IRC logs of #shogun for Tuesday, 2016-11-08

« Monday, 2016-11-07 Index Wednesday, 2016-11-09 »
--- Log opened Tue Nov 08 00:00:07 2016
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 250 seconds]02:58
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun02:59
shogun-buildbotbuild #1 of clang - static analysis is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/clang%20-%20static%20analysis/builds/103:02
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 256 seconds]03:11
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun03:14
@wikingyeey03:32
@wikingnightlies are back03:32
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun03:52
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 268 seconds]05:09
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun05:09
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 248 seconds]05:31
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun05:33
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 244 seconds]06:04
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun06:05
Saurabh7sukey should auto kick that US election bot :))08:51
@sukeyNew Comment on Issue #3517 by amarlearning: "@karlnapf please mention the sample template url ?" - https://github.com/shogun-toolbox/shogun/issues/3517#issuecomment-25906980208:58
-!- praisethemoon [~praisethe@41.226.248.123] has joined #shogun09:05
-!- praisethemoon [~praisethe@41.226.248.123] has quit [Changing host]09:06
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun09:06
sonne|workwiking: nice! now what is missing is the http://cloud.shogun-toolbox.org/notebook/neuralnets_digits.ipynb09:20
sonne|worke.g...09:20
@sukeyNew Comment on Issue #3517 by amarlearning: "@karlnapf  It would be great if you could mention the link to this website http://shogun.ml/showroom " - https://github.com/shogun-toolbox/shogun/issues/3517#issuecomment-25907418509:24
@sukeyNew Comment on Issue #3517 by amarlearning: "@karlnapf  It would be great if you could mention the link to this website http://shogun.ml/showroom09:25
@sukeyI mean the link to the file structure of this website! " - https://github.com/shogun-toolbox/shogun/issues/3517#issuecomment-25907418509:25
sonne|workand various ipython notebooks do not have a heading 1 level at their top09:25
sonne|workso the description is missing and we just have the stupid filename o_O09:25
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 250 seconds]09:27
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun09:33
sonne|workOK I see that newer ipython notebooks have no more heading fields but use markdown like # header level 110:19
sonne|workso we need to add that logic to the parser10:19
-!- sonne|work [~sonnenbu@88.134.176.135] has quit [Quit: Leaving.]11:18
-!- witness_ [uid10044@gateway/web/irccloud.com/x-oxsvymicjuoevouj] has joined #shogun12:14
@sukeyNew Comment on Pull Request "It is in the base class, but I guess i should add it here too :)" by Saurabh7: https://github.com/shogun-toolbox/shogun/pull/3427#discussion_r8697430712:58
@sukeyReceived a new pull_request_review event, just so you know.12:59
@sukeyNew Comment on Pull Request "Can't operate on different types with eigen, I tried. " by Saurabh7: https://github.com/shogun-toolbox/shogun/pull/3427#discussion_r8697441312:59
@sukeyReceived a new pull_request_review event, just so you know.12:59
@sukeyReceived a new pull_request_review event, just so you know.13:04
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Ping timeout: 260 seconds]13:58
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has joined #shogun14:07
protinoHi14:07
protinoI m facing problem in building code14:07
protinoso at present i have just set it up without any interfaces like Python .14:08
protinoCan someone help me ?14:09
protinoProcedure followed : cmake .. -DPythonModular=ON14:10
protinoThen make all gives error ..14:10
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has quit [Quit: Page closed]14:26
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has joined #shogun14:26
-!- witness_ [uid10044@gateway/web/irccloud.com/x-oxsvymicjuoevouj] has quit [Quit: Connection closed for inactivity]14:50
@sukeyReceived a new pull_request_review event, just so you know.15:15
@sukeyReceived a new pull_request_review event, just so you know.15:16
@sukeyNew Comment on Pull Request "You're right.. this is not necessary, but helps with readability and similar practice is followed everywhere in Shogun, so... :)" by lambday: https://github.com/shogun-toolbox/shogun/pull/3427#discussion_r8699564715:16
@sukeyNew Comment on Pull Request #3427 by lambday: "Thanks @Saurabh7 .. let me know when you update the PR.15:17
@sukey@karlnapf yeah I'll merge it then :)" - https://github.com/shogun-toolbox/shogun/pull/3427#issuecomment-25914708315:17
@sukeyReceived a new pull_request_review event, just so you know.15:18
-!- lambday [31cf37a7@gateway/web/freenode/ip.49.207.55.167] has joined #shogun15:32
-!- mode/#shogun [+o lambday] by ChanServ15:33
@sukeyPull Request #3518 "Adding news entry for templated LDA"  opened by c4goldsw - https://github.com/shogun-toolbox/shogun/pull/351815:35
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun15:36
-!- mode/#shogun [+o HeikoS] by ChanServ15:36
-!- c4goldsw [8a33f486@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.244.134] has joined #shogun15:38
c4goldswping wiking15:38
c4goldswHey HeikoS, are you around?  I submitted a patch for the news file.  I saw the result of the nightly build and saw that the LDA unit test has memory leaks.15:42
@HeikoSc4goldsw: gotta fix those :) are they new? caused by your patch?15:43
@HeikoSc4goldsw: currently on phone but will be half reading irc15:43
c4goldswThey're probably caused by my patch, I'd have to look at a previous build.15:44
c4goldswI'll get onto that then.15:44
@wikingk15:50
@wikingwe can merge15:50
@wikingthe news15:50
@sukeyPull Request #3518 "Adding news entry for templated LDA"  merged by vigsterkr - https://github.com/shogun-toolbox/shogun/pull/351815:50
@sukeyNew Commit "Merge pull request #3518 from c4goldsw/develop15:50
@sukeyAdding news entry for templated LDA" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/e4092bc9fc7051290d480ef75f0eba66738e3e1915:50
@wikingsukey, do you job15:50
@wikingHeikoS, i reckon i should teach sukey to actually do github commands :)15:51
c4goldswWho's sukey?15:51
@wikinghubot15:51
@wiking:)15:51
c4goldswDance sukey15:52
c4goldswDarn.15:52
@wikingshogun-buildbot, dance15:52
shogun-buildbot<(^.^<)15:52
shogun-buildbot<(^.^)>15:52
shogun-buildbot(>^.^)>15:52
shogun-buildbot(7^.^)715:52
shogun-buildbot(>^.^<)15:52
c4goldswsukey, dance.15:52
@wikingthat's a privilage for shogun-buildbot15:52
@wiking:)15:52
c4goldswHaha, fair enough.15:52
@wikingHeikoS, ping15:53
@wikingah fuck money!15:53
@wikinglisitsyn, ping russia15:53
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun15:54
@wikingmizari, i want your tld!15:56
c4goldswwiking, what do SG_REF() and SG_UNREF() do again?15:56
@wikingrefernce_counter++15:57
@wikingreference_counter--15:57
@wikingif reference_counter == 0  => free memory15:57
c4goldswGreat, thank you.15:58
@wikingnw15:58
c4goldswAddios.15:58
-!- c4goldsw [8a33f486@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.244.134] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]15:58
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has quit [Ping timeout: 260 seconds]16:00
shogun-buildbotbuild #51 of bsd1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/51  blamelist: Viktor Gal <vigsterkr@gmail.com>, christopher <christopher.goldsworthy@outlook.com>16:02
-!- zero1hac [~zerooneha@139.59.16.180] has quit [Ping timeout: 256 seconds]16:04
@sukeyNew Comment on Pull Request "@lambday  Actually that would mean I have to cast LHS and RHS feature matrix to double. I think just casting the norms is better (even if we have to cast the result matrix) , memory wise ?" by Saurabh7: https://github.com/shogun-toolbox/shogun/pull/3427#discussion_r8700667016:08
@sukeyReceived a new pull_request_review event, just so you know.16:08
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting]16:11
mizariwiking, come visit Siberia16:11
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun16:11
@wiking:D16:11
mizariwhat's with su tld anyway?16:12
mizaripretty boring, I think16:12
@wikingno it's not16:13
@wikingit's almost as good as .yu16:13
@wiking:)16:13
mizariwell, .yu is something!16:14
mizaribut .su?16:15
mizariI'm afraid I disagree16:15
@wiking:>16:15
@wikingbut it's cool that your provider could keep it16:15
@wikingthey took away all the .yu tlds16:16
@wiking:<16:16
mizaria shame16:16
@wikingexactly16:17
@wikingshogun-buildbot: force build --branch=develop 'bsd1 - libshogun'16:19
shogun-buildbotbuild #52 forced16:19
shogun-buildbotI'll give a shout when the build finishes16:19
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Ping timeout: 260 seconds]16:20
shogun-buildbotbuild #52 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/5216:21
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has quit [Ping timeout: 256 seconds]16:26
-!- fhal [~fhal@my83-216-95-118.cust.relish.net] has joined #shogun16:27
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 260 seconds]16:33
-!- zero1hac [~zerooneha@139.59.16.180] has joined #shogun16:33
@sukeyReceived a new pull_request_review event, just so you know.16:35
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has joined #shogun16:36
-!- mode/#shogun [+o HeikoS] by ChanServ16:37
@wikingHeikoS, ping16:43
@wikingHeikoS, https://github.com/Homebrew/homebrew-science/pull/458316:43
-!- praisethemoon [~praisethe@41.226.248.123] has joined #shogun17:01
-!- praisethemoon [~praisethe@41.226.248.123] has quit [Changing host]17:02
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun17:02
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has quit [Ping timeout: 246 seconds]17:02
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Read error: Connection reset by peer]17:10
-!- zero1hac [~zerooneha@139.59.16.180] has quit [Ping timeout: 260 seconds]17:19
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has joined #shogun17:21
-!- mode/#shogun [+o HeikoS] by ChanServ17:21
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has joined #shogun17:23
protinomake cookbook is giving error. Can someone help please ?17:24
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has quit [Quit: Leaving.]17:30
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has joined #shogun17:30
-!- mode/#shogun [+o HeikoS] by ChanServ17:30
protinoHi. can someone help me with make cookbook ?17:30
-!- HeikoS [~heiko@eduroam-int-pat-8-202.ucl.ac.uk] has quit [Ping timeout: 250 seconds]17:36
@wikingprotino, what's the error/17:36
@wiking?17:36
@wikingshogun.ml/cookbook/latest/index.html17:36
@wikinghttp://shogun.ml/cookbook/latest/index.html17:36
@wikingthat's 5.0.0's cookbooks17:37
protinoonce i have build the source code in /shogun/build17:42
protinothen i have to goto /shogun/doc/cookbook/build17:42
protinoand the go for cmake .. and then make cookbook .17:42
@wikingyou dont need to build the src17:43
@wikingto build cookbook17:43
@wikingand for building cookbook17:43
protinoyes...17:43
@wikingjust run make cookbook17:43
@wikingin shogun/build17:43
protinoinsider shogun/build ??17:44
protinoits showing no such target17:44
protinoprotino@protino:~/shogun/build$ make cookbook make: *** No rule to make target `cookbook'.  Stop17:44
@wikingthen you are missing some dependencies17:44
@wikingfor cookbok17:44
@wiking*cookbook17:44
protinoonce i have cloned the git repo.17:44
@wikingmake sure you have17:44
@wikingshogun/doc/cookbook/requirements.txt17:44
protinothe i need to create the build dir .17:44
@wikinginstalled17:44
@wikingall the required packages17:45
@wikingi.e pip install -r shogun/doc/cookbook/requirements.txt17:45
protinoohkk.. let me try. thanks .17:46
protinoI have again downloaded the shogun code. Then create build directory . Then cmake .. -DJavaModular=ON.17:50
protinoThe make cookbook is not coming as target. I have ran pip install -r shogun/doc/cookbook/requirements.txt .17:50
protino@wiking, i have searched on all MakeFile in build/ none of them contains cookbook18:01
@sukeyNew Comment on Issue #3513 by sudk1896: "@karlnapf : I did investigate usage of `KNN_KDTREE`, after some Googling I found that it is faster than the others only on low dimensional data. In fact when I ran the tests on the `usps` dataset, The results were -18:20
@sukey`Standard KNN took 1.9s18:20
@sukeyCovertree KNN took 1.0s18:20
@sukeyKDTree KNN took 3.7s`18:20
@sukeyHow do I make changes to the notebook (the multiclass KNN one in doc folder) ? It has a very weird format and I do not know how to add the code to it.18:20
@sukeyDo you also want me add a plot for the `KNN_KDTREE` as well like the rest ?18:20
@sukeyThanks." - https://github.com/shogun-toolbox/shogun/issues/3513#issuecomment-25919973718:20
@sukeyNew Comment on Issue #3513 by sudk1896: "@karlnapf : I did investigate usage of `KNN_KDTREE`, after some Googling I found that it is faster than the others only on low dimensional data. In fact when I ran the tests on the `usps` dataset, The results were -18:20
@sukey`Standard KNN took 1.9s18:20
@sukeyCovertree KNN took 1.0s18:20
@sukeyKDTree KNN took 3.7s`18:20
@sukeyHow do I make changes to the notebook (the multiclass KNN one in doc folder) ? It has a very weird format and I do not know how to add the code to it.18:20
@sukeyDo you also want me add a plot for the `KNN_KDTREE` as well like the rest ?18:20
@sukeyThanks." - https://github.com/shogun-toolbox/shogun/issues/3513#issuecomment-25919973718:20
@sukeyNew Comment on Issue #3513 by sudk1896: "@karlnapf : I did investigate usage of `KNN_KDTREE`, after some Googling I found that it is faster than the others only on low dimensional data. In fact when I ran the tests on the `usps` dataset, The results were -18:21
@sukey`Standard KNN took 1.9s`18:21
@sukey`Covertree KNN took 1.0s`18:21
@sukey`KDTree KNN took 3.7s`18:21
@sukeyHow do I make changes to the notebook (the multiclass KNN one in doc folder) ? It has a very weird format and I do not know how to add the code to it.18:21
@sukeyDo you also want me add a plot for the `KNN_KDTREE` as well like the rest ?18:21
@sukeyThanks." - https://github.com/shogun-toolbox/shogun/issues/3513#issuecomment-25919973718:21
@wikingprotino, after installing the dependencies18:40
@wikingyou have to re-run the cmake18:40
protinoyes.. now make cookbook ran18:42
protinoIOError: Generated listing /home/protino/shogun-shogun_5.0.0/build/examples/meta/python/gaussian_processes/gaussian_process_regression.py not found, it is expected to be created from the corresponding meta-example18:43
@wikingok cool18:43
@wikingmmm error18:43
@wikingbut wait18:43
@wikingthis is not when running make cookbook18:44
@wikingthis is make meta_examples18:44
protinonope i ran make cookbook .18:44
protinoprotino@protino:~/shogun-shogun_5.0.0/build$ make cookbook [  0%] Copying cookbook Sphinx static overrides to /home/protino/shogun-shogun_5.0.0/build/doc/cookbook/_build [  0%] Built target cookbook_sphinx_static [ 20%] Built target ctags [100%] Built target meta_examples [100%] Copying cookbook Sphinx source to /home/protino/shogun-shogun_5.0.0/build/doc/cookbook/_build [100%] Built target cookbook_copy_source [100%] Creating Quicks18:44
protinoi have even used a stable version 5.0.0 still i m getting same error .18:45
@wikingok just a second18:49
@wikingbtw18:57
@wikingcan you actually checkout the git repo18:57
@wikingdo the following in it18:57
@wikinggit submodule init18:57
@wikinggit submodule update18:57
@wikingand then ran those comes you tried18:57
protinosure.18:58
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun18:59
-!- mode/#shogun [+o HeikoS] by ChanServ18:59
@HeikoSprotino: still there?18:59
@HeikoSwiking: jo18:59
protinoYup.19:00
@wikingHeikoS, hey19:00
@HeikoSwiking: I would be scared to template that19:00
@HeikoSso messy19:00
@HeikoSprotino: what do you need help with?19:00
protinorunning make cookbook gives error .19:00
@HeikoSlisitsyn:19:01
@wikingprotino, plz try what i've said19:01
lisitsynyeah19:01
protinosure.19:01
lisitsynlets move here then19:01
protinodoing that .19:01
@HeikoSprotino: ah you already talked to wiking19:01
@wikinglisitsyn, putty19:01
protinohmm.19:01
@wikinglisitsyn, bitch please you answer to HeikoS here but not to me19:01
@wiking:)19:01
lisitsynwiking: haha19:02
lisitsynno19:02
lisitsynI thought it is better to keep it in email19:02
@HeikoSwiking: lisitsyn is my b***19:02
lisitsyndaddy19:02
@sukeyNew comment by vigsterkr    on Commit e4092bc9fc7051290d480ef75f0eba66738e3e19: good that you caught that  :) - https://github.com/shogun-toolbox/shogun/commit/e4092bc9fc7051290d480ef75f0eba66738e3e19#commitcomment-1974368719:02
@HeikoSI gotta go soon19:03
lisitsynmeh19:03
@wikinglisitsyn,19:03
@HeikoSbut lisitsyn so I didnt get your idea19:03
@HeikoSwith the distribution thing19:03
@wikingbut lets say19:03
@wikingyou have svm19:03
@wikingSVM(RangeDomain C)19:03
@wikingand then svm would have 2 class param19:03
@wikinga) actual value of C19:04
@wikingb) the domain of C19:04
@wiking?19:04
lisitsynuhmm I've lost some messages19:04
lisitsynI see your a) and b)19:04
lisitsynis that all?19:04
@wiking  <wiking> SVM(RangeDomain C)19:04
@wiking[13:02]  <wiking> and then svm would have 2 class param19:04
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Ping timeout: 252 seconds]19:04
lisitsynahh19:04
@wikingthat was above it19:04
lisitsynmaybe it makes sense to separate19:04
@wikingbecause we want to be able to enforce19:05
@wikingthe actual value of C19:05
@wikingalready on the API sde19:05
@wikingso that one can only give valid values19:05
@wikingor you would have an internal mapping19:05
@wikingonly visible in the implementation19:05
@wikingthat say19:05
lisitsynI would go for something like19:05
lisitsynyou have svm kernels etc19:06
@wikingparam C = RangeValue19:06
lisitsynthen you create some thing19:06
lisitsynlike19:06
lisitsynmake C be exp range 1 to 1019:06
lisitsynmake kernel a set of gaussian and linear19:06
@sukeyNew Comment on Issue #3513 by karlnapf: "You directly can open the notebook in your browser. Google for ipython notebook and how to use it. Then you send a standard patch on via github pull request. Please clear all output of the notebook before sending it. And also please provide an html link to a fully rendered version (say on gist)" - https://github.com/shogun-toolbox/shogun/issues/3513#issuecomment-25921219419:06
lisitsynyou pass this thing and then it tries to find something for you19:06
@HeikoSlisitsyn: this is not only for modelselection19:07
@HeikoSbut also for value checking in getter setter19:07
@HeikoSalso callback functions should be possible to define things like prior distributions to sample from19:07
lisitsynbut validity is much simpler19:07
@HeikoSthe linear range etc is just a very special case19:07
lisitsynyeah yeah but what I mean19:08
lisitsynI am ok with keeping svm having one value of C19:08
-!- mode/#shogun [+o lisitsyn] by wiking19:08
-!- mode/#shogun [-o lisitsyn] by ChanServ19:08
@wikingfucking ChanServ19:08
lisitsynit's perfect ok with me19:08
lisitsynnot sure we need to push domains there19:08
protino@wiking facing same error :(19:09
@wikingprotino, could you paste the whole error in pastebin.com19:09
@wikingand paste here the url plz19:09
@wikinglisitsyn, i mean the idea here was to have both solved in one19:09
protino(y)19:09
@wikingenforce value19:09
lisitsynboth validation and selection?19:09
@wikingand ease modelselection19:09
@wikingyep19:10
lisitsynwhat if you already know you need C=3.019:10
@sukeyIssue #3519 "write a NOTEBOOKS.md readme" opened by karlnapf - https://github.com/shogun-toolbox/shogun/issues/351919:10
@sukeyIssue #3519 "write a NOTEBOOKS.md readme" karlnapf added label: "entrance" - https://github.com/shogun-toolbox/shogun/issues/351919:10
@sukeyIssue #3519 "write a NOTEBOOKS.md readme" karlnapf added label: "Documentation" - https://github.com/shogun-toolbox/shogun/issues/351919:10
@wikingif that's a valid value for19:10
@wikingC19:10
@wikingthen you just pass that19:10
@wikingi mean RangeValue->get() would give you back the value19:11
@HeikoSin some sense19:11
@HeikoSthe values should be stored as before19:11
@wikingor any operator you wish19:11
@wikingyeah but should be encapuslated19:11
lisitsynok the thing we should decide19:11
lisitsynis19:11
@HeikoSbut just when registering the variables as parameters, once should be able to give domain information19:11
protino@wiking http://pastebin.com/sDjUi3bb19:11
lisitsyndo we go for composite parameters19:11
@HeikoSto have some model meta information in there without changing the representation of the variables in the class19:11
lisitsynI mean do we always assume svm has *a set of C values*19:12
lisitsyn(in general case)19:12
@HeikoSprotino: run "make meta_examples"19:12
lisitsynI somehow like it but it's a change19:12
@wikingprotino, and what happens when you run make meta_examples19:12
@HeikoSwiking: I think this might be cmake bug :)19:12
@HeikoScmake config error i mean19:12
lisitsynand it intertwines two systems together19:12
@wikingmmm weird19:12
@wikingbecause for me it triggered the meta_examples19:13
protino@wiking that ran successfully19:13
@wikingprotino, try now again make cookbook19:13
@sukeyIssue #3520 "`make cookbook` should have `make meta_examples` as dependency" opened by karlnapf - https://github.com/shogun-toolbox/shogun/issues/352019:13
protinoits failing with same error19:13
@sukeyIssue #3520 "`make cookbook` should have `make meta_examples` as dependency" karlnapf added label: "BUG" - https://github.com/shogun-toolbox/shogun/issues/352019:13
@sukeyIssue #3520 "`make cookbook` should have `make meta_examples` as dependency" karlnapf added label: "entrance" - https://github.com/shogun-toolbox/shogun/issues/352019:13
@sukeyIssue #3520 "`make cookbook` should have `make meta_examples` as dependency"- https://github.com/shogun-toolbox/shogun/issues/352019:13
@sukeyIssue #3520 "`make cookbook` should have `make meta_examples` as dependency" karlnapf added label: "cmake" - https://github.com/shogun-toolbox/shogun/issues/352019:13
@sukeyIssue #3519 "write a NOTEBOOKS.md readme"- https://github.com/shogun-toolbox/shogun/issues/351919:13
@wikingprotino,  The full traceback has been saved in /tmp/sphinx-err-ibn7bR.log, if you want to report the issue to the developers.19:13
@wikingcan you pastebin that content?19:13
protinoyep.19:14
@wikinglisitsyn, i mean for me that is quite elegant19:14
@wikingthat you can validate & have the domain at once19:14
lisitsynyes do we all agree19:14
lisitsynthat in general case there won't be single parameter of svm19:14
protino@wiking http://pastebin.com/BNHtyAxp19:14
lisitsynthere always be some set or distribution (or any other domain)19:15
@sukeyIssue #3519 "write a NOTEBOOKS.md readme"- https://github.com/shogun-toolbox/shogun/issues/351919:15
lisitsynHeikoS: wiking: ^19:15
@wikingHeikoS, i think you fired that issue before it actually made sense :)19:15
-!- zero1hac [~zerooneha@139.59.16.180] has joined #shogun19:15
@wikingprotino, i reckon you have a missing dependency19:15
@wikingthats why gaussian_process_regression.py is not generated19:16
@wikingand but cookbook is not aware that it should not look for it19:16
@wikinglisitsyn, yeah i reckon that that'd be a case19:16
@wikingi mean in worse case the parameter is any number19:16
@wiking:)19:16
@HeikoSwiking: ok we need to fix that as well I guess?19:16
@HeikoSah19:16
@HeikoSit is the guards19:16
protino@wiking thanks i will work on that :)19:17
@wikinglisitsyn, but yeah i would 'box' it still19:17
lisitsynbox?19:17
@wikinglike java boxing19:17
@wikingyou know19:17
lisitsynyes19:17
@wikingnot a primitive type anymore19:17
@wikingit's an object etc.19:17
lisitsynno primitive19:17
lisitsynobject19:17
lisitsynbut abstract object19:17
@wikingyep19:17
lisitsynand no templates please19:17
lisitsynthey are incompatible with interfaces19:17
lisitsyn:P19:17
@wiking:D19:17
@wikingtrue19:17
@wikingi was just starting a conversation man19:18
@wiking:D19:18
lisitsynok then lets have base class domain19:18
lisitsynan iterator + sampler19:18
@wikingthe reason i pinged you19:18
@sukeyNew Comment on Issue #3520 by karlnapf: "This comes since `FindMetaExamples.cmake` sometimes removes certain meta examples.19:18
@sukeyThe cookbook should still build fine, just ignoring that particular page.... " - https://github.com/shogun-toolbox/shogun/issues/3520#issuecomment-25921530519:18
@wikingbecause you are much better in this than any of us here19:18
@sukeyIssue #3520 "`make cookbook` fails when certain cookbooks are guarded"- https://github.com/shogun-toolbox/shogun/issues/352019:18
@wikingyeap the iterator sounds good19:18
@wikingand then what if it's like19:18
@wikingunset variable19:19
@wikingand AnyDomain19:19
lisitsynnot sure if we need any domain19:19
@HeikoSiterator ??19:19
@wikingthen basically nextValue() returns (double)random()19:19
@wiking?19:19
lisitsynah19:19
@HeikoSthink of optimization algorithms19:19
@HeikoSthey want to propose the next value19:19
@HeikoSfrom gradient or whatever19:19
lisitsynHeikoS: yes also gradient19:19
lisitsynor something19:19
@HeikoSbut the parameter itself cannot know that19:20
@HeikoSyou can only do that to generate a-priori values19:20
@HeikoSbut for me this is more about telling a parameter about its domain19:20
@wikingok but then you need like a validator function? :D19:20
@wikingboolean operator()19:20
@wiking:)19:20
@HeikoSbut then you can change representation19:20
@HeikoSto an unconstrained domain19:20
@HeikoSi.e. you do gradient descent for a positive variable19:20
@HeikoSyou just do gradient descent in log-space and exponentiate after every step19:21
lisitsynbut you don't change domain when you started optimization?19:21
@HeikoSthats different thing, that is runtime thing19:21
@wikingHeikoS, but that search would be done on primitive types no?19:21
@HeikoSwe are here talking about things that are fix at compile time as they are linked to the model19:21
@HeikoSwiking: yeah sure19:21
@wikingi mean you woulnd't like run the gradient search on those boxed parameters19:21
@HeikoSno you wouldnt19:22
@wikinglisitsyn, ^ ?19:22
@HeikoSso there is two things here: allowing users/coders to restrict values at their wishes19:22
@HeikoSbut there is also model parameters (like svm C) that have to be positive19:22
@HeikoSno matter what19:22
@HeikoSif if I search for C19:22
@HeikoSI do that in an unconstrained space, say real line19:23
@HeikoSand then I take exp after19:23
lisitsynwhy compile time?19:23
@HeikoSfor the actual value19:23
@HeikoSbecause C is always positive19:23
@HeikoSnot matter what19:23
lisitsynwe won't have any compile time thing at all19:23
@wikingyeah actually true19:23
@HeikoSsorry, I mean model definition time19:23
lisitsynah19:23
@wikingwe should not depend on compile time :P19:23
@HeikoSbut thats different to user selects subset of domain19:23
@wiking(nod)19:23
@HeikoSor like user defines search space for grid search19:24
lisitsynok ok lets make domain an entity19:24
lisitsynentitify it :D19:24
lisitsynwhat can we do with domain?19:24
@wiking:>>>19:24
lisitsynand what can domain tell us?19:24
lisitsynI have two things in mind19:24
@HeikoSthe whole idea came up because of this:19:24
lisitsynwe can retrieve all its values19:24
@HeikoShorrific construction:19:24
lisitsynand we can sample a value19:24
lisitsynwhat else?19:25
@HeikoShttps://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/GaussianKernel.cpp#L9919:25
@HeikoSand we have a few of those explicit reparametrisations19:25
@HeikoSsince they are needed19:25
lisitsynwell I don't see a problem with that - is it?19:26
@HeikoSbut here for kernel bandwidth, it just makes much more sense to store the log value19:26
@HeikoSwell automate this thing19:26
@HeikoSgit grep 'must be positive'19:27
@wikinglol19:27
@HeikoSI gotta run, sorry :) lets discuss more later19:27
@HeikoSI mean with me, you two go ahead :)19:27
@HeikoSprotino: https://github.com/shogun-toolbox/shogun/issues/352019:28
@HeikoSprotino: bundle nlopt, then this should work, but it is a bug in our configuration script19:28
@HeikoSprotino: check the FindMetaExamples.cmake file, there some of them are excluded if some dependency is not met19:28
lisitsynok let me think about it then19:28
@HeikoSif you install all dependencies (or bundle them), then it should work19:28
@HeikoSprotino: but you can also just use the website cookbook, and look at the generated example code listings locally19:29
protinoyep. well i wanted to contributed to cookbook so once i start writing them i would need to run them locally.19:30
protino*contribute19:30
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 265 seconds]19:34
@wikingprotino, if you are too lazy19:34
@wikingyou can always send in prs19:34
@wikingand check the output19:34
@wikingof the PR buildbot19:34
@wiking;)19:34
protino@wiking thanks. i will try to resolve it else resort to buildbot :)19:35
@wikingyeah but nlopt should solve your problem19:38
@wikingor send in the patch that skips cookbooks19:38
@wikingfor which you dont have the dependency19:38
@wikingeither is fine19:38
@wiking:)19:38
@sonney2kwiking, where is that script doing the ipynb parsing?19:48
@sonney2kwiking, seems like you didn't fix the title of nb's yet right?19:49
@wikingsonney2k, no idea :)19:49
@wikingsonney2k, all notebook relates stuff19:50
@wikingis in infra repo19:50
@wikingit's not autodeploy yet19:50
@wikingso ping me if you updated things in it19:50
-!- protino [0e8bc20c@gateway/web/freenode/ip.14.139.194.12] has quit [Ping timeout: 260 seconds]20:38
@sukeyNew Comment on Issue #3520 by IOcodegeass: "I faced this problem. And installing **nlopt** solved it.20:43
@sukeyThanks :)" - https://github.com/shogun-toolbox/shogun/issues/3520#issuecomment-25923809520:43
@sonney2kwiking, gmane is dead http://news.gmane.org/gmane.comp.ai.machine-learning.shogun20:56
@wikingyep20:57
@sonney2kso our mailinglist archive is gone20:57
@wikingi know20:57
@sonney2kdo we have a replacement20:57
@wikingwe need to enable20:57
@wikingarchive20:57
@wikingin our mailman20:57
@sonney2kit is not mailman20:57
@sonney2kor did you convert the list?20:57
@wikingno20:57
@wikingi didn't touch20:57
@wikingbut i remember a mailman -like admin site :D20:57
@sonney2kI thought that was for gsoc20:57
@sonney2kwiking, nice that we have a shogun tag on stack overflow now :)20:58
@wiking\p/21:00
@sonney2kbtw what happened to shogun-cloud?21:01
@sonney2kthe link is broken on the website too..21:01
@sonney2kand the interactive examples?21:01
@sonney2kall gone?21:01
@wikingsonney2k, working on it21:05
@wikingbut we didn't want to wait with the release21:05
@wikingsince it's a bigger revamp21:05
@wikingbut soon soon21:05
@wikingdemos is still existing21:06
@wikingmaybe redirection needs fixing21:06
@sonney2kwiking, the website doesn't cover them anymore (well not working now but still nice demos IMHO)21:07
@wikinghttp://demos.shogun-toolbox.org/regression/svr/21:07
@wikingworks21:07
@wikingjust no reference21:07
@sonney2kwiking, so where do we put the references?21:14
@wikinganywhere21:14
@wikingbut be ncie :)21:14
@sonney2kwe used to have screenshots and some wheel thingy21:16
@wikingmmmm21:16
@wikingshogun-web repo woudl have that for sure21:16
@sonney2khell nips.cc is sold out this year21:17
@wikinglol21:17
@wiking:)21:17
@wikingbcn is popular21:17
@sonney2kwell cheap to get to from europe..21:17
@sonney2kand cool21:18
@sonney2kagreed21:18
@sonney2kalmost as hip as berlin is :P21:18
@wikinghttps://arxiv.org/pdf/1605.06049v2.pdf21:18
@wiking:)21:18
@wikinghahahaha21:18
@sonney2kI don't know where to put it on the site21:20
@sonney2kwiking, can't kevin invent something nice?21:20
@wikingdunno21:21
@sonney2kI must be the super web guru by now...21:21
@wikinghe is not around for a long time21:21
@sonney2kof course not he likely has a life :P21:21
@wiking:>21:21
@sonney2kwiking, so we have sgtb.eu21:24
@sonney2kshogun.ml21:24
@sonney2kshogun-toolbox.org21:24
@wikingyes21:24
@sonney2kdid I forget any :P21:25
@wikingnot that i know of21:25
@sonney2kwiking, another broken link http://www.shogun-toolbox.org/page/about/contributions21:42
@sonney2kSee our list of integrated external libraries.21:42
@wikingmm21:42
@sonney2kwiking, okido ... let me get some sleep21:43
@sonney2kcu!21:44
@wiking:)21:46
@wikinggnite21:46
--- Log closed Wed Nov 09 00:00:08 2016
« Monday, 2016-11-07 Index Wednesday, 2016-11-09 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!