Open in new window / Try shogun cloud
--- Log opened Mon Oct 31 00:00:55 2016
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun05:17
-!- HeikoS [~heiko@104.220.67.105] has joined #shogun07:09
-!- mode/#shogun [+o HeikoS] by ChanServ07:09
-!- HeikoS [~heiko@104.220.67.105] has quit [Ping timeout: 268 seconds]08:05
-!- HeikoS [~heiko@104.220.67.105] has joined #shogun08:06
-!- mode/#shogun [+o HeikoS] by ChanServ08:06
-!- HeikoS [~heiko@104.220.67.105] has quit [Ping timeout: 244 seconds]08:12
-!- praisethemoon [~praisethe@41.226.248.123] has joined #shogun08:14
-!- travis-ci [~travis-ci@ec2-54-162-125-255.compute-1.amazonaws.com] has joined #shogun08:21
travis-ciit's Saurabh7's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/17190362108:21
-!- travis-ci [~travis-ci@ec2-54-162-125-255.compute-1.amazonaws.com] has left #shogun []08:21
-!- c4goldsw [8a33ffc3@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.255.195] has joined #shogun12:10
c4goldswping wiking12:10
-!- c4goldsw [8a33ffc3@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.255.195] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]12:21
-!- c4goldsw [8a33f99f@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.159] has joined #shogun12:53
-!- c4goldsw [8a33f99f@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.159] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]13:41
-!- c4goldsw [8a33f99f@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.159] has joined #shogun13:48
-!- c4goldsw [8a33f99f@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.159] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]14:00
-!- c4goldsw [8a33f005@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.240.5] has joined #shogun15:26
-!- praisethemoon [~praisethe@41.226.248.123] has quit [Ping timeout: 268 seconds]18:07
-!- c4goldsw [8a33f005@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.240.5] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]18:14
-!- c4goldsw [8a33f496@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.244.150] has joined #shogun18:48
CaBawiking: ping19:06
CaBais there any example on how to do model selection in case of MKL?19:07
@wikingCaBa, hey19:10
@wikingmmm19:10
@wikinghave you checked the notebooks?19:10
@wikingc4goldsw, hi man19:11
@wikingsorry we've been busy with the release19:11
@wikingand gsoc summit19:11
@wikingbut i'm gonna get to your PR asap19:11
CaBajust staring at this right now: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/CombinedKernel.cpp#L80519:11
-!- ajatgd [8ba74f02@gateway/web/freenode/ip.139.167.79.2] has joined #shogun19:11
CaBawiking: where can i search the notebooks?19:12
-!- ajatgd [8ba74f02@gateway/web/freenode/ip.139.167.79.2] has quit [Client Quit]19:12
CaBawiking: i don't really get how i can add the kernels with their own parameters to be learned19:13
-!- HeikoS [~heiko@104.220.67.105] has joined #shogun19:13
-!- mode/#shogun [+o HeikoS] by ChanServ19:13
CaBahey HeikoS19:13
@HeikoSCaBa: hihi19:13
@HeikoSall good?19:13
CaBaHeikoS: trying to figure aout MKL + MS19:13
CaBaHeikoS: no examples out there, are they?19:13
@HeikoSI am not sure19:13
@HeikoSbut what do you want to model select?19:14
@HeikoSonly thing you need is the regularisation parameter, no?19:14
CaBaHeikoS: the parameters of the individual kernesls19:14
@HeikoSthe MKL learns the kernel weights for you19:14
@HeikoSthe coefficients19:14
CaBaHeikoS: yes but how about the parameters of the used kernels?19:14
@HeikoSusually, one would put in multiple kernels with different parameters19:15
@HeikoSbut yeah you should also be able to model select them19:15
@HeikoSjust do the usual story of building a parameter tree19:15
@HeikoSit is a bit insane computationally though ;)19:15
@HeikoSthere is the ipython notebook on model selection19:16
CaBais there mkl in there?19:16
@HeikoSyou can open manually, but we are in fact preparing a new release atm and it will be on the website soon19:16
@HeikoSas in today/tomorrow19:16
@HeikoSno mkl in there19:16
CaBawell MS i've alreay been doing19:16
@HeikoSI see19:16
@HeikoSdoesnt it work with mkl?19:16
@HeikoSI can have a look19:17
@HeikoS(after we released)19:17
CaBai don't see how i could add that, maybe i don't get the idea19:17
@HeikoSyou build the parameter tree19:17
@HeikoSput in all the things you want19:17
@HeikoSand then as usual19:17
@HeikoSmkl might or might not work with that19:17
@HeikoSbut it is easy to make it work if it doesnt and I can help19:17
@HeikoSbut building the parameter tree should be straight forward19:17
CaBait isn't to me right now19:18
CaBathat's my problem19:18
@HeikoSok19:19
@HeikoSI can write something19:19
@HeikoStomorrow ok?19:19
@HeikoSI might do it in the plane19:19
CaBaany brief example on how to plug this together would be great19:19
@HeikoShave a 10 hr flight later19:19
CaBayeah thanks19:19
@HeikoScool will do19:19
@HeikoSCaBa: ping me if you dont hear anything19:19
CaBawill do19:19
CaBa;)19:19
CaBahave a safe trip19:20
@HeikoSthanks :)19:21
@HeikoSCaBa: hey do you want to contribute a cookbook page for modelselection?19:21
@HeikoSCaBa: in the style of this: http://shogun.ml/cookbook/latest/19:21
CaBaHeikoS: sure, if i get to figure it out :))19:22
CaBaHeikoS: is running multiple kernels on the same data the usual usecase for MKL?19:24
CaBaHeikoS: because that's not what I'm doing... every kernel works on it's own data in my case19:24
@HeikoSyeah it is fine19:24
@HeikoSbut depends on what you want to do19:25
@HeikoSwe have a noteobok on mkl, you saw that?19:25
CaBayep, just went through that19:25
CaBathe old school examples are a bit more exhaustive here i think ;)19:26
lisitsynhey guys19:26
CaBahey lisitsyn19:26
lisitsynHeikoS: you back?19:26
lisitsynthanks for the plan19:26
lisitsynglanced over it19:26
CaBabbl, going home19:26
@wikinglisitsyn,19:28
lisitsynwiking: hey19:29
@wikingwhen are you coming back to shogun? :D19:29
lisitsynis it an invitation?19:29
lisitsyn:P19:29
lisitsynI guess nips19:29
lisitsynnips anybody?19:29
@wiking:(19:30
@wikingnot this year again :(19:30
lisitsynmeh19:30
lisitsynyour employer19:31
lisitsyn:P19:31
@wiking:D19:31
@wikingso btw19:31
@wikingdo you have somewhere that SG_ADD patch19:31
@wikingthat is not yet 100%19:31
@wikingbut still something19:31
lisitsynuhmm yes19:31
lisitsynwant to finish it?19:31
@wikingyeah woudl be good to have it out there19:31
@wikingso either we finish it19:31
@wikingor somebody19:31
lisitsynit is terrible19:31
lisitsynI mean I got into it19:31
lisitsynand then I realize I'm in freaking narnia19:32
@wiking;D19:32
lisitsynall is fragile and19:32
lisitsynno idea how to unfuck it19:32
@wikingheheh no worries19:32
@wikingwe'll see19:32
@wikingjust have it somewhere public19:32
@wikingso we can play with it19:32
lisitsynwiking: https://github.com/shogun-toolbox/shogun/commit/b7823b95315068cda6fba24b41e11f46887064fd19:33
lisitsynthis branch19:33
@wiking\o/19:33
lisitsynwiking: you will get the basic idea19:34
lisitsynI cna describe what is fucked19:34
lisitsynthe thing is I had to patch any19:34
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Quit: Leaving]19:34
lisitsynto support policy different from default19:34
lisitsyndefault policy is like copy like a pro19:35
c4goldswwiking Hey there19:35
lisitsynso it's owning19:35
c4goldswIt's fine, no rush.19:35
lisitsynand I tried to add non-owning policy19:35
c4goldswHey HeikoS19:35
lisitsynbut here comes maths19:35
lisitsynyou have owning x owning19:35
lisitsynowning x non-owning19:35
lisitsynetc19:35
lisitsynand I am not even sure I covered everything19:36
@HeikoSCaBa: well yeah, the idea is to port them all :)19:36
lisitsynit's quite tough19:36
@HeikoSc4goldsw: hi there19:36
lisitsynthen as a desperate attempt I tried to add equals19:36
@wikinglisitsyn, ah i see ... let's see i mean this is definitely like 6.0.0 so we hav etime to figure out shit19:36
lisitsynok19:36
@wikingwe are gonna just fix notebooks19:37
@wikingnow19:37
@wikingand some builds19:37
@wikingand then release whatever is now in develop19:37
@wikingto test the new release pipeline + website19:37
lisitsynso tldr: it is conceptually fucked so I had to take time to think19:37
lisitsynthis means I context switched for an year19:37
lisitsynlol19:37
@wikingsure, we still have time19:37
@wikingtill 6.019:37
lisitsynmy context switching is like19:37
lisitsynyear by year19:37
lisitsynare you in the us still?19:38
@wikingyeah19:38
@wikingnext to heiko19:38
@wikingtrying to fix the last shit19:38
lisitsynslap him19:38
@wikingto be able to release today19:38
@wiking:D19:38
@HeikoSlisitsyn: what????19:38
-!- c4goldsw [8a33f496@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.244.150] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]19:38
lisitsynHeikoS: what? ;)19:39
@wikinganyhow let's see how the release pipeline works out19:39
@wiking :)))19:39
@wikinglisitsyn, have you seen that the buildbot is actually autodeploy the new website19:39
@wiking?: )19:39
lisitsynwiking: oh you rock19:39
@wikingwhen you push to shogun-web2 ? :D19:39
@wikingwe have now in buildbot like all sorts of automation like this19:39
@wikingnot only libshogun builds19:39
lisitsynoh this is good19:39
@wikingand basically thus19:40
@wikingwe have a full release pipeline19:40
@wikingthat creates all the artifacts19:40
@wikingthat is needed for a release19:40
@wikingonce you push/merge thing into master19:40
@wiking:D19:40
@wikingbut have never tested it yet19:40
@wikingso lets see how it works out19:41
@wiking:)))19:41
@wiking:)19:41
lisitsynproduction based testing19:41
lisitsynokie19:41
lisitsyn:P19:41
@wikingindeed19:41
@wikingwell fuck it19:41
@wikingwe just need to do it once19:41
@wikingand then it should be fine19:41
@wikingand then we can do more frequently releases19:41
@wikinghttps://hub.docker.com/r/shogun/shogun/19:43
@wikingthis is a libshogun nightly docker image :)19:43
@wikingbased on trusty19:43
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting]20:09
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun20:09
-!- HeikoS [~heiko@104.220.67.105] has quit [Ping timeout: 244 seconds]21:17
--- Log closed Tue Nov 01 00:00:57 2016