Open in new window / Try shogun cloud
--- Log opened Fri Feb 22 00:00:49 2013
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]00:10
-!- FSCV [~FSCV@173.254.212.46] has quit [Ping timeout: 264 seconds]03:06
shogun-buildbot_build #290 of nightly_default is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/29004:21
blackburnsonney2k: sonne|work: http://song-of-github.herokuapp.com/?username=sonney2k07:33
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Ping timeout: 255 seconds]08:08
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun08:16
-!- zxtx [~zv@c-24-18-130-24.hsd1.wa.comcast.net] has joined #shogun09:14
-!- wiking [~wiking@info2k1.hu] has joined #shogun09:26
-!- wiking_ [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer]09:26
-!- n4nd0 [~nando@n150-p30.kthopen.kth.se] has joined #shogun09:38
n4nd0blackburn: that's a cool thing!09:44
blackburnn4nd0: yeah09:44
blackburnhttp://song-of-github.herokuapp.com/?username=lisitsyn I sound funny too09:45
n4nd0blackburn: yeah, I listened to you too :)09:45
blackburnn4nd0: tapkee sounds intensive :D09:45
blackburnn4nd0: I had a period with commit streak for 27 days09:46
n4nd0blackburn: one can introduce projects too?09:46
blackburnthat was tapkee development09:46
blackburnno I mean tapkee on my timeline09:46
n4nd0or within your song?09:46
n4nd0ok09:46
blackburnlets listen to yours09:46
blackburnpretty sparse! :)09:46
n4nd0hehe yeah :(09:46
n4nd0it starts well but...09:47
blackburnmake that song happier n4nd009:48
blackburn:)09:48
n4nd0let's see :)09:48
n4nd0oh09:50
n4nd0today a do a yeah on github09:50
n4nd0year09:50
n4nd0year of contributions Feb 22 2012 - Fec 22 201309:50
blackburnn4nd0: I don't think so :) it just shows an year09:51
n4nd0lol I am stupid09:51
n4nd0:D09:52
blackburnn4nd0: https://github.com/lisitsyn same for me09:52
sonne|workn4nd0: have you seen? I put your changes live. now please fix the website :)10:01
sonne|workI think we should add the gsoc ideas / stuff back from 20{10,11,12}10:01
n4nd0sonne|work: hey! yeah I saw it yesterday at night10:01
n4nd0sonne|work: I think it is fixed, or do you see something wrong?10:02
sonne|worklooks a bit weird when we have no sub menus but still the menu line is empty10:02
sonne|workI mean in: http://shogun-toolbox.org/page/about/information10:03
sonne|workthe black line is empty10:03
n4nd0yeah I see what you mean10:03
n4nd0sonne|work: how would you like to see that?10:04
sonne|workto have them in the black line usually only when a sub-menu is available not?10:07
sonne|workno idea if that's easy to do. maybe it is easier to add another menu line below then?10:07
n4nd0let me think of it a moment10:08
sonne|workn4nd0: the other thing that is confusing me is that I wanted to have http://shogun-toolbox.org/page/Events/workshop201310:08
sonne|workto be a news subpage10:08
n4nd0sonne|work: I think the second one is easier10:08
sonne|workso it should display10:08
sonne|workUpdates | Program | Registration10:08
sonne|workbut no idea how to do that10:09
sonne|work(or if it is even possible)10:09
n4nd0it is similar to how it is done now10:09
n4nd0but the updates page is the main Workshop 201310:10
sonne|workn4nd0: a third idea - how about when you mouse hover over 'Workshop 2013' that it would then display a .js menu with Updates, Program Registration?10:10
n4nd0sonne|work: regarding the thing of the menu line below the current one10:10
sonne|workand then display that in the You are here: Events  /  workshop2013 / Updates10:11
sonne|worketc?10:11
n4nd0sonne|work: even if a new one is added, in the pages that do not have subpage, the menu line below would be empty10:11
sonne|workthen we don't need any new lines10:11
n4nd0sonne|work: so you mean that the sub-pages do not stay permanently but only popped up when the mouse is over the current parent10:12
n4nd0say in the current events page10:12
n4nd0when in workshop 2013, program and registration are not shown10:12
n4nd0they are shown when mouse over workshop 201310:12
sonne|workyes10:13
sonne|workand only in the you are here line this extra stuff is added10:13
n4nd0sonne|work: that can be done10:13
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has joined #shogun10:13
n4nd0so the mouse over would do the you are here line disappear?10:13
sonne|workn4nd0: no10:26
sonne|workn4nd0: the mouse over would just display a menu10:26
sonne|workwith e.g. Updates | Registration ...10:26
sonne|workand then when you click on updates10:27
sonne|workthe you are here line would display ...workshop -> updates10:27
sonne|workso one more level10:27
n4nd0and after clicking in, say Updates, the displayed menu Updates | Registration ... does not need to remain shown?10:28
n4nd0sonne|work: ^10:33
sonne|workn4nd0: yes10:36
n4nd0sonne|work: ok, I am not 100% but I think I should be able to do that with the little I've learnt about this stuff10:37
sonne|workn4nd0: one could think of displaying the hierarchy in the your are here line though, via You are here: Events  /  workshop2013 / (*updates* | registration | foo )10:38
sonne|workwhere *updates* is highlighted10:38
n4nd0ok10:38
n4nd0sonne|work: I see an issue10:38
sonne|workyes?10:38
n4nd0there would be a gap between the menu line showing e.g. Workshop, GSoC10:39
n4nd0and the the line You are here ...10:39
n4nd0so when the mouse is over Workshop or GSoC10:39
n4nd0the submenu can be shon10:39
n4nd0shown*10:39
sonne|workn4nd0: I don't understand11:09
sonne|workn4nd0: I meant some menu like this: http://demos.9lessons.info/DropMenu/MyDemo.html11:10
n4nd0sonne|work: draw for me in a paper what you want to have and send me a picture of it, it will take less time :D11:10
sonne|workso why shoudl there be a gap?11:11
n4nd0I didn't think of it like that11:11
n4nd0I didn't understand you meant a drop down menu11:11
n4nd0but a menu like the one we have now with the entries appearing along the same line11:12
n4nd0sonne|work: I can try to take a look to how to do this11:12
sonne|workn4nd0: yeah I meant a drop down menu11:13
n4nd0but no idea how long can this take me11:13
n4nd0sonne|work: what I have done now is probably rather simple and has taken me more than two days doing just that11:13
sonne|workwell just take the menu code from http://www.9lessons.info/2012/06/simple-drop-down-menu-with-jquery-and.html11:13
sonne|workI think it is rather easy to do then11:14
sonne|workyou only have to fill in the fields?11:14
n4nd0sonne|work: I have no idea about javascript11:14
n4nd0I have to see how it works together with django templates11:14
sonne|workn4nd0: no difference :)11:15
n4nd0there will probably be some issues like11:15
n4nd0when the menu drops down11:15
n4nd0does it cover the information under it?11:16
n4nd0such as the "You are here..." line11:16
n4nd0I will investigate a bit11:16
n4nd0sonne|work: btw, no issues in shogun-web yet?11:17
sonne|workn4nd0: no the menu is just displayed on top everything. once you click you have to change the you are here line no earlier11:18
sonne|workn4nd0: no issues AFAIK but I am not sure what you are looking for11:18
n4nd0sonne|work: roger that :)11:18
n4nd0https://github.com/shogun-toolbox/shogun11:19
n4nd0here there is a tab for issues11:19
n4nd0https://github.com/shogun-toolbox/shogun-web11:19
n4nd0and here, there is not11:19
n4nd0sonne|work: I wanted to write a couple of things that I think will be nice to remember - few fixes that the website needs11:20
sonne|workn4nd0: ok have a look at  https://github.com/shogun-toolbox/shogun-web again :D11:20
n4nd0sonne|work: that is what I meant, ty!11:20
sonne|workn4nd0: do you want me to create an issue and assign it to you :D11:27
n4nd0sonne|work: haha no need for it11:27
n4nd0that would put me more pressure to get it done fast :)11:27
n4nd0oops maybe it was not a good idea to say that...11:28
sonne|workn4nd0: if you do that by the end of today it is totally ok ;-)11:29
n4nd0sonne|work: lol I cannot do that, I have to take care of my degree project11:30
n4nd0it has been quite abandoned these days11:30
n4nd0sonne|work: btw, does the shogun logo on the top right corner appear overlapped with the FAQ page in your browser?11:33
n4nd0it does in mine11:33
n4nd0i remember blackburn said it looked ok in his11:33
-!- heiko [~heiko@nat-182-132.internal.eduroam.ucl.ac.uk] has joined #shogun11:34
sonne|workn4nd0: no it looks ok here - which browser?11:34
n4nd0sonne|work: chromium11:35
n4nd0sonne|work: it seems that it depends whether I zoom the page or not11:35
n4nd0but when default view is use - no zoom - it is overlapped11:35
sonne|workn4nd0: I am zooming in and out and no issues..11:36
sonne|workchrome here11:36
n4nd0chromium users are on their own then :D11:38
heikosonne|work, your patch refactor parameters to use SGVector as dims (for future ndarray support)        broke the streaming dense features, any reasons why you removed the lines in StreamingDenseFeatures.cpp ?12:49
heikoon another note, debug messages dont work anymore on my classes, I have no idea why this is12:51
blackburnheiko: does info wrok?13:02
heikoblackburn, no idea13:02
heikoprint does13:02
blackburnheiko: actually sonney2k added there branch prediction but I believe it should not disable anything13:03
heikoblackburn, see for example LinearTimeMMD.cpp line 13013:06
heikothis is not outputted13:06
heikobut if i add a SG_PRINT in the next line, I can see this13:07
heikoand blackburn, do we care if unit-tests leak memory?13:07
blackburnheiko: yeah I think so13:07
n4nd0heiko: stupid question, are you setting log_level with an SGObject?13:08
blackburnhah I guess so n4nd013:08
n4nd0just to be sure :)13:08
heikon4nd0, I am doing13:08
n4nd0heiko: all right, sorry13:08
heikosg_io->set_loglevel(MSG_DEBUG);13:08
heikoin my main13:08
heikobut doing it on the instance also doesnt help13:08
heikostrange thing is I CAN see other debug messages13:09
heikojust not all of them13:09
n4nd0aham13:09
blackburnheiko: ah so you just can't see that one??13:09
heikoblackburn, yes13:09
blackburnehm13:09
heikocode is executed however13:09
heikosince turning it into a print makes it appear13:09
blackburnquite interesting13:10
heikoindeed :D13:10
heikonot too important, I just noticed13:10
blackburnheiko: well what I could suggest is to remove branch predicting hint in SGIO13:10
blackburnand check13:10
blackburnI have no other idea and this shouldn't actually be the reason13:10
heikoblackburn, on another note: the unit-tests with valgrind is a quite sad thing13:11
blackburnheiko: you are just like russian tv13:12
heikoIs there actually a way of running single tests without having to create new files for this?13:12
blackburnno good news13:12
heikoblackburn why? :)13:12
heikoah13:12
heikoI am sorry13:12
blackburn:D13:12
blackburnjk13:12
heikowell I fixed many bugs today and yesterday :)13:12
blackburnheiko: what do you mean?13:12
heikothats good noews13:12
heikoblackburn, say I want to execute one test in one file13:12
heikofor fixing13:12
heikohow do I do this best?13:12
blackburnw/o make ?13:13
heikosince running them all is annoying if I only want to fix one13:13
blackburnfrankly I wanted that too but was lazy to check13:13
blackburnit is a good chance13:13
heikoI think it would be also good to have the valgrind statistics for the single files, not for all tests13:13
heikosince not easy to get an overview13:13
blackburnheiko: we should patch makefile for that I guess13:14
blackburnnow we just compile it into one executable13:14
heikoagreed :)13:15
heikoI am not sure whether one of my tests leaks but I dont want to go through them one by one :)13:16
-!- FSCV [~FSCV@173.254.212.46] has joined #shogun14:58
-!- FSCV_ [~FSCV@187.210.54.166] has joined #shogun15:35
-!- FSCV [~FSCV@173.254.212.46] has quit [Ping timeout: 260 seconds]15:38
-!- n4nd0 [~nando@n150-p30.kthopen.kth.se] has quit [Quit: leaving]16:12
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun16:32
sonne|workheiko: what did I remove?16:38
sonne|workheiko: what is with your LinearTimeMMD.cpp print stuff?16:39
sonne|workheiko: did you not set loglevels at the very beginning?16:39
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has quit [Ping timeout: 260 seconds]16:49
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Quit: Leaving.]17:01
heikosonne|work, https://github.com/shogun-toolbox/shogun/commit/49060b0a4fdc0a8beacf821e09ed04fab9e1367117:37
heikochanges in the file StreamingDenseFeatures.cpp17:37
heikoand I did set the loglevel at the very beginning17:37
-!- hoijui [~hoijui@141.23.95.39] has joined #shogun17:40
blackburnsonney2k: lets postpone release for one week - I will release tapkee as well17:46
blackburnso this way all patches will be applied17:46
blackburna few issues left in any case17:47
blackburnheiko: ^17:47
heikoblackburn, good! I think I will finish my stuff soon also17:47
heikomany warnings are still on me :)17:47
heikogithub is asleep18:12
-!- hoijui [~hoijui@141.23.95.39] has quit [Ping timeout: 260 seconds]18:13
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun18:15
shogun-notifier-shogun: Heiko Strathmann :master * 9759146 / src/shogun/features/streaming/StreamingDenseFeatures.cpp: https://github.com/shogun-toolbox/shogun/commit/97591469871d0de72857eb4233949db6cdee9afc18:15
shogun-notifier-shogun: re-added initialisation for current vector which prevents double free segfaults when streaming from fixed data in memory18:15
shogun-notifier-shogun: Heiko Strathmann :master * ff9a774 / src/shogun/statistics/LinearTimeMMD.cpp: https://github.com/shogun-toolbox/shogun/commit/ff9a77402fd742972d9810052d9eb996626fb91c18:15
shogun-notifier-shogun: moved SG_REF to a nicer place18:15
shogun-notifier-shogun: Heiko Strathmann :master * 253e9a6 / tests/unit/statistics/MMDKernelSelectionMax_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/253e9a6c41cb04a2367106cc839522e99ed7b9c518:15
shogun-notifier-shogun: removed prints18:15
shogun-notifier-shogun: Heiko Strathmann :master * 051da80 / src/shogun/kernel/CombinedKernel.cpp,src/shogun/kernel/CombinedKernel.h: https://github.com/shogun-toolbox/shogun/commit/051da8060260b566bef9b01b980510f0885267ec18:15
shogun-notifier-shogun: -removed KernelToCombinedKernel and replaced by a (proper) obtain_from_generic18:15
shogun-notifier-shogun: -some better error messages18:15
shogun-notifier-shogun: -added possibility to init with single features which are then applied to all subkernels (needed by MMD stuff)18:15
shogun-notifier-shogun: Heiko Strathmann :master * 76463c5 / src/shogun/statistics/LinearTimeMMD.cpp,src/shogun/statistics/LinearTimeMMD.h: https://github.com/shogun-toolbox/shogun/commit/76463c5e8621bf1adb258376b5dc4b4578f8260618:15
shogun-notifier-shogun: some minor updates18:15
shogun-notifier-shogun: Heiko Strathmann :master * c637489 / src/NEWS: https://github.com/shogun-toolbox/shogun/commit/c637489e4282ced0f3879c6961a1a18b6b59c3e218:15
shogun-notifier-shogun: mentioned Gaussian blobs generator18:15
shogun-buildbot_build #224 of rpm1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/22418:24
shogun-buildbot_build #859 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/85918:25
shogun-buildbot_build #819 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/819  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:00
-!- hoijui [~hoijui@141.23.95.39] has joined #shogun19:12
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Ping timeout: 245 seconds]19:13
shogun-buildbot_build #820 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/820  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:14
shogun-notifier-shogun: Heiko Strathmann :master * b2ecc90 / examples/undocumented/python_modular/statistics_mmd_kernel_selection.py: https://github.com/shogun-toolbox/shogun/commit/b2ecc905f37e84bfeed14e266490175f4ffbe77519:18
shogun-notifier-shogun: extended example with a combined kernel weight selection19:18
shogun-notifier-shogun: Heiko Strathmann :master * a528805 / examples/undocumented/ (2 files): https://github.com/shogun-toolbox/shogun/commit/a52880556c117d224d7b3d869299f964524486c119:18
shogun-notifier-shogun: added other kernel selection methods for illustrational purposes19:18
shogun-notifier-shogun: Heiko Strathmann :master * 41aef00 / src/shogun/statistics/MMDKernelSelectionMedian.h: https://github.com/shogun-toolbox/shogun/commit/41aef009835ee9136a63d643758b67e9fb6c2f9119:18
shogun-notifier-shogun: fixed get_name string19:18
shogun-notifier-shogun: Heiko Strathmann :master * b78b284 / examples/undocumented/python_modular/statistics_linear_time_mmd.py: https://github.com/shogun-toolbox/shogun/commit/b78b2848fadea1fa05794486b1cf68f950017cd319:18
shogun-notifier-shogun: added computation of type I and II errors19:18
shogun-notifier-shogun: Heiko Strathmann :master * 3822694 / examples/undocumented/python_modular/statistics_quadratic_time_mmd.py: https://github.com/shogun-toolbox/shogun/commit/38226942fffa1769174c1f1dfaed06ace13b033819:18
shogun-notifier-shogun: reduced number of samples to be faster19:18
shogun-notifier-shogun: Heiko Strathmann :master * 18e8181 / examples/undocumented/python_modular/statistics_mmd_kernel_selection.py: https://github.com/shogun-toolbox/shogun/commit/18e818104f502713347f6e1db75e519dc413f70b19:18
shogun-notifier-shogun: commented out all pylab stuff19:18
shogun-notifier-shogun: Heiko Strathmann :master * 5ebfc14 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/5ebfc14aad2e4f8a71dbc434d34a3e50212cbb0519:18
shogun-notifier-shogun: Merge pull request #891 from karlnapf/master19:18
shogun-notifier-shogun:19:18
shogun-notifier-shogun: some example fixed and updates19:18
shogun-buildbot_build #551 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/551  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:24
shogun-buildbot_build #552 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/55219:39
shogun-buildbot_build #822 of deb3 - modular_interfaces is complete: Failure [failed git]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/822  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:42
heikolol github is offline19:42
blackburncool19:49
heikocannot commit19:49
heikoehm push19:49
heikoand the buildbot cannot do anything19:50
blackburnheiko: yeah web doesn't work as well19:50
heikowell19:50
heikothis means:19:50
heikotime to go home :)19:51
heikoso byebys blackbird ;)19:51
heikoehm blackburn19:51
blackburnhaha19:51
blackburnhahah19:51
blackburnokay19:51
blackburnheiko: doesn't your irc client supports auto nick completion?19:51
heikowhat is that?19:52
heikoblackburn:19:52
heikooh19:52
blackburnwell in pidgin I usually write he[Tab]19:52
heikoyes :)19:52
heikoit does19:52
heikoblackburn: hello19:52
blackburnson[Tab]19:52
heikonice19:52
blackburnwi[Tab]19:52
blackburn:D19:52
heikowow that saves time with all those strange nicknames :)19:52
heikoblackburn: thanks for the tip19:52
heikoi used to type everything19:52
blackburnheiko: well at least you train in typing :D19:53
heikoyeah :)19:53
heikoblackburn: I am quite fast these days19:53
heikoblackburn: the makefile change in the unit-tests would be great to have if you have some time19:53
blackburnheiko: yeah I will have some time this weekend19:53
heikoblackburn: nice :)19:53
blackburnheiko: I am now having a luv time with openmp - trying to parallelize some things in tapkee19:54
heikoI am also close19:54
heikoyeah this is a good idea19:54
heikois openmp for multicore or also such that different processes can be started?19:54
heikofor cluster etc19:54
heikosince the kernel mmd stuff can also be parallelised19:54
heikobut makes more sense to do this massively19:54
blackburnwell mainly multicore - there exists some crazy openmp for clusters but no idea how it works19:55
heikook19:55
heikoI am not too keen on multicore parallelisations, a lot of work for not so much speedup19:55
heikounless there will be 20xcore cpus19:55
heikobut on a cluster19:55
heikothis would be cool19:55
blackburnwith openmp it should be easy but I don't get something :D19:55
blackburnI slowed down implementation - trying to understand why19:56
heikohehe dont you have a multicore machine?19:56
blackburnI think I have19:56
blackburnthere are 4 I believe19:56
heikoyeah same here19:57
heikoI have cluster access here at work, did some things on 250 processes at once19:57
heikothis is really fast :)19:57
heikoso if there is an automagic way to have this in shogun, this would be awesome19:57
heikobut difficult to program19:57
blackburnheiko: nothing automagic is possible with that I think19:57
heikomaybe I will also stick with openmp19:58
blackburnheiko: I am constantly thinking about 'workers'19:58
heikoI will soon start working with graphlab, there these things are meant to be automagic19:58
blackburnthen we could build something really scalable19:58
heikoblackburn: explain19:58
blackburnheiko: if we define some 'workers', that know what to do19:59
blackburnthey could be run both multicore and multimachine19:59
heikobut how does information exchange work?20:01
blackburnheiko: with messages20:01
heikook20:01
heikoand these are either implemented in in shared memory or via network etc20:01
blackburnheiko: yeah there are some solutions for that20:01
blackburnzeromq, etc20:01
heikopretty cool20:01
heikohowever, unrealistic to implement in shogun, we would have to change everything :D20:02
blackburnheiko: yeah..20:02
heikomaybe for new stuff20:02
heikobtw thats what the graphlab guys are intending20:02
heikoI will spy them out a bit and then tell you what its like :)20:02
blackburnheiko: yeah that's an old idea actually20:03
shogun-buildbot_build #821 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/821  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>20:03
blackburnheiko: main reason I think about that is that my job is related to multiagent systems and I get damaged by that20:03
heikoblackburn, btw we could also launch multiple processes for the tests and examples20:03
heikosince sloooow20:04
heikoI see :)20:04
heikoblackburn, I would be keen on those things20:04
blackburnheiko: I know some great library for message passing (but in java)20:04
blackburnakka that is20:04
blackburnI am very satisfied to use it20:04
blackburnheiko: but to be honest I feel totally absolutely wrong doing one-core in 201320:08
blackburnheiko: it is just like using calculators to solve differential equations20:10
-!- hoijui [~hoijui@141.23.95.39] has quit [Ping timeout: 256 seconds]20:32
wikingblackburn: mpi?20:48
wikingthat's pretty much straightforward20:48
blackburnwiking: come on that's like fortran for parallel stuff20:48
wikingwell still20:48
blackburnin a bad sense20:48
wikingit's pretty stable20:48
wikingand works very well on clusters20:48
wikingthen again of course not on desktop machines with n cores20:49
wikingit's intended for clusters i admit :P20:49
wikingbut mpi is used widely for a reason when u are dealing with c++ and multinodes20:49
blackburnyeah I know20:49
wikingbut then go with M/R setup20:50
wikingi.e. like mahout20:50
blackburnyeah everything is done actually I understand :)20:50
wikingheheh well mahout could take some total refurbishment :P20:50
wikingbut hadoop has c++ interface20:50
wikingit's pretty shit i must admit :)20:51
wikingthe people at yahoo should be killed for that one ;)20:52
blackburnheh20:52
wikingapparently they are pretty much the only left big users of that interface20:52
wikingbut it's horrible20:52
wikingtotally out of sync with the core API20:53
wikingso it pretty much calls functions there are deprecated for like 2 years now20:53
wikingand stuff like this20:53
wikingafaik we are better of with that line of opencl20:53
wikingthat's not done yet20:53
wiking;)20:53
wikingthere's graphlab and mahout for distributed machine learning20:54
wikingbut both of them pretty much are going along with multiple nodes with n core cpus20:54
wikinggpu based stuff is not much in use20:54
wikingTheano is pretty much one ML lib that uses cuda20:55
blackburnyeah20:55
wikingso that viennaCL stuff last year was a good start in that sense imo20:55
wikingbut i think we had this conversation like 2 months ago or so20:55
wiking:))))20:55
blackburnwiking: I still think something is wrong with GPU computing20:56
wikingwell you wouldn't think that when u would have a nvidia tesla in your machine ;P20:56
wikingof course the memory maintainance is pretty shit20:56
wikingmoving constantly between the actual ram and the gpu memory20:57
wikingcan be a pain in the ass20:57
blackburnwell yeah may be I wouldn't think so but something is wrong20:57
wikingmy concern is always this20:57
wikingthe bottleneck of moving the memory content back and forth20:57
wikingi mean sometimes it's not even worth to move things into the xmm registers20:57
wikingso then think about moving it to a different memory...20:58
blackburnwiking: any idea why for loop with openmp gets slower than serial? :D21:02
blackburncan't make any sense of it :(21:03
blackburnno sync at all21:04
wikingwell threading21:07
blackburnwiking: hmm changing compiler helped21:11
blackburnwiking: I admit gcc terribly sucks in everything21:11
blackburnah no21:12
blackburn:D21:12
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has joined #shogun21:17
shogun-notifier-shogun: Heiko Strathmann :master * 5a99140 / tests/unit/statistics/LinearTimeMMD_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/5a99140e0e0374159aa1a7bd1d53f8f0301d84bf21:58
shogun-notifier-shogun: added unit test for linear time MMD21:58
shogun-notifier-shogun: Heiko Strathmann :master * eb8bf03 / examples/undocumented/libshogun/statistics_linear_time_mmd.cpp: https://github.com/shogun-toolbox/shogun/commit/eb8bf033374ecf088acfccee111a36f35523123421:58
shogun-notifier-shogun: made a more illustrative example now that tests are in unit-tests21:58
shogun-notifier-shogun: Heiko Strathmann :master * 8b83c55 / / (2 files): https://github.com/shogun-toolbox/shogun/commit/8b83c55fb1ee5aef3827ccef0a2a9beaa919dbf021:58
shogun-notifier-shogun: Merge pull request #892 from karlnapf/master21:58
shogun-notifier-shogun:21:58
shogun-notifier-shogun: new tests new example21:58
blackburnheiko: crazy active today21:59
heikoblackburn :)21:59
heikoI just left for the pub when we last talked21:59
heikoand had a PR open21:59
blackburnballmer's peak?21:59
heikoand now I am back drunk :D21:59
heikosending it :)21:59
heikobut you  know, I currently get paid to work on shogun/MMD :)22:00
heikoso thats why active all day22:00
blackburnreally?22:00
heikoyes22:00
blackburnyou were doing some other things before22:00
blackburnso now you are paid shogun developer for real, hmm nice22:00
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun22:00
heikoI should rather say: to finish MMD implementation22:01
heikoI alwas try to sneak this in22:01
heikonext thing is kernel belief propagation, but its supposed to go to graphlab, so lets see22:01
heikowill go home now, have a good evening!22:01
blackburnkernel BP?22:01
blackburninteresting22:01
heikoyeah, good stuff!22:01
heikobye22:02
blackburnokay see you22:02
-!- heiko [~heiko@nat-182-132.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]22:02
blackburnfer n4nd022:02
blackburnwhat's up?22:02
n4nd0hello hello22:02
n4nd0mm I am reading you people are putting into practice Ballmer's peak :DD22:03
blackburnyeah heiko tried22:03
n4nd0blackburn: are these people Russian? http://www.youtube.com/watch?feature=player_embedded&v=XsFBuXIpdVg22:09
n4nd0ah no22:09
blackburnn4nd0: well in means of madness yes22:10
blackburn:D22:10
n4nd0they are talking Portuguesse22:10
n4nd0caralho22:10
blackburnbut not originally22:10
blackburnhah22:10
blackburnn4nd0: have you heard that: "when you are drunk and think of doing something crazy - russian already did that"22:10
n4nd0I will never learn to spell Portuguese correctly :D22:10
n4nd0haha no I didn't22:11
-!- cwidmer [8ca3fe9d@gateway/web/freenode/ip.140.163.254.157] has joined #shogun22:13
shogun-buildbot_build #555 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/555  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>22:33
shogun-buildbot_build #823 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/82322:37
-!- FSCV [~FSCV@187.210.54.166] has joined #shogun23:03
-!- FSCV_ [~FSCV@187.210.54.166] has quit [Read error: No route to host]23:03
-!- cwidmer [8ca3fe9d@gateway/web/freenode/ip.140.163.254.157] has quit [Quit: Page closed]23:08
-!- FSCV [~FSCV@187.210.54.166] has quit [Read error: No route to host]23:10
-!- FSCV [~FSCV@187.210.54.166] has joined #shogun23:10
-!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has quit [Ping timeout: 244 seconds]23:41
--- Log closed Sat Feb 23 00:00:49 2013