [Ping timeout: 264 seconds]09:59
-!- nube [~rho@116.90.228.137] has joined #shogun10:02
-!- nube [~rho@116.90.228.137] has quit [Ping timeout: 245 seconds]10:20
-!- nube [~rho@36.253.75.217] has joined #shogun10:41
-!- nube1 [~rho@116.90.228.137] has joined #shogun10:51
-!- nube [~rho@36.253.75.217] has quit [Ping timeout: 248 seconds]10:55
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has joined #shogun10:55
lambdaysonney2k: hi.. I changed our CMath::abs(complex64_t a) to not to make use of std::abs() but rather calculate it10:57
lambdaysonney2k: I think this should work in bsd..10:57
lambdaythe SGVector test error also arises from abs()...10:57
lambdaybut I can't check whether it works or not in bsd.. should I keep the tests as 1e-15 and if this time it doesn't work then make it 1e-14?10:58
lambdaysonne|work:10:59
sonne|worklambday: yes makes sense11:00
lambdaysonne|work: okay.. then sending a PR with just that11:02
@iglesiasglambday, I can merge that now11:09
@iglesiasgwhy did you change the constant in the comparisons?11:09
lambdayiglesiasg: okay :) thanks11:09
@iglesiasgthat is the new value you get when you run it with CMath::abs?11:09
lambdayiglesiasg: no.. this is what I get using bc11:09
@iglesiasgbc?11:09
lambdaysimilar to what sonne|work: wrote this morning11:09
lambdayarbitrary precision calculator or something11:10
@iglesiasgaham ok11:10
@iglesiasglambday, so did you use another program to get that value?11:10
lambdayiglesiasg: yes11:10
lambdayfor these two values11:10
@iglesiasglambday, ok, matlab or so?11:10
lambdayno.. its called bc11:11
@iglesiasgand it runs well on your machine11:11
@iglesiasgaaah ok, got it now :)11:11
lambdayhttp://www.gnu.org/software/bc/manual/html_mono/bc.html11:11
@iglesiasgI thought bc was a generic name for arbitrary precision calculators11:11
@iglesiasgthough I didn't get where the shortform could come from11:11
lambdayiglesiasg: my mistake :( I should have been more clear with it.. actually I didn't know about this until this morning11:12
@iglesiasglambday, no problem :)11:12
@iglesiasglet's just check that the test is fine on travis build11:12
sonne|workiglesiasg: matlab just uses double too11:12
@iglesiasgand later hope it goes well on bsd11:12
@iglesiasgsonne|work, aham11:13
lambdayyes.. the value that c++ program gives lie within the range 1e-15 of the value that bc gives11:13
sonne|workso it suffers the same issues11:13
@iglesiasgsonne|work, this arbitrary precision calculator use a list of digits?11:13
@iglesiasgor similar idea?11:13
sonne|workyou tell it how many digits you want11:13
shogun-notifier-shogun: lambday :develop * c8cc427 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/c8cc427435b1ab974b66afe409b5f7d4b9803dff11:14
shogun-notifier-shogun: CMath::abs(complex64_t) changed to not to use std::abs()11:14
shogun-notifier-shogun: Fernando Iglesias :develop * a851de1 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/a851de1f6dd25e4405f34d08edd9451f1f34f9f311:14
shogun-notifier-shogun: Merge pull request #1148 from lambday/develop11:14
shogun-notifier-shogun:11:15
shogun-notifier-shogun: CMath::abs(complex64_t) changed to not to make use of std::abs()11:15
lambdaythanks iglesiasg: hope this works11:15
* lambday keeping his fingers crossed11:15
@iglesiasgsonne|work, I actually think there is something in matlab to set more precision than double11:17
sonne|workiglesiasg: no there is not11:17
@iglesiasgok :)11:17
sonne|workiglesiasg: type 1-1.0111:18
sonne|workin matlab11:18
sonne|workand set format to the best possible11:18
sonne|workthen you will see11:18
@iglesiasgsonne|work, http://www.mathworks.se/help/symbolic/vpa.html11:19
sonne|workIIRC11:19
sonne|workformat long11:19
sonne|work1-1.0111:19
@iglesiasgI can be wrong11:19
sonne|workfor some symbolic stuff yeah11:20
-!- naywhaya1e [~ryan@spoon.lugatgt.org] has joined #shogun11:24
-!- Netsplit *.net <-> *.split quits: @sonney2k, thoralf, naywhayare11:30
-!- Netsplit over, joins: @sonney2k, thoralf11:31
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun11:34
@iglesiasglambday, it seems the test went fine on BSD11:35
@iglesiasgat least this time :)11:35
lambdayiglesiasg: ah, finally :D11:36
lambdayyippie :D11:36
@iglesiasgmmm although it is building again11:36
-!- nube1 [~rho@116.90.228.137] has quit [Quit: Leaving.]11:36
-!- nube [~rho@116.90.228.137] has joined #shogun11:37
lambdaybuilding twice? one for the commit and another for merge?11:37
@iglesiasgI think so11:38
@iglesiasglambday, what I find funny is that it seems it first builds the commit with the merge11:39
lambdayyes..11:39
shogun-buildbotbuild #905 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/905  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:39
lambdaycyg1 :'(11:40
* lambday feels like watching last extra 3 mins in a football match11:40
@iglesiasglambday, cygwin is not your fault11:40
@iglesiasgit was already screwed :D11:40
@iglesiasgit actually configured well this time11:41
lambdaysse2 not found or something11:41
lambdaynot enabled11:41
lambdayyay... bsd works :D11:43
@lisitsynbah11:43
@lisitsyndid anyone touch libocas recently?11:43
@iglesiasgno idea11:44
@iglesiasgnothing that I can remember11:44
-!- HeikoS [~heiko@nat-173-83.internal.eduroam.ucl.ac.uk] has joined #shogun11:49
-!- mode/#shogun [+o HeikoS] by ChanServ11:49
thoralfsonney2k: I put all local changes into a single commit, but I cannot commit them into the remote branch, since it already contains the "old" commits.11:49
thoralfsonney2k: Would push --force mess something up?11:50
shogun-buildbotbuild #906 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/906  blamelist: lambday <heavensdevil6909@gmail.com>11:50
lambdayhi HeikoS11:54
@HeikoShi11:54
lambdaychecking your comments11:55
@HeikoScommenting ... :)11:55
@HeikoSlambday: done11:55
lambdayHeikoS: okay.. I'm checking :)11:55
lambdayHeikoS: alright..11:56
lambdayI'm making the changes then11:56
@iglesiasglisitsyn, do you know why the buildbot builds first the merge commit and then the other?11:58
@iglesiasgit looks weird to me, since the other one is done first11:58
sonne|workthoralf: well not for the PR11:58
sonne|workbut for the general repository yeah kaboom time :D11:58
thoralfsonne|work: So you don't know other ways than 1. a new PR or 2. push --force?12:00
thoralfsonne|work: "but for the general repository yeah kaboom time" <-- parse error ;)12:00
sonne|workthoralf: push --force is what we usually do12:01
sonne|work(for PRs!)12:01
@lisitsyniglesiasg: no don't know why that12:02
sonne|workiglesiasg: I think it is building the master branch for some reason12:02
@iglesiasgsonne|work, mmm but it seems to build the changes were are pushing these days, that go to develop branch12:03
sonne|workiglesiasg: well you can see what it built http://shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun12:04
@iglesiasgsonne|work, yep, that is develop12:05
sonne|workno idea then12:05
wikinglisitsyn: not that i know of12:16
wiking(libocas)12:16
@iglesiasgwiking, any idea why shogun buildbot builds the commit in that order?12:19
@iglesiasgsay there is a PR with a commit then someones merge it12:20
@iglesiasgit builts first the merge, then the original commit12:20
lambdayHeikoS: could please have a look at : https://github.com/shogun-toolbox/shogun/pull/1140/files#L19R50812:23
wikingiglesiasg: lol... sorry no idea12:28
@iglesiasgthen we leave it as x-file :D12:29
@HeikoSlambday: good that you found that one!12:34
lambdayHeikoS: more unit-tests? :(12:35
lambdayokay adding12:35
@HeikoSlambday: not your fault, but this nicely illustrates why one should always add at least one per method that makes sure no nonsense is produced12:35
@HeikoSif you could add one for this, this would be very helpful12:36
lambdayHeikoS: yes..12:37
lambdayeven I missed a few for complex and SGVector.. I'll make sure that each method is called at least once.. (may not need a separate test case for each method then)12:38
lambdayor a separate one for each one is better?12:38
-!- Netsplit *.net <-> *.split quits: @sonney2k, thoralf12:40
@HeikoSseperate is better12:41
lambdayalright12:41
@HeikoSlambday: but depends on how much work you want to invest12:41
@HeikoSif its separated, its way easier to track errors12:41
@HeikoStests should be as simple as possible12:42
@HeikoShave  alook at the SGObject tests I recently added12:42
lambdayokay12:42
@HeikoS20 of them or so, since it allows to quickly see whats wrong12:42
lambdayand for the parameters, I should add test cases too in the TParameter tests, right?12:43
lambdayfor complex64_t I mean12:43
@HeikoSlambday: question is what to test there12:43
@HeikoSbut actually, yes that would be helpful12:43
lambdaycompare_ptype12:43
@HeikoSsince then we have a fine grid of whats going on12:43
@HeikoSrather than "serialization doesnt work"12:43
lambdayokay.. I'll check it12:44
lambdaywill add that too in the PR then? including the edits for error msgs?12:44
-!- Netsplit over, joins: @sonney2k, thoralf12:45
@HeikoSlambday: yes please do12:48
lambdayHeikoS: alright :)12:48
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun12:55
-!- nube [~rho@116.90.228.137] has quit [Remote host closed the connection]12:58
-!- iglesiasg [~iglesiasg@193.147.77.22] has quit [Quit: Leaving]13:14
-!- gsomix [~Miranda@83.234.54.93] has joined #shogun13:15
gsomixpassed13:15
@lisitsynaway? teacher passed away? :D13:16
gsomixlisitsyn: huh, no. then I'd have failed the exam!13:20
thoralfgsomix: Dying teacher implies failed exam?13:21
thoralfgsomix: Drivers license? :)13:21
gsomixthoralf: no, just exam at university.13:22
thoralfgsomix: I wasn't serious. ;)13:23
thoralfgsomix: I was just wondering, which kind of exam fails if the teacher passes away... :)13:23
gsomixthoralf: ok. I'm a little inadequate now. >:313:24
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun13:26
-!- mode/#shogun [+o iglesiasg] by ChanServ13:28
-!- gsomix [~Miranda@83.234.54.93] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]13:48
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving]13:52
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun14:12
@lisitsynvotjakovr: hey, when do you have exams?14:13
-!- naywhaya1e is now known as naywhayare14:14
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]14:15
votjakovrlisitsyn: hi, in june, but i'm already free14:18
@HeikoSvotjakovr: hi!14:20
@HeikoShow is it going with the GPs?14:20
votjakovrHeikoS: hi, please wait a minute, i'd like to send a PR with moving of latent features to fitc14:22
@lisitsynvotjakovr: no exams?14:26
votjakovrlisitsyn: almost :)14:28
-!- Netsplit *.net <-> *.split quits: @sonney2k, thoralf14:30
votjakovrHeikoS: mmm... sorry, integration tests are failed, i'm finding the reason14:31
@HeikoSvotjakovr: did you change any class members? variables? or any implementation details?14:31
votjakovrHeikoS: yep, but diff is empty14:32
-!- nube [~rho@116.90.228.137] has joined #shogun14:33
-!- Netsplit over, joins: @sonney2k, thoralf14:34
-!- HeikoS [~heiko@nat-173-83.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]14:46
-!- HeikoS [~heiko@nat-173-83.internal.eduroam.ucl.ac.uk] has joined #shogun15:16
-!- mode/#shogun [+o HeikoS] by ChanServ15:16
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has joined #shogun15:19
lambdayHeikoS: unit-test for float64_t seems okay?15:21
@HeikoSlambday: traivs fails15:21
@HeikoSCMath::sinh(a) evaluates to 162.07812441272409,162.07812441272406545067 evaluates to 162.07812441272407, and1E-15 evaluates to 1.0000000000000001e-15.15:21
lambdayerr15:22
lambdayno idea why this happens!15:22
lambdayit passed all the tests here :(15:23
@HeikoSdont know, maybe some library issues?15:23
@HeikoSI will get some lunch now, be back in an hour15:23
lambdayI think so15:23
lambdayHeikoS: okay, even I'm leaving.. be back in an hour too15:23
lambdaywill fix it then15:23
@HeikoSok, see you then!15:24
lambdaysee you :)15:24
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Quit: Page closed]15:27
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has quit [Quit: lambday]15:28
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Read error: Connection reset by peer]15:34
-!- nube [~rho@116.90.228.137] has quit [Quit: Leaving.]15:36
-!- vgorbati [~vgorbati@212.2.159.34] has joined #shogun16:05
-!- vgorbati [~vgorbati@212.2.159.34] has quit [Read error: Operation timed out]16:08
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun16:15
-!- vgorbati [~vgorbati@212.2.159.34] has joined #shogun16:17
votjakovrHeikoS: there is something strange: one of two python integration tests reports about an error and diff is empty, but another one is OK.16:20
@HeikoSvotjakovr: indeed strange16:20
@HeikoSis this related to an example where you changed things?16:20
-!- mode/#shogun [+o iglesiasg] by ChanServ16:20
votjakovrHeikoS: yep16:21
@HeikoSvotjakovr: then maybe rewind your changes a bit to see which part borke it?16:22
@HeikoSbroke16:22
votjakovrHeikoS: huh... Ok, I'll try to do so...16:23
@HeikoSvotjakovr: maybe also show the full error message to me or sonney2k first16:24
votjakovrDetailled failures:16:26
votjakovrregression_gaussian_process_modular.py setting 1/216:26
votjakovr====DIFF================16:26
votjakovr====EOT================16:26
votjakovrThat's all16:27
@HeikoSthats weird indeeed16:27
@HeikoSno idea what can cause this16:27
votjakovrbtw i when run this test manually and plot some stuff, everything is OK16:28
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has joined #shogun16:29
@HeikoSI see16:29
@HeikoSso then try to find out when this occured16:29
@HeikoSthis usually solves such problems16:29
lambdayHeikoS: hi16:33
lisitsyn1votjakovr: HeikoS: looks like a crash for me16:36
lisitsyn1could be that I mean16:36
-!- Netsplit *.net <-> *.split quits: @sonney2k, thoralf16:36
votjakovrlisitsyn1: I think so, but I when run this test manually and it doesn't report anything16:38
votjakovrlisitsyn1: I mean segfaults, etc16:39
-!- Netsplit over, joins: @sonney2k, thoralf16:49
@HeikoSlisitsyn1: do we have group lasso in shogun?16:50
@HeikoSand in particular, group lasso for logistic regression (aka classification) ?16:50
@HeikoSsonne|work, sonney2k ^16:50
sonne|workHeikoS: IIRC lisitsyn added that last year16:51
-!- sonney2k [~shogun@7nn.de] has quit [Ping timeout: 270 seconds]16:51
-!- sonney2k [~shogun@7nn.de] has joined #shogun16:52
lisitsyn1HeikoS: yes multiclass group lasso is here16:52
@HeikoSlisitsyn1: which class?16:52
@HeikoSand: does it work? :D16:52
lisitsyn1HeikoS: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/multiclass/MulticlassTreeGuidedLogisticRegression.cpp16:52
lisitsyn1HeikoS: in general it is tree group lasso16:53
lisitsyn1HeikoS: well I hope it works16:53
@HeikoSno documentation  ....16:53
@HeikoSwhich algorithm is implemented in there?16:53
lisitsyn1HeikoS: algorithm of learning?16:54
@HeikoSso is it based on the multiclass framework and does OvO ?16:54
lisitsyn1HeikoS: OvR16:54
@HeikoSand how is the lasso problem solved16:54
@HeikoSokay so its core is logistic regression I guess?16:54
lisitsyn1HeikoS: euclidean projection16:54
lisitsyn1IIRC it is called that16:54
@HeikoSwhere did you take this from?16:55
@HeikoSwhere is the algorithm described?16:55
lisitsyn1HeikoS: http://www.public.asu.edu/~jye02/Software/SLEP/16:55
lisitsyn1HeikoS: http://www.public.asu.edu/~jye02/Software/SLEP/manual.pdf16:55
lisitsyn1HeikoS: tree_mcLogisticR16:55
@HeikoScool thanks!16:56
@HeikoSmaybe this should be noted in the shogun documentation16:56
@HeikoSotherwise no-one will ever use it16:56
lisitsyn1HeikoS: optimization wise it is some moreau-yosida proximal stuff16:56
@HeikoSespecially in case you suddenly die ;)16:57
lisitsyn1implemented in the 'altra' function16:57
lisitsyn1HeikoS: so basically accelerated projected gradient descend or so16:57
@HeikoSlisitsyn1:  is there an example how to use it?16:57
lisitsyn1HeikoS: let me check16:58
lisitsyn1HeikoS: https://github.com/shogun-toolbox/shogun/blob/develop/examples/undocumented/python_modular/classifier_featureblock_logistic_regression.py16:59
lisitsyn1like that16:59
@HeikoScool thanks!16:59
lisitsyn1HeikoS: the group stuff is the same16:59
@HeikoSso this block stuff defines the groups?16:59
lisitsyn1yes16:59
@HeikoSI in fact never used such methods, so forgive me my stupid questions16:59
lisitsyn1HeikoS: haha no that's good question because API is stupid17:00
lisitsyn1HeikoS: you want to regularize groups of features right?17:00
@HeikoSyes17:01
@HeikoSLet me explain the problem:17:01
lisitsyn1like L1 of (L2 of (1,2,3)) + (L2 of (4,5,6))17:01
@HeikoSmaybe you have an idea17:01
@HeikoSI have a dataset17:01
lisitsyn1HeikoS: ah wait17:01
@HeikoSit consists of frequency measurements of the brain - with a few hundred sensors17:01
lisitsyn1do you need multiclass17:02
@HeikoSyes17:02
lisitsyn1or binary?17:02
lisitsyn1alright17:02
@HeikoSthe sensors are place on the head of a guy17:02
lisitsyn1HeikoS: unfortunately I'd have to ask you to try and let me know if it produces something unreasonable17:02
@HeikoShaha ok17:02
lisitsyn1ok lets continue17:02
lisitsyn1so each feature is17:03
-!- lisitsyn1 is now known as lisitsyn17:03
lisitsynampl of some frequency17:03
lisitsyn?17:03
@HeikoSyes, its preprocessed I have  the FFt coefficients17:03
@HeikoSI want to group sensors17:03
lisitsyndo you have groups of sensors?17:04
lisitsynor do you want to group them?17:04
@HeikoSmakes sense to group them17:04
@HeikoSthe resulting weights are probably sparse17:04
@HeikoSbut the sparsity structure depends on the sensor positions17:04
@HeikoSso I want to try to learn this17:04
lisitsynhmm can you group sensors into some hierarchy?17:05
lisitsynHeikoS: it is important to have these groups already17:06
lisitsynit can't learn the groups itself17:06
@HeikoSi know thats fine17:06
lisitsynHeikoS: so the simplest way to go is17:07
lisitsynlet me check17:07
lisitsynHeikoS: root is all features17:07
lisitsynHeikoS: tree = IndexBlockTree(IndexBlock(0,n_features))17:08
lisitsynthen if you want no L2 of all features do tree.get_root_block.set_weight(0.0)17:08
lisitsynand then17:09
@HeikoSwhats the difference between IndexBlockGroup and tree?17:09
lisitsynHeikoS: just how they treat the IndexBlock17:09
lisitsynHeikoS: I mean Tree uses subblocks as childs17:09
lisitsynGroup is just a set of blocks17:09
@HeikoSso if I have indices17:10
@HeikoS[0,1,2,3,4,5,6,7,8,9]17:10
@HeikoSand I want groups17:10
@HeikoS[0,1,2], [3,4,5], [6,7,8,9]17:10
lisitsynthe code will be17:10
lisitsyntree = IndexBlockTree(IndexBlock(0,10))17:10
lisitsyntree.get_root_block().add_subblock(IndexBlock(0,3))17:11
lisitsyntree.get_root_block().add_subblock(IndexBlock(3,6))17:11
lisitsyntree.get_root_block().add_subblock(IndexBlock(6,10))17:11
lisitsynthat's tree17:11
lisitsynroot        [0,1,2,3,4,5,6,7,8,9]17:11
lisitsynchildren [0,1,2], [3,4,5], [6,7,8,9]17:11
@HeikoSand I want the regulariser to be \lambda \sum ||groups||_217:11
lisitsynHeikoS: I didn't implement pure group lasso but tree lasso17:12
lisitsynso you have to put weights here17:12
@HeikoSok, can I simulate this?17:12
@HeikoSokay17:12
lisitsynyes17:12
@HeikoShow? :)17:12
lisitsynjust set the weight of root to 017:12
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Ping timeout: 250 seconds]17:12
lisitsynand you get L1/L217:12
lisitsynotherwise you get L1/L2 of groups17:12
lisitsynplus L2 of all features17:13
@HeikoSok17:13
@HeikoStree.get_root_block.set_weight(0.0)17:13
@HeikoSnice17:13
-!- mode/#shogun [+o sonney2k] by ChanServ17:14
@HeikoSbtw, this should really be documented somewhere (sorry to be annoying :)17:14
lisitsynHeikoS: really ugly API17:14
@HeikoSwhy?17:14
lisitsynI want to do something with it but don't know what17:14
@HeikoSits fine17:14
@HeikoSjust one needs to share the knowledge on how to use it17:14
lisitsynHeikoS: about its working capabilities17:15
lisitsynI *know* binary thing works17:15
lisitsyncan't say that about multiclass tree thing though17:15
lisitsynbut I will be able to find a bug rather quick17:15
@HeikoSif binary works and it uses the multoclass framework than it should be fine right?17:15
@HeikoSbut there are also multinomial formulations of this17:16
@HeikoSanyway17:16
lisitsynHeikoS: no that's different solvers17:16
@HeikoSI am glad there is something17:16
@HeikoSreally?17:16
@HeikoSI see17:16
@HeikoSdid you ever test it?17:16
@HeikoSlike on a toy multiclass problem17:16
lisitsynHeikoS: yes I tried to use it on road signs data17:16
@sonney2klisitsyn, HeikoS - that you are both here. When do we put up a workshop schedule?17:16
@sonney2kI mean finalize the schedule!17:16
lisitsynwhere HoG cells are groups17:16
@sonney2klisitsyn, did it improve anything?17:17
lisitsynHeikoS: it produced reasonable results but worse than SVM17:17
lisitsynsonney2k: no became worse17:17
@sonney2klisitsyn, yeah as expected :D17:17
@sonney2kMKL is just crap17:17
lisitsynsonney2k: too many parameters (each group has weight)17:17
@sonney2kcomplex crap17:18
lisitsynhaha yes17:18
@sonney2kHeikoS, lisitsyn workshop!17:18
lisitsyneverything complex sucks it seems17:18
@sonney2k1 month17:18
lisitsynsonney2k: yes yes17:18
lisitsynwhat's about it?17:18
@sonney2kwe need to fix the schedule17:18
@HeikoSsonney2k:  is there a list of people that applied?17:18
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun17:18
@sonney2kas in when & where17:18
@sonney2kHeikoS, did you apply?17:18
-!- mode/#shogun [+o iglesiasg] by ChanServ17:18
@HeikoSsonney2k:  no why should I?17:18
@sonney2kHeikoS, then do it17:19
-!- gsomix [~Miranda@85.26.235.51] has joined #shogun17:19
@sonney2kHeikoS, how do we know who will come and what topics we want to see presented/discussed?17:19
@sonney2klisitsyn, did you?17:19
lisitsynsonney2k: applying17:20
lisitsyndone17:21
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun17:22
@sonney2kanyway we need to come up with a fixed workshop program17:22
@HeikoSsonney2k: is there a list?17:22
lisitsynHeikoS: list of what?17:22
@sonney2klisitsyn, can you give HeikoS access to the list please17:22
@sonney2kworkshop document17:23
@HeikoSfound it17:23
@HeikoSdidnt know there was this google doc17:23
@HeikoS20 people17:23
lisitsynHeikoS: shared with you :)17:23
lisitsynI'd say 1717:23
lisitsynha17:23
lisitsynthoralf: one more :)17:24
thoralf:)17:24
@sonney2kthoralf, will you come btw?17:24
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun17:24
shogun-notifier-shogun: Thoralf Klein :develop * 4146e2b / src/shogun/mathematics/Math.h,tests/unit/mathematics/Math_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/4146e2bb6159f01b743c4c6e226714fdda12331f17:24
shogun-notifier-shogun: Added test for sorting zero-sized vectors:17:24
shogun-notifier-shogun:17:24
shogun-notifier-shogun: - CMath::qsort(* T, ...)17:24
shogun-notifier-shogun: - CMath::qsort(** T, ...)17:24
shogun-notifier-shogun: - CMath::qsort_index(...)17:24
shogun-notifier-shogun: Soeren Sonnenburg :develop * 7e999bd / src/shogun/mathematics/Math.h,tests/unit/mathematics/Math_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/7e999bdd3d9c74866e25e3fa628f70fba8b6118517:24
shogun-notifier-shogun: Merge pull request #1145 from tklein23/develop17:24
shogun-notifier-shogun:17:24
@sonney2kiglesiasg, registered and cheng will also be there17:24
shogun-notifier-shogun: Unit tests for CMath::qsort(): Trying to sort empty lists with qsort(*T) and qsort(**T)17:24
lisitsynsonney2k: cheng is already here17:25
thoralfsonney2k: I just registered.17:25
@HeikoSsonney2k: I suggest we count votes on topics and take the n most voted ideas17:26
@HeikoS    tree.get_root_block().add_subblock(IndexBlock(0,3))17:27
@HeikoSAttributeError: 'IndexBlock' object has no attribute 'add_subblock'17:27
@HeikoSlisitsyn:  ^17:28
lisitsynHeikoS: sub_block may be?17:28
@HeikoSyep17:29
-!- vgorbati [~vgorbati@212.2.159.34] has quit [Quit: vgorbati]17:29
lisitsynsonney2k: so what should we fix?17:30
@sonney2klisitsyn, the schedule!17:30
@HeikoSsonney2k: have we decided on which topics we present?17:30
@sonney2kas in who gives a talk about what, what discussions, hands ons thats17:30
@sonney2kHeikoS, nope17:30
@sonney2kwe have to do that17:30
@HeikoSsonney2k, lisitsyn so my suggestion on this, just count17:30
@HeikoSfind volunteers17:31
@HeikoSand then assign them the topics17:31
lisitsynyeah lets do that this way17:31
@HeikoSI can do that17:31
@HeikoSdo we have a list of people that want to talk about things?17:31
lisitsynHeikoS: do you want?17:31
@HeikoSlisitsyn: maybe we could put names, emails, and topics somewhere17:32
@sonney2koh HeikoS and please don't forget the e.V. thing. the WS is just one month away17:32
@HeikoSsonney2k: yes, havent forgot it17:32
@sonney2kI don't really want to talk but hey I guess I have to17:32
@sonney2k1 talk only though :)17:32
@HeikoSI guess so :D17:32
shogun-buildbotbuild #907 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/907  blamelist: Thoralf Klein <thoralf.klein@zib.de>17:33
@sonney2ksry gtg, can we talk about this further in the evening?17:33
@sonney2kthoralf, you are famous -17:33
@HeikoSsonney2k, lisitsyn so noone has in fact submitted any abstracts17:33
* sonney2k argh if I only knew why cygwin has problems ...17:33
@HeikoSI guess this means we only have our own talks17:34
@sonney2kHeikoS, yes no researchers that need a plane / train ticket17:34
@sonney2kin the evening then... cu17:34
@sonney2kafk17:34
-!- shogun-notifier- [~irker@7nn.de] has quit [Ping timeout: 256 seconds]17:38
-!- gsomix [~Miranda@85.26.235.51] has quit [Ping timeout: 256 seconds]17:41
-!- shogun-toolbox [~shogun@7nn.de] has quit [Ping timeout: 256 seconds]17:41
--- Log closed Wed Jun 05 17:41:01 2013
--- Log opened Wed Jun 05 17:41:12 2013
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun17:41
-!- Irssi: #shogun: Total of 15 nicks [3 ops, 0 halfops, 0 voices, 12 normal]17:41
shogun-buildbotbuild #1075 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1075  blamelist: Thoralf Klein <thoralf.klein@zib.de>17:41
-!- gsomix_ [~Miranda@85.26.235.51] has joined #shogun17:41
-!- Irssi: Join to #shogun was synced in 6 secs17:41
shogun-buildbotbuild #1076 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/107617:47
-!- gsomix_ [~Miranda@85.26.235.51] has quit [Ping timeout: 264 seconds]17:52
-!- sonney2k [~shogun@7nn.de] has quit [Ping timeout: 264 seconds]17:52
-!- sonney2k [~shogun@7nn.de] has joined #shogun17:53
-!- gsomix [~Miranda@37.61.180.242] has joined #shogun17:55
-!- nube1 [~rho@36.253.6.136] has joined #shogun17:56
-!- Netsplit *.net <-> *.split quits: lambday, nube, van5117:57
-!- Netsplit over, joins: van5118:05
-!- mode/#shogun [+o sonney2k] by ChanServ18:05
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Quit: Page closed]18:08
-!- nube1 [~rho@36.253.6.136] has quit [Ping timeout: 256 seconds]18:08
-!- nube [~rho@36.253.6.136] has joined #shogun18:08
-!- nube [~rho@36.253.6.136] has quit [Ping timeout: 256 seconds]18:13
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has joined #shogun18:13
-!- nube [~rho@49.244.97.103] has joined #shogun18:22
lambdayHeikoS: please take a look at the PR when you're free.18:59
lambdayI'm taking off for the day..18:59
lambdayciao18:59
-!- lambday [67157f4e@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.78] has quit [Quit: lambday]19:00
-!- Netsplit *.net <-> *.split quits: @sonney2k, nube, shogun-buildbot, flxb_, van5119:05
-!- vgorbati [~vgorbati@212.2.159.34] has joined #shogun19:05
-!- Netsplit over, joins: nube, van51, @sonney2k, shogun-buildbot, flxb_19:08
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 246 seconds]19:11
-!- lisitsyn [~blackburn@37.61.180.242] has joined #shogun19:13
@sonney2klisitsyn, HeikoS so what should be the topics?19:25
@sonney2kHeikoS, it looks like lots of people want MMD/HSIC :)19:26
@HeikoSsonney2k: can do :)19:26
@HeikoSI will do the vote list tomorrow19:26
@sonney2kvan51, around?19:34
van51sonney2k: yeap19:34
@sonney2kvan51, could you make sense of what I wrote yesterday about dotfeatures?19:35
van51sonney2k: yea19:35
van51sonney2k: by reading it on the logs19:35
@sonney2kand how to do the CHashedDocFeatures / CHashedDocPreprocessor19:35
@sonney2k(Doc is not the best word but no idea what else to use)19:36
@sonney2kvan51, I think this is no more than 1 day of work for you BTW19:36
van51sonney2k: question, the preprocessor can be used offline?19:36
@sonney2kvan51, offline as in?19:36
@sonney2knot sure what you mean19:36
van51I mean use it with a file as input and a file as output19:36
votjakovrHeikoS: I've found that difference between a and b in compare function in tester.py approx. 10e-12 on test that gives empty diff19:37
van51and then read that file through a load method19:37
van51or something19:37
@sonney2kvan51, ohh actually preprocessor won't work - you need a converter19:37
@HeikoSvotjakovr: say again, I did not get that19:37
@sonney2kvan51, ok one step back. which interface do you intend to use?19:37
@sonney2kpure C++?19:37
van51sonney2k: yea19:38
@sonney2kor e.g. python_modular19:38
@sonney2kvan51, ok then well you can create your stringcharfeatures19:38
@sonney2kload some strings from disk19:38
@sonney2kand then apply a converter to convert them to DenseFeatures19:38
@sonney2ksince it is all about wordcounts you can use uint16_t or int32_t19:39
@sonney2kfor dense features19:39
@sonney2kso you would convert StringFeatures<char> -> CDenseFeatures<int32_t> in your converter19:40
@sonney2kand then of course you can print the whole matrix19:40
van51sonney2k: ah yea19:40
votjakovrHeikoS: I've changed implementation of exact inference method a little bit. Python integration test gives an error and empty diff. I've found that that the difference between a and b in compare function in tester.py approx. 10e-1219:40
van51sonney2k: this sounds ok to me19:40
@sonney2khmmhh maybe CSparseFeatures<int32_t> is better considering that you use hashes :)19:40
van51sonney2k: I am not sure that the vector will be sparse19:41
@sonney2kvan51, it will be: look19:41
@sonney2khash(word1) -> some large int19:42
@sonney2kand you do this for all words in the docuemnt19:42
@sonney2kso depending on the number of bits you use for your hash19:42
van51sonney2k: yes but from what I understand it's for dimensionality reduction and afterwards you mod that number to a specific range 0...d19:42
@sonney2ksay 2019:42
@sonney2kyou have a 1048576 dimensional vector you create of a text document19:43
@sonney2kvan51, well if you specify number of bits to be say 819:43
@sonney2kthen it will be dense19:43
@sonney2kbut we the more bits we can afford the better the results19:44
@sonney2kso it is usually sparse19:44
@sonney2kIIRC spam filters use sth like 32bit19:44
van51sonney2k: ok, got it19:44
votjakovrHeikoS: May be it's better to regenerate test data?19:44
@sonney2kvan51, anyway you should really get started with that19:44
van51sonney2k: I will today19:45
van51sonney2k: btw since we're at it19:45
van51sonney2k: how do I specify the number of bits in the hashes from CHash?19:46
van51hash functions*19:46
@sonney2kvan51, well just use bitwise &19:48
@sonney2kas in x & ((1<<8)-1)19:48
@sonney2kwould give you an 8 bit hash19:48
@sonney2kand btw use murmurhash319:49
van51sonney2k: ok I will19:50
van51sonney2k: thanks, you've been really helpful :)19:50
-!- nube [~rho@49.244.97.103] has quit [Quit: Leaving.]19:57
@sonney2kvan51, you just have to do this once. afterwards you will realize how simple this is19:59
votjakovrHeikoS: i've just sent a PR, please have a look at it.20:02
-!- vgorbati [~vgorbati@212.2.159.34] has quit [Quit: vgorbati]20:08
@sonney2kvan51, I've found the reason for the crash with the feature iterator20:10
@sonney2kvan51, bad news is that it is not easy to fix20:11
van51sonney2k: what is it?20:11
@sonney2kit is indeed the conversion to SGSparseVector and reference counting20:11
@sonney2kthe correct patch will be very intrusive20:11
@sonney2kno idea what it will break besides that but I suspect serialization and lots of other stuff20:12
@sonney2kso I will need HeikoS help then too20:12
@sonney2kvan51, ok let me try...20:12
van51sonney2k: good luck :)20:12
-!- vgorbati [~vgorbati@212.2.159.34] has joined #shogun20:17
van51sonney2k: it's in the get_sparse_feature_vector() method?20:18
@HeikoSvotjakovr:  looking...20:24
@sonney2kvan51, no it is the whole CSparseFeatureMatrix class20:25
votjakovrHeikoS: thanks :)20:25
@sonney2kvan51, it needs to be converted to internally use SGSparseMatrix20:25
@HeikoSsonney2k: reference counting for sparse vectors?20:25
@sonney2kvan51, otherwise refcounts are kaboom20:25
@sonney2kHeikoS, we have that already20:25
@sonney2kHeikoS, but SGSparseMatrix is not yet used and I will need help with serialization support for that20:26
@HeikoSsonney2k: I see20:26
@HeikoSwho wnats to have that?20:26
@sonney2kHeikoS, sorry what?20:28
@HeikoSsonney2k: how come you are working on this?20:28
@sonney2kHeikoS, CSparseFeatureMatrix is half-way converted to use our refcounted stuff20:28
@HeikoSI see20:28
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun20:28
@sonney2kwe have crashers now due to it not being done completely20:29
@sonney2kso I am doing it now (or better trying to do it)20:29
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun20:29
shogun-notifier-shogun: Roman Votyakov :develop * cd27d05 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/cd27d057b1c1a7cd2e5a22741c9b7be1e1d8e89620:29
shogun-notifier-shogun: Redesigned GP framework20:29
shogun-notifier-shogun: Heiko Strathmann :develop * 53f4ea1 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/53f4ea1ff26387eb1d85988c3f547c7e8f94676a20:29
shogun-notifier-shogun: Merge pull request #1150 from votjakovr/develop20:29
shogun-notifier-shogun:20:29
shogun-notifier-shogun: Redesigned GP framework20:29
@HeikoSsonney2k: I see, sounds like a nasty one20:29
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has joined #shogun20:29
@HeikoSat some point we have to fix all these parameter issues and bugs20:30
@HeikoSalso streaming ....20:30
@HeikoSbut so much work :(20:30
@HeikoSvotjakovr: what is the next step?20:31
@HeikoSvotjakovr: pulling GP class out?20:31
@sonney2kvotjakovr, woaaah massive patch!20:31
votjakovrHeikoS: I'd like to remove check_members methods20:31
votjakovrHeikoS: Also solve the problem with recomputing cholesky20:32
@HeikoSvotjakovr: go ahead ! I agree20:32
@HeikoSyep also on that20:32
@HeikoSsonney2k:  votjakovr is like a steam train :)20:32
votjakovrHeikoS: I find that update_parameter_hash() will be very helpful with that20:33
votjakovrhehe :D20:34
@HeikoSvotjakovr: yep :)20:34
@sonney2kHeikoS, hehe yeah I would say the same about lambday ;)20:34
@sonney2kHeikoS, and you and lisitsyn btw :)20:34
@HeikoSyeah lambday rocks totally20:34
@sonney2kmaybe even more like a start destroyer 'we break for nobody'20:35
@sonney2k:D20:35
lisitsynRE!20:35
@HeikoShaha :)20:35
@sonney2klisitsyn, I finally understand why you didn't even try to convert sparse featurs to use SGSparseMatrix20:35
* sonney2k suffers20:35
lisitsynsonney2k: why?20:36
lisitsynbecause it is mega pain in the ass you mean?20:36
lisitsynlike megaupload20:37
shogun-buildbotbuild #909 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/909  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>20:38
@HeikoSwhaaaaaat?20:38
votjakovrhmmm20:38
@HeikoSvotjakovr: its cyg, dont worry :)20:38
@sonney2kHeikoS, btw you should ask Olivier if he wants to come to the workshop :)20:39
@HeikoSsonney2k: will do20:40
@HeikoShe seemed ultra busy though20:40
@HeikoSas everyone always seems20:40
@HeikoSweird isnt it?20:40
lisitsynwhat is the magical moment it comes better with our busy deals20:41
lisitsynsonney2k: everything is getting worse with years, right?20:41
lisitsyn:D20:41
@HeikoSwhen my evil-nano-robots conquered the planet20:42
@sonney2klisitsyn, yes it is now that you have lots of time20:42
lisitsynsonney2k: haha ok!20:42
@sonney2konce you become as old as me you would wish you would have spent your lots and lots of time you had more wisely :D20:43
lisitsynsonney2k: tell me what I should spend it on then!20:43
@sonney2kdoing cool stuff20:43
lisitsynsonney2k: haha!20:43
@sonney2klisitsyn, so of course working on shogun :D20:43
@sonney2kor HeikoS nano robots20:44
@sonney2kbut powered by shogun20:44
@HeikoSof course they are20:44
@sonney2khmmhh20:44
lisitsynsonney2k: let see what you say when you turn 5020:44
@HeikoSthey are in fact powered on unit-tests20:44
@sonney2klisitsyn, hey I am not there yet20:44
lisitsynI will turn like 3620:44
lisitsyn:D20:44
@sonney2kI first have to survive this decade20:45
lisitsynsonney2k: oh I'll turn your current year when you turn 5020:45
@sonney2kwhat?20:45
lisitsynsonney2k: when you are 50 I am 3620:45
@sonney2kHeikoS, lisitsyn does anyone use get_feature from our features? and the on-the-fly preprocessors?20:45
@HeikoSI dont know20:46
lisitsynsonney2k: I never used on-the-fly thing20:46
@sonney2kI used to use them but haven't for a long time20:46
lisitsynwell it may be a good thing but20:46
lisitsynsomething is wrong20:46
@sonney2kit was working when I wrote it - I made sure20:46
@sonney2kbut now with all that refactoring and no tests20:47
@sonney2kno idea20:47
lisitsynsonney2k: no it works I think20:47
lisitsynI mean something is wrong with its api and etc20:47
lisitsynit is not attractive to use20:47
lisitsynsonney2k: see a big shot in our list20:47
@sonney2kwhat woudl you do better?20:47
@sonney2kit is very easy to do20:47
lisitsynsonney2k: if I had idea I would do that or say :)20:47
@sonney2kjust attach a list of preprocessors20:47
@sonney2kdone20:47
lisitsynI usually say I don't like X because I don't know how to improve it20:48
@sonney2kno idea how it could be even easier20:48
@sonney2kI am a bit worried about the Cache in all this20:48
lisitsynsonney2k: I'd like to drop some code20:48
lisitsyn:D20:48
@sonney2klisitsyn, do you know some LRU cache?20:49
lisitsynsonney2k: basics - never implemented20:49
lisitsynyou mean implementation?20:49
@sonney2kI would rather use some well written thing and drop ours20:49
shogun-notifier-shogun: lambday :develop * 23d134a / / (10 files): https://github.com/shogun-toolbox/shogun/commit/23d134aa08d65d5d8df380c46c4140af9ea7639820:49
shogun-notifier-shogun: error msgs fixed for complex64_t, unit-tests added20:49
shogun-notifier-shogun: Heiko Strathmann :develop * 6a47272 / / (10 files): https://github.com/shogun-toolbox/shogun/commit/6a47272a5d5982979faddf98d551eef60ee04c6020:49
shogun-notifier-shogun: Merge pull request #1149 from lambday/develop20:49
shogun-notifier-shogun:20:49
shogun-notifier-shogun: error msgs fixed for complex64_t, unit-tests added20:49
@sonney2kyeah something we could use20:49
lisitsynlets google20:49
@sonney2kHeikoS, you do that on purpose do you? Just when I branched you mega merge20:49
@HeikoSehm?20:50
@HeikoSwhat?20:50
lisitsynsonney2k: a lot of implementations20:50
lisitsynsonney2k: what is it for?20:50
@sonney2klisitsyn, our on-the-fly preprocessors use a cache to cache feature vectors20:51
@sonney2kpreprocessed ones I mean20:51
lisitsynsounds complex!20:51
@sonney2kI know you never used it but our DenseFeatures don't have to be in memory20:51
lisitsynhaha yes I know20:52
lisitsynbut never tried20:52
@HeikoSI gotta go now20:52
@HeikoSsee you guys tomorrow!20:52
-!- mode/#shogun [+o lisitsyn] by ChanServ20:52
-!- HeikoS [~heiko@nat-173-83.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]20:52
@sonney2kto late to say bye20:52
shogun-buildbotbuild #1078 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1078  blamelist: Roman Votyakov <votjakovr@gmail.com>20:52
shogun-buildbotbuild #910 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/910  blamelist: Roman Votyakov <votjakovr@gmail.com>20:53
@sonney2klisitsyn, could you please change this to 1e-14 http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1078/steps/test/logs/stdio20:53
@sonney2klooks like fabs is b0rken in old libc's20:54
@lisitsyncrazy precision20:54
@sonney2klisitsyn, why?20:54
@lisitsynsonney2k: I am scaried what these guys are going to compute20:55
@lisitsynintergalactic spaceships20:55
@lisitsynor what20:55
@sonney2klisitsyn, I was already scared when I read the proposal20:55
@lisitsynsonney2k: I believe they are going to compute 4d space warping to fly away from this planet20:55
@sonney2k*yay* it is becoming kind of summer here :D20:55
@lisitsynsonney2k: bayessian theoretical aspects of teleportation20:56
@lisitsynsonney2k: we should ask if nasa uses such a precision20:58
@lisitsynthey must be e-3 most of the times20:59
@lisitsyn:D20:59
@sonney2klisitsyn, only our SVMs are 1e-321:00
@lisitsynsonney2k: what about the others?21:00
@lisitsynoops I am into merging issue21:04
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving]21:11
shogun-notifier-shogun: Sergey Lisitsyn :develop * e496115 / tests/unit/lib/SGVector_unittest.cc,tests/unit/mathematics/Complex_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/e4961157662ed92cac9139fd4eca1913a454877121:17
shogun-notifier-shogun: Lowered precision of complex unit-tests21:17
@lisitsynfinally21:17
-!- vgorbati [~vgorbati@212.2.159.34] has quit [Quit: vgorbati]21:20
-!- shogun-toolbox [~shogun@7nn.de] has quit [Ping timeout: 256 seconds]21:21
--- Log closed Wed Jun 05 21:21:27 2013
--- Log opened Wed Jun 05 21:21:36 2013
-!- shogun-t1olbox [~shogun@7nn.de] has joined #shogun21:21
-!- Irssi: #shogun: Total of 15 nicks [2 ops, 0 halfops, 0 voices, 13 normal]21:21
-!- Irssi: Join to #shogun was synced in 9 secs21:21
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has left #shogun ["QUIT :Leaving."]21:23
-!- shogun-toolbox [~shogun@7nn.de] has quit [Ping timeout: 256 seconds]21:28
@sonney2kargh out of battery21:29
shogun-notifier-shogun: van51 :feature/model_selection_syntax * b1b8570 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/b1b85707654bed214973ec65aa8c9742143a31c021:31
shogun-notifier-shogun: Changed CList to CDynamicObjectArray in CCombinedDotFeatures (WIP)21:31
shogun-notifier-shogun: lambday :feature/model_selection_syntax * d453390 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/d453390e7ecd231fea02d4e8259a9279ea30ca7d21:31
shogun-notifier-shogun: CMath::abs(complex64_t) changed to not to use std::abs()21:31
shogun-notifier-shogun: Thoralf Klein :feature/model_selection_syntax * a361752 / src/shogun/mathematics/Math.h,tests/unit/mathematics/Math_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/a361752210386dec5027ffc84427276b39895b8d21:31
shogun-notifier-shogun: Added test for sorting zero-sized vectors:21:31
shogun-notifier-shogun:21:31
shogun-notifier-shogun: - CMath::qsort(* T, ...)21:31
shogun-notifier-shogun: - CMath::qsort(** T, ...)21:31
shogun-notifier-shogun: - CMath::qsort_index(...)21:31
shogun-notifier-shogun: Roman Votyakov :feature/model_selection_syntax * 5b94f09 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/5b94f09a19731d6dc13160169efdddbf10897df121:31
shogun-notifier-shogun: Redesigned GP framework21:31
shogun-notifier-shogun: Sergey Lisitsyn :feature/model_selection_syntax * f9cdccf / / (4 files): https://github.com/shogun-toolbox/shogun/commit/f9cdccf2cf48cf4168437cbc3741288a209a272721:31
shogun-notifier-shogun: A prototype of builder-like modelselection syntax21:31
@lisitsynsonney2k: these guys are tiresome21:31
@lisitsynehmm I get e+180 difference21:31
@lisitsynin FITCInferenceMethod21:31
@lisitsynvgorbati: hey how are ya doin?21:31
@lisitsynsonney2k: here goes a new branch21:32
-!- iglesiasg [d58f3220@gateway/web/freenode/ip.213.143.50.32] has joined #shogun21:34
-!- mode/#shogun [+o iglesiasg] by ChanServ21:34
@lisitsynoh shit21:35
shogun-buildbotbuild #911 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/911  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>21:42
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 245 seconds]21:47
shogun-buildbotbuild #1079 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/107921:49
shogun-notifier-shogun: Sergey Lisitsyn :feature/model_selection_syntax * 8ac4ed7 / src/shogun/modelselection/ModelSelectionParameters.h: https://github.com/shogun-toolbox/shogun/commit/8ac4ed73c9edf590a2b8954b5bda49e49fbfe37421:58
shogun-notifier-shogun: Some documentation for model selection range helper21:58
-!- Netsplit *.net <-> *.split quits: @lisitsyn, hushell, flxb_, sonne|work, gsomix, shogun-notifier-, shogun-buildbot, @sonney2k, naywhayare, wiking, (+2 more, use /NETSPLIT to show all of them)22:02
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun22:28
-!- Netsplit over, joins: @iglesiasg, hushell, shogun-notifier-, @lisitsyn, flxb_, shogun-buildbot, @sonney2k, gsomix, thoralf, naywhayare (+2 more)22:28
-!- shogun-t1olbox [~shogun@7nn.de] has quit [Ping timeout: 240 seconds]22:39
--- Log closed Wed Jun 05 22:39:25 2013
--- Log opened Wed Jun 05 22:39:33 2013
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun22:39
-!- Irssi: #shogun: Total of 16 nicks [4 ops, 0 halfops, 0 voices, 12 normal]22:39
-!- Irssi: Join to #shogun was synced in 6 secs22:39
-!- shogun-t1olbox [~shogun@7nn.de] has quit [Ping timeout: 240 seconds]22:41
-!- HeikoS [~heiko@nat-173-83.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]22:50
-!- iglesiasg [d58f3220@gateway/web/freenode/ip.213.143.50.32] has quit [Quit: Page closed]22:55
--- Log closed Thu Jun 06 00:00:33 2013