Open in new window / Try shogun cloud
--- Log opened Thu Feb 14 00:00:46 2013
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]00:06
-!- cwidmer [8ca3fe9d@gateway/web/freenode/ip.140.163.254.157] has quit [Quit: Page closed]01:03
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]01:19
-!- FSCV_ [~FSCV@173.254.212.46] has quit [Quit: Leaving]01:49
-!- FSCV [~FSCV@189.139.160.181] has joined #shogun04:09
-!- FSCV_ [~FSCV@204.45.110.163] has joined #shogun04:11
-!- FSCV [~FSCV@189.139.160.181] has quit [Ping timeout: 240 seconds]04:14
-!- FSCV_ [~FSCV@204.45.110.163] has quit [Quit: Leaving]04:17
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun04:47
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has joined #shogun05:32
-!- sumit_ [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun07:18
-!- sumit_ [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]08:02
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun08:07
-!- sumit_ [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun08:21
sumit_Hey blackburn,Haven't talked to you for long.Denoising Auto-encoders/Stack Denoising Auto-Encoders will be the next to work upon|However,The devotion/work fervour slows down a bit as the midterm exams are nearing! :P08:26
sumit_Anyways I shall get back to you,This Time Versed with The workflows at shogun :)08:26
sumit_Thanks  :)08:26
blackburnsumit_: wait wait no need to implement it one by one without integration08:28
blackburnit would be better if you finish rbm and integrate it - I can't use your standalone code anyhow08:29
@sonney2kblackburn, we have an issue with de-serializing sparse features08:34
@sonney2keverything works ok08:34
@sonney2kbut the inplace constructor & sgvector make it crash once the object is SG_UNREF'd08:34
blackburnsonney2k: is that the issue I commented I can't reproduce it?08:34
@sonney2kblackburn, yes because you forgot to destroy the object08:35
blackburnah08:35
@sonney2kblackburn, but i've put in a unit test now08:35
blackburnI see08:35
blackburnsonney2k: so it is a unit test now and it fails?08:35
blackburnah reading yesterday logs08:36
@sonney2kblackburn, yes but I commented out the failure08:36
@sonney2kgtg08:36
@sonney2kcu08:36
blackburnokay then I'll take a look once I get to that08:38
blackburnsonney2k: sonne|work as of SO idea - feel free to suggest better naming :)08:38
sumit_Oh Sorry I had to go somewhere Blackburn, All Right , I shall focus on RBM first , render it useful and proceed further :)08:42
blackburnalright08:43
sumit_:)08:43
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun09:34
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 244 seconds]10:08
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun10:27
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]10:29
wikingbtw: we should as well use mocking macros since we have them: e.g. EXPECT_CALL11:04
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Ping timeout: 246 seconds]11:32
wikingsonney niiice hostname :)11:34
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun11:46
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun12:29
blackburnwiking: isn't my hostname nice? :D13:13
blackburnI want nice hostname too!13:13
wiking:>13:26
wikingkgru13:26
wikingkenguruu13:26
wiking:>>13:26
blackburnwiking: heh, 'knowledge genesis'13:27
-!- sumit_ [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]13:33
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]13:57
-!- FSCV [~FSCV@204.45.110.163] has joined #shogun14:19
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun14:36
wikingblackburn: here?15:07
blackburnwiking: ja15:10
wikingblackburn: have u manage to get clang_complete work under tests/unit ?15:13
wikingfor some reason it does not work for me15:13
blackburnwiking: hmm I didn't try15:14
blackburnwiking: what does clist show?15:14
wikingblackburn: full vim command?15:15
blackburnwiking: :clist15:16
wikingE42: no errors ?15:17
blackburnhmm15:19
wikingblackburn: have u seen this: https://github.com/exclipy/clang_indexer15:28
wikingbut i'm missing Index.h :)15:28
blackburnhmm15:31
blackburnwiking: ctags works ok I think though15:32
wikinglolface15:34
wikingthis is some heavy shit: http://www.sublimetext.com/15:34
blackburnit is always told sublime is cool but I didn't try yet :)15:36
wikingi'm just checking it out15:38
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]15:43
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:12
-!- blackburn [~blackburn@188.168.128.140] has joined #shogun16:49
-!- heiko [~heiko@027d8b72.bb.sky.com] has joined #shogun17:19
-!- blackburn [~blackburn@188.168.128.140] has quit [Quit: Leaving.]17:20
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has joined #shogun17:33
KMcQuistenHello all17:33
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Quit: Page closed]17:53
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun18:04
shogun-notifier-shogun: Heiko Strathmann :master * e4b0d2b / tests/unit/features/CGaussianBlobsDataGenerator_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/e4b0d2b1e4130780ecc50a555ac2f14e2a0389ee18:04
shogun-notifier-shogun: my first unit-test to ensure that Gaussian blobs data is produced roughly the right way18:04
shogun-notifier-shogun: Heiko Strathmann :master * 09c467a / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/09c467a5694e89d9a3bcb0e9f3f0ff5742de3ffd18:04
shogun-notifier-shogun: examples work again so use them18:04
shogun-notifier-shogun: Heiko Strathmann :master * 5903252 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/59032523db4e58f95b32fe93449b999f7e7fcea318:04
shogun-notifier-shogun: Merge pull request #876 from karlnapf/master18:04
shogun-notifier-shogun:18:04
shogun-notifier-shogun: my first unit test18:04
heikowiking, ^18:04
heikofinally :)18:04
shogun-buildbot_build #822 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/822  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>18:08
-!- heiko [~heiko@027d8b72.bb.sky.com] has quit [Quit: Leaving.]18:09
-!- blackburn [~blackburn@188.168.128.140] has joined #shogun18:12
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has quit [Ping timeout: 245 seconds]18:45
@sonney2kblackburn, I start to miss some features of eclipse in vim :D19:24
@sonney2ktoo much java coding I guess :/19:24
blackburnsonney2k: you should try intellij19:24
@sonney2kblackburn, b'cause it's made by Russians?19:25
blackburnsonney2k: no, best java ide19:25
@sonney2kblackburn, why - what is better wrt eclipse19:25
blackburnsonney2k: autocompletion19:26
blackburn(at least)19:26
@sonney2kI don't see what could be improved19:26
blackburnmuch faster and you start to think it knows better what you want to write19:26
@sonney2kI mean what can it autocomplete or in a better way thatn eclipse?19:26
blackburnwhat I don't like - eclipse autocomplete is by demand19:27
@sonney2kyou mean intellij *always* autocompletes?19:27
blackburnsonney2k: most of time19:27
blackburnwhen you write19:28
blackburnMessage <your cursor here>19:28
blackburnit suggests message19:28
blackburnand so on19:28
blackburnsonney2k: if you don't trust me - ask mikio :)19:28
@sonney2kahh ok that eclipse cannot do but in eclipse you would write new Message() and then it would suggest Message message19:29
@sonney2kas variable19:29
@sonney2kit is not about trust :)19:29
@sonney2kI cannot change anyways but I am tempted to use the refactoring features w/i shogun19:29
@sonney2kdid anyone try to use shogun & eclipse?19:29
blackburnsonney2k: oh it won't work in idea then :)19:30
blackburnsonney2k: well I tried19:31
@sonney2kintellj lost then :)19:31
blackburnI've been using eclipse as an IDE first two months of shoguning or so19:31
@sonney2kI would rather code with one ui for c++ / java / python19:31
@sonney2kblackburn, so tell me how to do it?19:31
blackburnsonney2k: do what?19:31
@sonney2kuse eclipse for shogun19:32
@sonney2kI am totally clueless19:32
blackburnsonney2k: hmm I am too now actually19:32
blackburnsonney2k: so you need refactoring?19:33
@sonney2kblackburn, what I like about eclipse is to see all functions/classes etc19:33
blackburnsonney2k: yeah it is possible in vim as well though19:34
blackburnwith some headache just like everything in linux19:34
@sonney2kahh it can use an existing makefile project19:35
blackburnsonney2k: yes I think so19:35
@sonney2kI am just doing this19:35
@sonney2kquestion is whether I can only use this for the src/shogun dir or src/19:35
blackburnsonney2k: if we get shogun to cmake we could generate projects for everything19:36
blackburneven for visual scheisse 201019:36
blackburn:D19:36
blackburnsonney2k: well to navigate you can use any folder I think19:38
blackburnmakefile would be used only for 'build project'19:38
blackburnsonney2k: works quite ok but so slow comparing to vim19:41
@sonney2kblackburn, well I would want to use it completely obviously19:41
blackburnsonney2k: I feel cumbersome in eclipse :D19:42
@sonney2kme too19:43
@sonney2kwhat sucks is that you have to give it all INCLUDES19:44
@sonney2kmanually somehow :/19:49
wikingmmm this sublimetext is actually good19:50
wikingbut i miss some vim shortcuts :)19:50
@sonney2kuhh making progress20:00
@sonney2kblackburn, does eclipse have a git plugin?20:04
@sonney2kprobably woa?20:04
blackburnsonney2k: yes it does IIRC20:04
shogun-notifier-shogun: Soeren Sonnenburg :master * ac7aa2a / src/shogun/classifier/mkl/MKL.h: https://github.com/shogun-toolbox/shogun/commit/ac7aa2a4d0da9576f5e90a821d1231acba7d009820:10
shogun-notifier-shogun: fix compile error / forgotten glpk include20:10
@sonney2kok I will give eclipse a try20:11
@sonney2klets see whether it behaves20:11
@sonney2kblackburn, you don't happen to know how we get serializaton to work with SGVector stuff right?20:12
@sonney2kI am asking because of the inplace issue...20:12
@sonney2k*hate*20:13
shogun-buildbot_build #823 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/823  blamelist: Soeren Sonnenburg <sonne@debian.org>20:13
blackburnsonney2k: no, I don't - what is the problem whough?20:14
blackburnth*20:14
@sonney2kblackburn, wasn't that even the problem when you attempted to convert SGString* stuff20:14
blackburnsonney2k: inplace? yeah it was a problem20:15
@sonney2kblackburn, that we when we de-serialize allocate some void* of SGVectors20:15
blackburnahh20:15
blackburnargh hate20:15
@sonney2kyeah as nice as this SGVector stuff is as tough it is with serialization20:16
@sonney2kit needs an extra treatment in the (de)-serialization code20:16
@sonney2kheiko is the expert there ...20:17
blackburnsonney2k: any guidance to line of code?20:21
@sonney2kparameters.cpp IIRC20:23
shogun-notifier-shogun: Soeren Sonnenburg :master * 59719ba / tests/unit/classifier/svm/SVMOcas_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/59719ba652c35ac206feb39945e262b83f7663f620:24
shogun-notifier-shogun: require single thread usage & initialize rnd number generator20:24
shogun-buildbot_build #824 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/82420:28
blackburnsonney2k: you should like it http://www.tldrlegal.com/20:37
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun20:42
-!- heiko [~heiko@027d8b72.bb.sky.com] has joined #shogun21:10
@sonney2kblackburn, I think I know how to fix it - just add SGSparseVector to lib/memory.h21:21
@sonney2klet me try21:21
n4nd0blackburn: around?21:23
n4nd0sonney2k: around? :)21:28
@sonney2kn4nd0, ?21:31
n4nd0I think it would be very convenient to introduce cmake in shogun21:31
@sonney2khaha21:31
n4nd0blackburn commented something before related to eclipse projects21:31
n4nd0but also, I think it could handle very well the fact of selecting to build different parts of shogun21:32
n4nd0I mean sometimes we don't need to rebuild everything for example when we are developing a new part21:32
n4nd0sonney2k: haha why do you think it is funny? :D21:32
@sonney2kn4nd0, not so sure what you mean but I have no strong preferences21:33
n4nd0it would be something pretty big to do probably21:33
@sonney2kn4nd0, b'cause blackburn was saying this too me21:33
@sonney2kand I said it is a massive undertaking21:33
@sonney2kso nothing we can do before gsoc starts21:33
n4nd0so I thought that we could include as part of one of GSoC projects21:33
n4nd0yeah sure, but what about *during* GSoC?21:33
n4nd0one of the projets related to usability improvements21:34
@sonney2kI think this should be done by some core developer though21:34
n4nd0sonney2k: why?21:34
@sonney2kbut I don't see the benefits for myself21:34
n4nd0but anyway a student is a core develooper, at least during the summer of his/her project21:34
@sonney2kit involves all parts of the code and we have lots of small customized tests21:34
n4nd0mmm I think it is still possible to do it with the tests21:35
n4nd0and I think there are big benefits21:36
n4nd0it becomes easier to handle new dependencies than modifying the current configure script21:36
n4nd0and what I meant before with selecting what parts of the code to build21:38
n4nd0say we are introducing a new preprocessor21:38
n4nd0why to build all the kernels, classifiers, etc?21:38
@sonney2kn4nd0, not sure if you said sth - I ran out of battery22:27
n4nd0sonney2k: check22:33
n4nd021:35 < n4nd0> but anyway a student is a core develooper, at least during the summer of his/her project22:33
n4nd021:35 <@sonney2k> it involves all parts of the code and we have lots of small customized tests22:33
n4nd021:36 < n4nd0> mmm I think it is still possible to do it with the tests22:33
n4nd021:37 < n4nd0> and I think there are big benefits22:33
n4nd021:37 < n4nd0> it becomes easier to handle new dependencies than modifying the current configure script22:33
n4nd021:39 < n4nd0> and what I meant before with selecting what parts of the code to build22:33
n4nd021:39 < n4nd0> say we are introducing a new preprocessor22:33
n4nd021:39 < n4nd0> why to build all the kernels, classifiers, etc?22:33
wikingmy 2 cents: cmake transition should be fairly easy. we don't really use any very very funky stuff in ./configure script22:39
wikingi mean there are far more bigger libraries that uses cmake22:41
n4nd0we could use their CMakeLists.txt as inspiration22:42
wikingand a lot of .cmake file exists already for various library tests, moreover some libraries already provide a cmake file as well22:42
wikingwho's?22:42
wikingah i was meaning itk and vtk22:42
wikingthey are big libraries22:42
wikingas well as opencv22:42
n4nd0and pcl22:43
wikingopencv really has some funky shit in it22:43
wikingand most of the libraries that we use are used by opencv as well22:43
wikingbut yeah afaik it's just CMakeList.txt + a cmake folder with all the funky finder scripts22:44
n4nd0it looks like that22:44
wikingbut i think we should not waste a gsoc slot for this task at all22:44
wikingit's far too simple imho22:44
n4nd0I didn't mean a whole slot for this22:44
@sonney2kI don't think it is simple.22:44
@sonney2kbut if someone has time to do it feel free22:45
n4nd0I think my opinion is a bit in between yours22:45
wikingwell yeah we already have some cmakefile.txt22:45
wikingthat's a good start22:45
n4nd0I don't think it is a piece of cake like wiking does but I think it should be doable in max two weeks of work for a gsocer22:45
n4nd0:)22:45
wikingn4nd0: max 2 days22:45
@sonney2kand I think 1-2 months22:46
n4nd0wiking: pi rule22:46
wikingand the 2nd day is testing :)22:46
n4nd0say one week at least22:46
wikingnop22:46
wikingwho cannot do this in 2 days22:46
wikingdoes not deserve to be a shogun developer22:46
n4nd0haha c'mon... say we don't get any cmake expert22:46
wiking;DDDD22:46
n4nd0:DDD22:46
wikingheheh yeah i was a bit ironic here22:46
wikingbut yeah i reckon that could be done within 2 weeks22:46
n4nd0sonney2k: why do you think it would be soo much work?22:47
wikingn4nd0: well we are already wasting time!22:47
wikingout of those hours22:47
wikinginstead of talking about it, let's just fucking doit22:47
wiking:)))22:47
@sonney2kn4nd0, it is a lot of very non-standard stuff22:47
n4nd0wiking: this is management time :)22:48
wikingpffff22:48
wikingpetting time22:48
wikingit's just <nop> <nop>...<nop>22:48
wiking:D22:48
n4nd0hehe22:48
n4nd0sonney2k: what non-standard stuff?22:48
@sonney2kn4nd0, you have to integrate gcc / g++ all the lib detection, optimization settings, swig, doxygen, version header and dependency generation22:49
@sonney2kand that for java,lua,ruby,csharp,python,r,octave ...22:50
n4nd0sonney2k: that is actually the most difficult thing I see22:50
n4nd0I am unsure how would it be together with SWIG22:50
wikingn4nd0: http://blog.cryptomilk.org/2011/01/15/cmake-java-support/22:50
n4nd0and its bindings22:50
n4nd0I will look around if I see any other project using swig + cmake22:51
wikingn4nd0: http://www.swig.org/Doc1.3/Introduction.html#Introduction_build_system22:52
n4nd0I think that the rest of the stuff would be very doable22:52
n4nd0wiking: aham, that's nice22:52
n4nd0wiking: do you also think that it would be positive to introduce cmake?22:52
@sonney2kn4nd0, in any case - what is much more important is to get shogun 2.1 out22:52
@sonney2kbefore gsoc students start to do PRs22:53
n4nd0sonney2k: agree22:53
wikingn4nd0: well since cmake is getting more and more supported by various libs22:53
wikingit does make life easier22:53
n4nd0so I understand that we should try to get valgrind errors as close to zero as possible22:53
wikingbut honestly22:53
n4nd0sonney2k: what else?22:53
wikingwhat cmake can do an autoconf + automake script can do as well22:54
@sonney2kn4nd0, all the warnings down to 022:54
@sonney2klook at the build bots22:54
n4nd0sure the warnings22:54
@sonney2kdo some tests, update README22:54
@sonney2kand bugs down if possible22:54
wikingi'm not soo much fan of current ./configure script (the mplayer style) + but it works... i prefer the configure.ac stuff.... but yeah it works now, and it's not soooo hard to add extra stuff into that configure script, only thing is that it's going to grow and grow and grow with new stuff, and it's really not easy to make it modular, like cmake... that's kind of like the only positive thing i think about cmake. that it can be broken down into smaller pi22:56
wikingbut imho cmake is more of an eye-candy22:57
wikingi mean i tried cross-compiling with the current setup shogun, and it's possible22:57
n4nd0sonney2k: can you remind me how do you do from vim to compile so it breaks every time a warning is found?22:58
n4nd0I think you mentioned that once22:58
wikingi.e. with this aspect we wouldn't benefit too much from cmake - the most advertised thing about cmake is that makes your project pretty much cross-platform22:58
n4nd0wiking: that is a very good advantage - cross-platform22:59
wikinglol: http://mjtsai.com/blog/2013/02/14/adobe-photoshop-1-0-1-source-code/22:59
wikingn4nd0: but that one pretty much works with current setup, except native windows :D22:59
@sonney2kn4nd0, shogun's code is not cross platform23:00
n4nd0ok so I doesn't make matter that much to use cmake in this aspect :)23:00
@sonney2kwiking, is that a joke?23:01
wikingn4nd0: well as said it's more about have a very clean/sterile build system23:01
wikingsonney2k: what?23:01
n4nd0haha23:01
n4nd0the adobe thing I guess23:01
wikingsonney2k: no i mean it's 1.0 :D23:01
@sonney2kyeah but pascal on 68k?23:01
n4nd0yeah but I can feel some irony too23:01
wikingi think this is kind of like the version they created for star wars23:02
n4nd0uncommented but well structured code!23:02
n4nd0it seems to be very clean code hehe23:02
wikingbtw i hope u know that the reason we have adobe photoshop in the first place is star wars23:02
wiking:23:02
wiking:P23:02
wikingthey developed the first version of adobe photoshop for that :)23:02
wikingpretty funny stuff :>23:02
n4nd0I didn't know that23:03
wiking"Then his brother, John Knoll, who was in charge of the special effects department for Industrial Light and Magic on the 1st Star Wars movie showed interest in the project and recommended that it be developed into a full image editing program. Hence it was named 'Display'. The Knoll brothers then renamed Display to Image-Pro. Image-Pro could be used for rack card printing too and that wasn't the only thing it could do."23:05
wiking"Image-Pro had many advanced features for displaying images so in the year 1988, Knoll brothers started to sell Image-Pro commercially. And for this purpose they renamed Image-Pro to "Photoshop".23:05
wikingsonney2k: yeah i mean 68k is because of mac23:06
wikingi guess23:06
@sonney2kwiking, yeah but not C?23:06
wikingsonney2k: mmm mac:))23:07
wikingand come oooon23:07
wikingit's done by a phd student23:07
blackburnre!23:07
wikingi mean by definition he had to use a funky fucking language to write it :P23:07
wikingas he was doing his phd :D23:07
wikingand back in 80s pascal was like the first language everybody learned @ uni for programming23:08
wikingi even had to fucking learn that language23:08
wiking:=23:08
wiking:= for defintion :D23:08
wikingthe only more funkier stuff i ever saw was delphi23:09
wiking :D23:09
blackburnstudents here still learning it as a first language23:09
n4nd0blackburn: join the cmake discussion! :D23:09
blackburnno joke, they are really doing that23:10
wikingblackburn: :>23:10
blackburnn4nd0: I have to try that and check23:10
blackburnI don't know23:10
wikingblackburn: indeed23:10
wikingi mean it's a clean syntax23:10
wikingbut ugly23:10
wiking:DD23:10
n4nd0wiking: in my Spanish uni it is still the first language...23:10
@sonney2kyou guys just do the cmake thing and once it works I don't mind using it - I don't have strong preferences23:10
wikingprocedure cmake23:11
wiking:D23:11
blackburnsoeren is a old jedi he could be right we could get troubles23:11
n4nd0yes23:11
n4nd0the fact that he estimates 2 months to make it work scares me23:11
blackburnI don't know - for student - maybe yes23:12
wikingn4nd0: and then on the end we will cmake shogun for android ;)23:12
blackburnfor me it is not23:12
blackburnshogun on android could be kind of funny, don't know how useful though23:13
wikingblackburn: well if we could separete the code easily from learning and classifying23:13
wikingthen actually it is interesting23:13
blackburnwiking: could you elaborate what to separate?23:13
wikingi mean just to compile the part that can read a learned model and do classification23:14
blackburnahh23:14
wikingblackburn: like for example a video codec23:14
wikingnormally u will only need a decoder23:14
n4nd0blackburn: who doesn't to apply SVMs to the pictures taken with the phone camera?23:14
wikingon a handheld device23:14
wikingbut encoder is not necessarily required23:14
blackburnn4nd0: ah linear svm is easy to write - why to bother with library23:14
wikingblackburn: well as i showed u earlier23:15
wikingios has liblinear and libsvm23:15
n4nd0blackburn: jk :)23:15
blackburnwiking: yeah, I was surprised23:15
wikingso for some reason they use it23:15
wikingas the license is there23:15
wikingfor what i have no clue23:15
blackburnliblinear and libsvm are great mature tools23:15
wikingi guess with a jailbroken ios23:15
wikingone could easily find out where those libs are used23:16
blackburnwiking: I tired to google but didn't manage to discover anything Iirc23:16
wikingbut afaik they would never go with shogun23:16
blackburniirc*23:16
wikingas it's gplv323:16
wiking:D23:16
blackburnhaha righto, gpl is tough23:16
wikingwell lgpl ;)23:17
blackburnwiking: do you know that blog? http://nuit-blanche.blogspot.com/23:17
wikingdunno about andro practices23:17
wikingif they go with gplv3 projects23:18
wikingblackburn: nou haven't seen it b423:18
blackburnwiking: I will be trying to convince and then support the author being a mentor23:18
wikingcool23:18
blackburnhe knows (some mature words sonney2k denies me to say) a lot23:18
blackburnabout compressive sensing23:19
wiking!@#$23:19
wiking:D23:19
blackburnand matrix factorization23:19
@sonney2kblackburn, I am off to bed - so you can switch to free speech mode23:19
blackburnsonney2k: zaebis!23:20
blackburnsonney2k: thanks! :D23:20
blackburnsonney2k: you tried to go to bed a few times before as I can see ;)23:20
@sonney2kblackburn, no I didn't spent the evening talking to some lawyer :/23:21
blackburnsonney2k: so you have stolen that car? I told you not to do that23:21
@sonney2kblackburn, you must have forgotten :)23:21
@sonney2kl8r23:21
n4nd0sonney2k: good night23:22
wikinggnite'23:23
wikingblackburn: time series analysis would be cool to have as well in shogun23:23
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]23:24
blackburnwiking: still pursuing quant stuff? :)23:26
blackburnn4nd0: main problem with any phones integration (as I see it) - we don't integrate to pipelines nicely23:30
blackburnimage -> features -> labels23:30
n4nd0mmm I am not sure what you mean, sorry23:30
blackburnn4nd0: we can't compute any image features23:32
blackburnn4nd0: then may be we can integrate with opencv/etc? still no23:32
n4nd0blackburn: aha! I see what you mean23:32
blackburnit is a general problem - how to design such api that would be so flexible23:33
n4nd0blackburn: how have you done when you used shogun for CV problems?23:33
n4nd0e.g. you road signs classification23:33
blackburnn4nd0: just computed features and then feed it to shogun23:33
blackburnwon't work with any embedding23:34
blackburnbut actually may be with trained svm it is not a problem23:34
blackburnwiking: n4nd0: do you happen to know akka library?23:37
blackburnabout it I meant23:38
n4nd0blackburn: you mentioned it not so long ago right?23:38
blackburnn4nd0: I do not remember :D23:38
n4nd0because I remember browsing this page some days ago23:38
n4nd0hehe23:38
n4nd0aah no I think I checked when I read a paper about FastSLAM with map/reduce23:39
blackburnit is really hot23:39
blackburnsomewhat 2-3 times slower than erlang23:39
n4nd0hot because it is slower? mmm23:40
blackburnhah23:40
blackburnno, because any other approach would be 200-300 times slower may be :)23:40
blackburnlike idiot way: actors made of threads23:40
blackburnn4nd0: actually tests differ and some say it is even faster than erlang23:41
n4nd0why is erlang the mark to beat?23:42
blackburnn4nd0: because it is the best actor-based thing23:42
n4nd0ahm I didn't know it is so appropriate for large scale systems23:42
blackburnn4nd0: I am talking about message passing only23:42
blackburnnot math or methods call, etc23:43
blackburnn4nd0: yes it is useful for large-scale (like some game servers for example) systems23:43
blackburnand I am using it for multiagent system I am implementing23:44
n4nd0facebook chat is done in erlang I think23:44
n4nd0trains?23:44
blackburnyeah trains23:44
blackburnn4nd0: actor-like systems are *so* easy to understand and design sometimes23:51
blackburnyou just describe messages and reactions23:51
blackburnloose coupling, bla bla :)23:51
n4nd0can you put me an example of a message and a reaction?23:55
--- Log closed Fri Feb 15 00:00:45 2013