--- Log opened Tue Jun 11 00:00:40 2013 | ||
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:28 | |
-!- FSCV [~FSCV@187.210.54.166] has quit [Quit: Leaving] | 02:30 | |
-!- lambday [67157d36@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.54] has joined #shogun | 02:46 | |
-!- nube [~rho@49.244.74.29] has quit [Quit: Leaving.] | 03:59 | |
shogun-buildbot | build #424 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/424 | 04:08 |
---|---|---|
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 255 seconds] | 04:42 | |
-!- nube [~rho@116.90.239.13] has joined #shogun | 05:51 | |
-!- nube [~rho@116.90.239.13] has quit [Ping timeout: 252 seconds] | 06:19 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 06:20 | |
-!- nube [~rho@116.90.239.3] has quit [Client Quit] | 06:21 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 06:21 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 06:30 | |
lambday | sonney2k: sonne|work: moin :) | 07:47 |
lambday | HeikoS: linear operator and dense matrix operator added, unit-tested, valgrinded | 07:48 |
-!- lambday [67157d36@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.54] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 07:57 | |
-!- blafasel [555a4c82@gateway/web/cgi-irc/kiwiirc.com/ip.85.90.76.130] has joined #shogun | 08:27 | |
-!- blafasel [555a4c82@gateway/web/cgi-irc/kiwiirc.com/ip.85.90.76.130] has quit [Client Quit] | 08:28 | |
-!- lisitsyn [~lisitsyn@83.234.54.170] has joined #shogun | 08:36 | |
-!- lisitsyn [~lisitsyn@83.234.54.170] has quit [Quit: Leaving.] | 08:51 | |
-!- lambday [67157d36@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.54] has joined #shogun | 09:08 | |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 09:18 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 09:18 | |
sonne|work | lambday: moin moin :) | 09:25 |
sonne|work | lambday: btw that is the name of a wiki | 09:25 |
sonne|work | http://moinmo.in/ | 09:25 |
sonne|work | lambday, HeikoS I will likely merge the sparse branch tonight | 09:25 |
lambday | sonne|work: wiki! :P | 09:26 |
lambday | sonne|work: you mean sparse matrix/vector related things? | 09:27 |
lambday | sonne|work: by the way, when I tried to run make shogun on my laptop (an ancient one, 1GB ram, intel core2 duo) it gave this "SSE2 not enabled" error.. earlier it used to work now after I updated it doesn't :( | 09:29 |
lambday | (I work on my lab system though, won't harm the work) | 09:30 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 09:31 | |
-!- mode/#shogun [+o lisitsyn] by ChanServ | 09:31 | |
lambday | lisitsyn: morning :) | 09:31 |
@lisitsyn | lambday: hey | 09:31 |
* lambday is in a state of trance | 09:31 | |
lambday | :) | 09:31 |
@lisitsyn | lambday: day already here ;) | 09:31 |
lambday | haha | 09:31 |
@lisitsyn | oh damn I am going to have a dense day | 09:32 |
@lisitsyn | debug time | 09:32 |
lambday | me too :( have this meeting in the afternoon with prof | 09:32 |
lambday | my thesis topic decided.. sequence kernel related stuffs.. gotta have to read papers on that | 09:33 |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds] | 09:44 | |
sonne|work | lambday: which kernels exactly? | 09:46 |
sonne|work | lambday: btw maybe you didn't do a configure on your old machine but used the config from some other machine? | 09:47 |
sonne|work | lambday: and yes sparsematrix/vector thigns :D | 09:47 |
lambday | sonne|work: I'm not sure yet.. there is this paper he gave me to read... haven't gone through yet.. http://www.cs.nyu.edu/~mohri/pub/lsk.pdf | 09:49 |
lambday | sonne|work: no I configured, that's why its even more confusing :-/ | 09:50 |
sonne|work | lambday: yeah I know these kernels | 09:51 |
lambday | I'll again check it... may be I'm just doing something stupid | 09:51 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 09:51 | |
lambday | sonne|work: ah, be prepared to get a LOT of questions from me :P | 09:51 |
lambday | awesome!! :D | 09:51 |
sonne|work | though I don't think they are relevant in practise :/ | 09:52 |
lambday | sonne|work: :( | 09:52 |
sonne|work | these transducer based string kernels are very slow | 09:52 |
sonne|work | it is a nice framework though | 09:52 |
lambday | actually they are doing something hybrid I guess... this one and some asymetric cost function.. something something | 09:53 |
lambday | I'll let you know | 09:53 |
-!- nube1 [~rho@116.90.239.13] has joined #shogun | 10:05 | |
-!- nube [~rho@116.90.239.3] has quit [Read error: Connection reset by peer] | 10:06 | |
sonne|work | lambday: they do multiple kernel learning on top of that yes | 10:11 |
@lisitsyn | sonne|work: hahah is there any algorithm you don't call MKL? | 10:15 |
sonne|work | lisitsyn: whats wrong? this is the true mkl algorithm they use | 10:16 |
sonne|work | these guys published a couple of mkl papers and it is only reasonable (like we did) to use their string kernel framework as example | 10:17 |
@lisitsyn | sonne|work: no that's just funny coincidence | 10:22 |
@lisitsyn | I mean most of the times you talk about ML it is MKL ;) | 10:22 |
sonne|work | well I read all those MKL papers back then and did lots of MKL algorithms/experiments | 10:22 |
sonne|work | so what do you expect? | 10:23 |
@lisitsyn | sonne|work: and wrote some too | 10:24 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun | 11:04 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 11:05 | |
@iglesiasg | hello people | 11:05 |
@lisitsyn | iglesiasg: hey there | 11:05 |
@iglesiasg | lisitsyn: how is it going? | 11:05 |
@lisitsyn | iglesiasg: fine, and you? | 11:06 |
@iglesiasg | lisitsyn: fine too, getting way too warm around here! | 11:06 |
@lisitsyn | iglesiasg: hehe | 11:06 |
@iglesiasg | lisitsyn: I think I could be doing better in Russia :P | 11:07 |
@iglesiasg | http://25.media.tumblr.com/6a3630f152bd3d00d23dde888706fe55/tumblr_mo0gw9HtPU1rnyeudo1_500.jpg | 11:07 |
@lisitsyn | iglesiasg: not too warm here ;) | 11:07 |
-!- nube1 [~rho@116.90.239.13] has quit [Quit: Leaving.] | 11:50 | |
-!- lambday [67157d36@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.54] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 12:14 | |
-!- HeikoS1 [~heiko@nat-169-135.internal.eduroam.ucl.ac.uk] has joined #shogun | 12:37 | |
HeikoS1 | lisitsyn: hi! | 12:40 |
HeikoS1 | wiking: hi! | 12:40 |
HeikoS1 | any news on the stuff? | 12:40 |
@lisitsyn | HeikoS1: hey | 12:41 |
@lisitsyn | it seems I have dropped | 12:41 |
@wiking | HeikoS1: which stuff? the autogenerate stuff? | 12:41 |
@lisitsyn | clone usages | 12:41 |
HeikoS1 | lisitsyn: nice thanks! | 12:41 |
HeikoS1 | wiking: yes | 12:41 |
@wiking | HeikoS1: i have one straight solution | 12:42 |
@wiking | HeikoS1: i.e. have a python script that generates the cpp file | 12:42 |
@wiking | but not based on a template file | 12:42 |
@wiking | because now to write a fucking python script that processes a template file and then generates based on that a cpp file | 12:42 |
@wiking | is just like overkill | 12:42 |
@wiking | i mean then we should define a new template file syntax etc etc etc | 12:43 |
@wiking | and i would really not like to do it | 12:43 |
HeikoS1 | wiking: yes sure, do that | 12:43 |
HeikoS1 | just a script that generates the file | 12:44 |
HeikoS1 | no template stuff | 12:44 |
HeikoS1 | but do the file in such way that we can change this later maybe | 12:44 |
HeikoS1 | i.e. other unit-tests which use *all* classes | 12:44 |
HeikoS1 | or just make a very simple template | 12:44 |
HeikoS1 | all occurences of " __CLASSNAME__" are replaced by all class names | 12:45 |
HeikoS1 | I think I would prefer that | 12:45 |
@wiking | yesh i did something like that | 12:45 |
HeikoS1 | then its very easy to create the template file | 12:45 |
HeikoS1 | and I can change the tests more importantly | 12:45 |
HeikoS1 | lisitsyn: did you push that to develop? still many clones around on my just rebased fork | 12:48 |
@lisitsyn | HeikoS1: that's sgvector clones | 12:48 |
HeikoS1 | lisitsyn: I see | 12:48 |
@lisitsyn | HeikoS1: I don't know if it should be called other way | 12:49 |
HeikoS1 | lisitsyn: no I think its fine | 12:49 |
HeikoS1 | since thats what it dies right? | 12:50 |
@lisitsyn | HeikoS1: yes just clashes | 12:50 |
@lisitsyn | with clone of sgobject | 12:51 |
HeikoS1 | lisitsyn: is that a problem? | 12:51 |
@lisitsyn | HeikoS1: yes and no | 12:53 |
@lisitsyn | :) | 12:53 |
@lisitsyn | I mean clashes are bad but unified is good | 12:54 |
HeikoS1 | I dont really see the clash | 12:54 |
HeikoS1 | do you mean two different objects have a method with the same name? | 12:54 |
HeikoS1 | but if the method does the same things, thats fine right? | 12:54 |
HeikoS1 | or is it swig-related? | 12:54 |
HeikoS1 | lisitsyn: cool merged the code, seems to not break stuff :) | 12:56 |
@lisitsyn | HeikoS1: logical clash | 12:56 |
@lisitsyn | not like issue | 12:56 |
@lisitsyn | just consitency of api | 12:56 |
HeikoS1 | I see | 12:57 |
HeikoS1 | well I can live with that :) | 12:57 |
HeikoS1 | wiking: let me know once this works, I will test/push immediately then | 13:02 |
@wiking | blah | 13:02 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Quit: Page closed] | 13:13 | |
HeikoS1 | lisitsyn: just dicovered c++'s auto | 13:16 |
HeikoS1 | thats nice | 13:16 |
sonne|work | HeikoS1: please don't use auto yet though | 13:27 |
HeikoS1 | sonne|work: okay, didnt plan to anyway | 13:27 |
HeikoS1 | just saw it | 13:27 |
HeikoS1 | sonney2k: it is quite useful for iterating STL types though | 13:30 |
HeikoS1 | but I never use those in shogun | 13:30 |
thoralf | type inference for c++? | 13:34 |
HeikoS1 | thoralf: yes | 13:34 |
thoralf | Cool. Ever tried? Any drawbacks? | 13:34 |
thoralf | c++1x is c++11 now - must have missed that. | 13:35 |
thoralf | What happens when using "auto" variables in printf? ;) | 13:36 |
sonne|work | I don't know why you guys are so excited about it? it makes code much harder to read | 13:39 |
@lisitsyn | haha | 13:40 |
@lisitsyn | sonne|work: for (vector<int>::const_iterator it=begin() | 13:41 |
@lisitsyn | is more readable than | 13:41 |
@lisitsyn | for (auto = begin ..? | 13:41 |
@lisitsyn | err | 13:41 |
@lisitsyn | for (auto it = begin . | 13:41 |
sonne|work | in a for loop not but do this auto someplace somehwere | 13:41 |
@lisitsyn | depends of course | 13:42 |
@lisitsyn | decltype may be better | 13:42 |
@lisitsyn | and of course explicit type | 13:43 |
@lisitsyn | HeikoS1: more to come for you actually | 13:48 |
@lisitsyn | nullptr | 13:48 |
@lisitsyn | move semantics (this one you probably won't notice) | 13:48 |
@lisitsyn | std::thread may be | 13:48 |
HeikoS1 | lisitsyn: I like that c++ is still developing | 13:56 |
HeikoS1 | sonne|work: well this auto is mostly intended to reduce code in loops (all examples are loops) | 13:56 |
sonne|work | for that it makes sense totally | 13:57 |
@lisitsyn | HeikoS1: yes that's good even though it is way too difficult already | 13:58 |
HeikoS1 | lisitsyn: what does nullptr do? | 13:58 |
HeikoS1 | can one catch null pointer exceptions? | 13:59 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun | 14:01 | |
HeikoS1 | lisitsyn: have you seen roman lately? | 14:01 |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 14:01 | |
sonne|work | lisitsyn: std::thread looks nice... | 14:02 |
@lisitsyn | HeikoS: no I didn't | 14:03 |
@lisitsyn | HeikoS1: no but they are more safe | 14:03 |
@lisitsyn | here is the example posted everywhere | 14:03 |
@lisitsyn | void func(int n); void func(char *s); func( NULL ); // guess which function gets called? | 14:03 |
@lisitsyn | I am not sure what happens on dereferencing nullptr | 14:03 |
HeikoS1 | lisitsyn: I see :) | 14:03 |
HeikoS1 | yeah NULL==0 is a horrible construct | 14:04 |
@lisitsyn | HeikoS1: ahh it fails to compile | 14:04 |
HeikoS1 | std::thread looks indeed nice | 14:04 |
@lisitsyn | HeikoS1: if not assigned to any pointer | 14:04 |
@lisitsyn | if it is assigned so so | 14:05 |
@lisitsyn | you get basic pointer so you can't catch it | 14:05 |
@lisitsyn | HeikoS1: *nullptr fails to compile though | 14:05 |
HeikoS1 | lisitsyn: okay but that cannot be checked at compile time always right? | 14:05 |
@lisitsyn | HeikoS1: yeah that's runtime thingy | 14:06 |
@lisitsyn | HeikoS1: you may catch null pointers with shared/scoped/../_ptr though | 14:06 |
HeikoS1 | lisitsyn: cool | 14:07 |
HeikoS1 | lisitsyn: on the other hand | 14:07 |
HeikoS1 | thats all very complicated | 14:07 |
HeikoS1 | normal poitners are complicated already | 14:07 |
@lisitsyn | HeikoS1: pointers are no-go for me | 14:08 |
HeikoS1 | eigen has a very nice approach in my eyes :) | 14:08 |
HeikoS1 | and openCV is also almost pointer free | 14:08 |
HeikoS1 | everything is done on stack | 14:08 |
HeikoS1 | and the rest is hidden | 14:08 |
@lisitsyn | HeikoS1: that's how it should be done | 14:09 |
HeikoS1 | lisitsyn: well not for us | 14:09 |
@lisitsyn | HeikoS1: why not? | 14:09 |
HeikoS1 | (only if its rewritten) | 14:09 |
@lisitsyn | hah | 14:09 |
HeikoS1 | but lets not annoy sonne|work too much with that :) | 14:09 |
@lisitsyn | HeikoS1: such a monster is unsupportable in my eyes | 14:10 |
@lisitsyn | :) | 14:10 |
HeikoS1 | what? | 14:11 |
@wiking | what magic do i miss here: /Users/wiking/shogun/src/shogun/lib/SGVector.h:54: Warning 503: Can't wrap 'operator short*' unless renamed to a valid identifier. | 14:12 |
sonne|work | swig | 14:14 |
sonne|work | I guess ruby. ignore | 14:14 |
@wiking | sonne|work: noup it's python now | 14:15 |
@lisitsyn | wiking: ignore ignore ignore it | 14:16 |
@wiking | ok | 14:16 |
@lisitsyn | wiking: not you | 14:16 |
@lisitsyn | but in swig_ignores | 14:16 |
@lisitsyn | or how is it called | 14:16 |
@wiking | /Users/wiking/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9318:55: error: use of undeclared identifier 'Integer' | 14:16 |
@wiking | fuck | 14:16 |
@wiking | whyyyy | 14:16 |
sonne|work | IIInteger? | 14:16 |
@lisitsyn | ehm I am not sure what happens here now | 14:16 |
sonne|work | lisitsyn: we shouldn't need to add it to ignores | 14:17 |
@lisitsyn | sonne|work: oh you know what is Integer do you? | 14:17 |
@lisitsyn | ;) | 14:17 |
@wiking | /Users/wiking/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:206720:45: error: use of undeclared identifier 'NPY_FLOAT32'; did you mean 'PT_FLOAT32'? | 14:17 |
@lisitsyn | java jokes are just like gay jokes | 14:17 |
sonne|work | IIRC operators and are supported in most recent versions of swig | 14:17 |
@wiking | it went wrong with numpy include i guess | 14:17 |
@lisitsyn | sonne|work: I wdnt export them | 14:17 |
sonne|work | IDK | 14:18 |
@wiking | ok | 14:18 |
@wiking | wtf am i missing here | 14:18 |
@wiking | https://gist.github.com/vigsterkr/5756383 | 14:19 |
sonne|work | some numpy.h include? | 14:20 |
sonne|work | or python.h? | 14:20 |
sonne|work | IDK | 14:20 |
@wiking | sonne|work: but i'm just swiging the same modshogun.i | 14:21 |
@wiking | ;) | 14:21 |
sonne|work | c++'ing you mean | 14:22 |
@wiking | well yeah after generating from modshogun.i | 14:22 |
sonne|work | lisitsyn, HeikoS1 I don't mind using references only. | 14:23 |
HeikoS1 | me neither | 14:24 |
@lisitsyn | sonne|work: references? | 14:24 |
@lisitsyn | like &? | 14:24 |
@lisitsyn | what's swig support of that? | 14:24 |
sonne|work | well we write the typemaps | 14:24 |
sonne|work | so anything is possible | 14:25 |
@lisitsyn | references as class fields are not very good | 14:25 |
sonne|work | what did you mean above with pointer free code? | 14:26 |
@lisitsyn | sonne|work: shared_ptr | 14:26 |
@iglesiasg | lisitsyn, sonne|work : what do we have around Shogun for linear SVMs solved in the primal? | 14:26 |
@lisitsyn | references are not any better | 14:26 |
@iglesiasg | I just want to have a simple example where I can get weight vector and bias | 14:26 |
@lisitsyn | iglesiasg: liblinear? | 14:27 |
@iglesiasg | lisitsyn: let me see | 14:27 |
@iglesiasg | lisitsyn: yeah! it was straightforward to make it work with it, thanks! | 14:39 |
-!- FSCV [~FSCV@187.210.54.166] has joined #shogun | 14:42 | |
@wiking | wohoooo | 14:58 |
@wiking | cmake swig interface is oooon | 14:58 |
@wiking | heh i missed a linker option :) | 14:59 |
sonne|work | wiking: haha :) | 15:13 |
sonne|work | but sounds excellent | 15:13 |
@wiking | ah and i missed a cpp file in python | 15:35 |
@wiking | :) | 15:35 |
-!- nube [~rho@49.244.71.138] has joined #shogun | 15:58 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 16:26 | |
shogun-notifier- | shogun: Heiko Strathmann :develop * 512d94f / src/shogun/lib/DataType.cpp,src/shogun/lib/DataType.h: https://github.com/shogun-toolbox/shogun/commit/512d94f6a6abb1a1fe43eb22d4dbe65e2347c855 | 16:26 |
shogun-notifier- | shogun: refactored sizeof_ptype, sizeof_stype to make them available from static context | 16:26 |
shogun-notifier- | shogun: Heiko Strathmann :develop * a5435a5 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/a5435a543f2265fc4409ac487f33fc72d7e7b663 | 16:26 |
shogun-notifier- | shogun: first parts of copy_ptype with unit tests | 16:26 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 883ac63 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/883ac6316c1401812621067b8890205715c6b684 | 16:26 |
shogun-notifier- | shogun: -refactored compare_stype a bit | 16:26 |
shogun-notifier- | shogun: -added copy_stype method with first bits | 16:26 |
shogun-notifier- | shogun: -added first unit test for copy_stype | 16:26 |
shogun-notifier- | shogun: Heiko Strathmann :develop * caef78d / / (8 files): https://github.com/shogun-toolbox/shogun/commit/caef78dbe9682e5b6aa77ad8feb25dd70e8be497 | 16:26 |
shogun-notifier- | shogun: initial (working) attempt for a general purpose clone method of CSGObject. Comes with first basic unit tests. More to come | 16:26 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 855b1be / / (8 files): https://github.com/shogun-toolbox/shogun/commit/855b1be43183afdde4b7e2edea0e42d867be04ca | 16:26 |
shogun-notifier- | shogun: Merge pull request #1167 from karlnapf/develop | 16:26 |
HeikoS1 | wiking: there you go! :) still waiting for the automagic tests | 16:26 |
@wiking | hehe yeah me too | 16:29 |
@wiking | :) | 16:29 |
@wiking | i want to finish this installation problem with modpython | 16:29 |
@wiking | and then i'll switch back to the develop branch | 16:29 |
shogun-buildbot | build #934 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/934 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 16:50 |
shogun-buildbot | build #935 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/935 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 16:51 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 16:53 | |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Ping timeout: 250 seconds] | 17:15 | |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun | 17:19 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 17:21 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 17:21 | |
shogun-notifier- | shogun: Roman Votyakov :feature/sparse_matrix * cd27d05 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/cd27d057b1c1a7cd2e5a22741c9b7be1e1d8e896 | 17:23 |
shogun-notifier- | shogun: Redesigned GP framework | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 53f4ea1 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/53f4ea1ff26387eb1d85988c3f547c7e8f94676a | 17:23 |
shogun-notifier- | shogun: Merge pull request #1150 from votjakovr/develop | 17:23 |
shogun-notifier- | shogun: | 17:23 |
shogun-notifier- | shogun: Redesigned GP framework | 17:23 |
shogun-notifier- | shogun: lambday :feature/sparse_matrix * 23d134a / / (10 files): https://github.com/shogun-toolbox/shogun/commit/23d134aa08d65d5d8df380c46c4140af9ea76398 | 17:23 |
shogun-notifier- | shogun: error msgs fixed for complex64_t, unit-tests added | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 6a47272 / / (10 files): https://github.com/shogun-toolbox/shogun/commit/6a47272a5d5982979faddf98d551eef60ee04c60 | 17:23 |
shogun-notifier- | shogun: Merge pull request #1149 from lambday/develop | 17:23 |
shogun-notifier- | shogun: | 17:23 |
shogun-notifier- | shogun: error msgs fixed for complex64_t, unit-tests added | 17:23 |
shogun-notifier- | shogun: Sergey Lisitsyn :feature/sparse_matrix * e496115 / tests/unit/lib/SGVector_unittest.cc,tests/unit/mathematics/Complex_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/e4961157662ed92cac9139fd4eca1913a4548771 | 17:23 |
shogun-notifier- | shogun: Lowered precision of complex unit-tests | 17:23 |
shogun-notifier- | shogun: lambday :feature/sparse_matrix * 43c4b4e / tests/unit/base/Parameter_unittest.cc,tests/unit/mathematics/Math_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/43c4b4e94fb392a290e8784379e0e9eb6fc9415e | 17:23 |
shogun-notifier- | shogun: complex64_t unit-tests added for TParameter | 17:23 |
shogun-notifier- | shogun: Fernando Iglesias :feature/sparse_matrix * 44c8ae8 / tests/unit/base/Parameter_unittest.cc,tests/unit/mathematics/Math_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/44c8ae8e63b218919cec4167419c2a799272cc8b | 17:23 |
shogun-notifier- | shogun: Merge pull request #1151 from lambday/develop | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 9850f44 / tests/unit/regression/gp/ (2 files): https://github.com/shogun-toolbox/shogun/commit/9850f44aa213fab987e30c179e58a5bcbbdbed9f | 17:23 |
shogun-notifier- | shogun: Merge pull request #1159 from votjakovr/develop | 17:23 |
shogun-notifier- | shogun: | 17:23 |
shogun-notifier- | shogun: Fixed memory leak in GaussianLikelihood and StudentsTLikelihood unit tests | 17:23 |
shogun-notifier- | shogun: Sergey Lisitsyn :feature/sparse_matrix * 818a18b / tests/unit/converter/TDistributedStochasticNeighborEmbedding_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/818a18beed6343a860178f2347ecdf3ba0232112 | 17:23 |
shogun-notifier- | shogun: Added missed unrefs in T-SNE unit test. Closes #1155 | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 512d94f / src/shogun/lib/DataType.cpp,src/shogun/lib/DataType.h: https://github.com/shogun-toolbox/shogun/commit/512d94f6a6abb1a1fe43eb22d4dbe65e2347c855 | 17:23 |
shogun-notifier- | shogun: refactored sizeof_ptype, sizeof_stype to make them available from static context | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * a5435a5 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/a5435a543f2265fc4409ac487f33fc72d7e7b663 | 17:23 |
shogun-notifier- | shogun: first parts of copy_ptype with unit tests | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 883ac63 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/883ac6316c1401812621067b8890205715c6b684 | 17:23 |
shogun-notifier- | shogun: -refactored compare_stype a bit | 17:23 |
shogun-notifier- | shogun: -added copy_stype method with first bits | 17:23 |
shogun-notifier- | shogun: -added first unit test for copy_stype | 17:23 |
shogun-notifier- | shogun: Thoralf Klein :feature/sparse_matrix * 4055e99 / src/shogun/kernel/Kernel.cpp,tests/unit/kernel/CustomKernel_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/4055e99fcc324d85c3badce8df6892deb4dcae5c | 17:23 |
shogun-notifier- | shogun: CKernel: Fixed uninitialized "lhs_equals_rhs" | 17:23 |
shogun-notifier- | shogun: CCustomKernel: Fixed memory leak in unit test. | 17:23 |
shogun-notifier- | shogun: Sergey Lisitsyn :feature/sparse_matrix * 87a3fce / src/shogun/kernel/Kernel.cpp,tests/unit/kernel/CustomKernel_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/87a3fcebceaedf8a82785e406b59301f32779687 | 17:23 |
shogun-notifier- | shogun: Renamed liblinear structures to avoid name clashes in future | 17:23 |
shogun-notifier- | shogun: Sergey Lisitsyn :feature/sparse_matrix * 13616ed / src/ (9 files): https://github.com/shogun-toolbox/shogun/commit/13616ede4921bc33cd5a5ac965cf0d6bbb56f7fd | 17:23 |
shogun-notifier- | shogun: Remove redundant clone, drop MT composite machine | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * caef78d / / (8 files): https://github.com/shogun-toolbox/shogun/commit/caef78dbe9682e5b6aa77ad8feb25dd70e8be497 | 17:23 |
shogun-notifier- | shogun: initial (working) attempt for a general purpose clone method of CSGObject. Comes with first basic unit tests. More to come | 17:23 |
shogun-notifier- | shogun: Heiko Strathmann :feature/sparse_matrix * 855b1be / / (8 files): https://github.com/shogun-toolbox/shogun/commit/855b1be43183afdde4b7e2edea0e42d867be04ca | 17:23 |
shogun-notifier- | shogun: Merge pull request #1167 from karlnapf/develop | 17:23 |
shogun-notifier- | shogun: | 17:23 |
shogun-notifier- | shogun: first attempt for general clone method | 17:23 |
shogun-notifier- | shogun: Soeren Sonnenburg :feature/sparse_matrix * ae1b8d8 / / (7 files): https://github.com/shogun-toolbox/shogun/commit/ae1b8d8c820185adc163d92a9187ff3f3406af42 | 17:24 |
shogun-notifier- | shogun: make sparse features use SGSparseMatrix underneath | 17:24 |
shogun-notifier- | shogun: Soeren Sonnenburg :feature/sparse_matrix * 434ca62 / src/shogun/features/SparseFeatures.cpp,src/shogun/features/SparseFeatures.h: https://github.com/shogun-toolbox/shogun/commit/434ca62168ec28bbfe657b4bdadd1df8cb602b8d | 17:24 |
shogun-notifier- | shogun: don't change the api return sparse transposed features | 17:24 |
shogun-notifier- | shogun: Soeren Sonnenburg :feature/sparse_matrix * ad7b340 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/ad7b340a54005964c8b7fe4f138fa588c8c1af23 | 17:24 |
shogun-notifier- | shogun: fix feature iterators for sparsefeatures and combined dot features | 17:24 |
shogun-notifier- | shogun: Soeren Sonnenburg :feature/sparse_matrix * 125cd93 / : https://github.com/shogun-toolbox/shogun/commit/125cd937a4a787cca04dd1041a5b8e59e61c7d52 | 17:24 |
shogun-notifier- | shogun: Merge branch 'feature/sparse_matrix' of github.com:shogun-toolbox/shogun into feature/sparse_matrix | 17:24 |
shogun-notifier- | shogun: | 17:24 |
HeikoS1 | what was that? | 17:24 |
@sonney2k | HeikoS1, a rebase on develop | 17:26 |
HeikoS1 | I see | 17:26 |
@sonney2k | van51, I've finally fixed the crashers in the feature iterators | 17:26 |
HeikoS1 | sonney2k: unit test them ! :) | 17:26 |
@sonney2k | van51, the situation is pretty delicate | 17:27 |
van51 | sonney2k: nice :D | 17:27 |
van51 | sonney2k: what do you mean? | 17:27 |
@sonney2k | and I don't like some inconsistency we are having | 17:28 |
@sonney2k | van51, reason was basically that we have SGVector in the interator | 17:28 |
@sonney2k | iterator | 17:28 |
@sonney2k | which was allocated with SG_MALLOC | 17:28 |
@sonney2k | even though it needs to be allocated with new bla_iterator() | 17:28 |
@sonney2k | so kaboom | 17:29 |
@sonney2k | other iterators don't need that | 17:29 |
@sonney2k | HeikoS1, it is already due to van51 | 17:29 |
@wiking | HeikoS1: ah actually i can finish now the begging machine i hope :) | 17:29 |
HeikoS1 | sonney2k: I see, nice! | 17:29 |
HeikoS1 | wiking: yes looking forward to see that | 17:29 |
@sonney2k | wiking, beg me plenty! | 17:30 |
HeikoS1 | but pls do the test before, I dont know whether clone works everywhere | 17:30 |
van51 | sonney2k: that was also the inconsistency you mentioned or there is something more? | 17:30 |
@wiking | beg me sweet :) | 17:30 |
HeikoS1 | wiking: will be pretty cool with this subsampling | 17:30 |
HeikoS1 | I coded this up a few times by hand for different shogun machines | 17:30 |
HeikoS1 | its very useful | 17:30 |
@wiking | mmm but yeah lets see my latest cmake hacking | 17:30 |
HeikoS1 | in practice | 17:30 |
@wiking | if this works out finally | 17:31 |
@wiking | then i just need to fix the other templates for modular interfaces :P | 17:31 |
@wiking | althoguh i still didn't do static interface | 17:31 |
@wiking | dont we wanna drop that anyways? :D | 17:31 |
HeikoS1 | its very outdated indeed | 17:32 |
@sonney2k | wiking, I would but some people here use matlab :D | 17:32 |
@sonney2k | ohh and R | 17:32 |
@wiking | doh | 17:32 |
@sonney2k | wiking, we would need an individual with superhuman powers to write swig wrappers for these languages | 17:33 |
@sonney2k | van51, no the other one was that we didn't return SGSparseVectors' properly refcounted | 17:35 |
@sonney2k | van51, so 2 reasons for the crasher/leak | 17:35 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Ping timeout: 250 seconds] | 17:36 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 260 seconds] | 17:37 | |
van51 | sonney2k: I see. and you 've fixed them both ? | 17:38 |
van51 | if that's the case I'll complete that unit test and make a PR by tonight | 17:38 |
@sonney2k | van51, I haven't pushed to develop yet though | 17:39 |
thoralf | sonney2k: So the sparse vectors don't leak any more? | 17:39 |
@sonney2k | currently in 'feature/sparse_matrix' | 17:39 |
@sonney2k | thoralf, where did they leak? | 17:39 |
@sonney2k | maybe not but no idea | 17:39 |
@sonney2k | argh | 17:39 |
@sonney2k | gtg | 17:39 |
@sonney2k | cu | 17:39 |
thoralf | sonney2k: Two weeks ago, an issue I fixed with a hack. | 17:40 |
thoralf | sonney2k: Tradeoff between possible leak and segfault. ;) | 17:40 |
-!- lisitsyn [~lisitsyn@188.168.2.227] has joined #shogun | 17:51 | |
shogun-buildbot | build #463 of ubu1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ubu1%20-%20libshogun/builds/463 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 17:59 |
-!- travis-ci [~travis-ci@ec2-54-235-63-20.compute-1.amazonaws.com] has joined #shogun | 17:59 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7986900 | 17:59 |
-!- travis-ci [~travis-ci@ec2-54-235-63-20.compute-1.amazonaws.com] has left #shogun [] | 17:59 | |
-!- thoralfk [~thoralf@enki.zib.de] has joined #shogun | 18:05 | |
-!- lisitsyn [~lisitsyn@188.168.2.227] has left #shogun [] | 18:07 | |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * b52248c / / (4 files): https://github.com/shogun-toolbox/shogun/commit/b52248cdf4de7e2c4daf59d2327bec069004b76d | 18:08 |
shogun-notifier- | shogun: WIP: modular interface building by cmake | 18:08 |
shogun-notifier- | shogun: modshogun_doxygen.i generation is still missing | 18:08 |
@wiking | fuck doxygen.i | 18:10 |
@wiking | :) | 18:10 |
-!- thoralf [~thoralf@enki.zib.de] has quit [Ping timeout: 246 seconds] | 18:11 | |
-!- thoralfk is now known as thoralf | 18:11 | |
-!- FSCV [~FSCV@187.210.54.166] has quit [Quit: This computer has gone to sleep] | 18:15 | |
-!- travis-ci [~travis-ci@ec2-50-19-67-126.compute-1.amazonaws.com] has joined #shogun | 18:19 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7988228 | 18:19 |
-!- travis-ci [~travis-ci@ec2-50-19-67-126.compute-1.amazonaws.com] has left #shogun [] | 18:19 | |
-!- FSCV [~FSCV@187.210.54.166] has joined #shogun | 19:25 | |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun | 19:37 | |
-!- hoijui [~hoijui@dslb-088-075-034-035.pools.arcor-ip.net] has joined #shogun | 19:51 | |
-!- HeikoS1 [~heiko@nat-169-135.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 20:13 | |
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Remote host closed the connection] | 20:14 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has left #shogun ["PING 1370976958"] | 20:56 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 21:08 | |
-!- hoijui [~hoijui@dslb-088-075-034-035.pools.arcor-ip.net] has quit [Quit: Leaving] | 21:10 | |
-!- lisitsyn [~lisitsyn@109-226-90-135.clients.tlt.100megabit.ru] has joined #shogun | 21:19 | |
-!- mode/#shogun [+o lisitsyn] by ChanServ | 21:20 | |
@lisitsyn | HeikoS: oh I don't remember exactly | 21:23 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 21:26 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 330aadc / / (9 files): https://github.com/shogun-toolbox/shogun/commit/330aadce2c90d72022019b96ecfe8e2b8cafc1bb | 21:26 |
shogun-notifier- | shogun: convert CSparseFeatures to use SGSparseMatrix | 21:26 |
shogun-buildbot | build #936 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/936 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:33 |
@sonney2k | wiking, you should check whether swig by now supports doxygen? | 21:36 |
@wiking | there's a branch from gsoc 2012 of swig | 21:38 |
@wiking | that supports it | 21:38 |
@wiking | but i dont know if it has been merged or what | 21:38 |
@lisitsyn | what's uppppp | 21:47 |
@lisitsyn | wiking: can I help you with cmaking it? | 21:48 |
-!- travis-ci [~travis-ci@ec2-54-235-22-108.compute-1.amazonaws.com] has joined #shogun | 21:51 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7994284 | 21:51 |
-!- travis-ci [~travis-ci@ec2-54-235-22-108.compute-1.amazonaws.com] has left #shogun [] | 21:51 | |
shogun-buildbot | build #464 of ubu1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ubu1%20-%20libshogun/builds/464 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:11 |
-!- gsomix [~Miranda@109.169.140.87] has joined #shogun | 22:13 | |
gsomix | I'm at home, I'm at home | 22:13 |
shogun-buildbot | build #1228 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/1228 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:14 |
shogun-notifier- | shogun-data: Soeren Sonnenburg :master * 312c4b7 / testsuite/tests/ (14 files): https://github.com/shogun-toolbox/shogun-data/commit/312c4b75015045cad974f0fda0969fb245760af1 | 22:24 |
shogun-notifier- | shogun-data: update sparse outputs due to sparse matrix change | 22:24 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * abd44eb / data: https://github.com/shogun-toolbox/shogun/commit/abd44ebaf7c008502cf3fbbae8a139056d4c9ce1 | 22:25 |
shogun-notifier- | shogun: require new data version | 22:25 |
@sonney2k | gsomix, good for you good for you | 22:26 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 22:27 | |
@sonney2k | van51, alright I've committed to develop - so you can continue | 22:28 |
gsomix | sonne|work: yep, yep :) did you see my code? is it good for merge (only test for buffer is needed)? | 22:28 |
@sonney2k | gsomix, new policy is no merge w/o test | 22:28 |
van51 | sonney2k: ok, thanks | 22:28 |
@sonney2k | van51, well thanks for finding the issue | 22:28 |
van51 | sonney2k: btw did you have a look at Olivier's milestones? | 22:29 |
@sonney2k | van51, yes I did | 22:29 |
@sonney2k | I think you can do more but just get it started :) | 22:30 |
@sonney2k | van51, btw did quoc say what his plans are? | 22:30 |
van51 | sonney2k: haven't talked to him again since the hangout | 22:30 |
gsomix | sonney2k: I know that. I mean "potentially". ok | 22:30 |
van51 | sonney2k: i'm thinking of logging in gchat now and talking a bit to him | 22:31 |
van51 | he told me he had a matlab implementation I could have a look at | 22:31 |
shogun-buildbot | build #937 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/937 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:32 |
@sonney2k | van51, so ask him for that. That would be very handy to test if the implementation is correct | 22:35 |
van51 | sonney2k: yea I just did but he's idle now | 22:36 |
van51 | for the meantime I'll get that test done | 22:36 |
van51 | in* | 22:37 |
-!- yoo_ [575b08cb@gateway/web/freenode/ip.87.91.8.203] has joined #shogun | 22:37 | |
gsomix | ok, time to sleep | 22:39 |
-!- yoo_ [575b08cb@gateway/web/freenode/ip.87.91.8.203] has quit [Quit: Page closed] | 22:43 | |
@sonney2k | van51, I don't know vw well enough so we will have to find out about milestone2 | 22:45 |
@sonney2k | van51, regarding milestone 3 we have quadratic (actually polynomial) dot features in shogun | 22:45 |
@sonney2k | but yes we should use hashing here | 22:45 |
van51 | sonney2k: I read about vw these days I don't think it will be hard | 22:46 |
van51 | sonney2k: Olivier's paper was actually very helpful and clarifying | 22:46 |
@sonney2k | yes it should not be difficult | 22:47 |
@sonney2k | so in the end we would have hashing for n-grams and say bag-of-words | 22:47 |
@sonney2k | and quadratic features | 22:47 |
van51 | sonney2k: yeah | 22:48 |
van51 | with streaming availability | 22:49 |
@sonney2k | yeah once you have one version working you can easily generalize this to streaming*dotfeaturs | 22:49 |
van51 | sonney2k: if I'm not missing something, n-grams and bag-of-words will just be two different tokenizers | 22:49 |
@sonney2k | van51, exactly | 22:50 |
van51 | sonney2k: good | 22:50 |
-!- travis-ci [~travis-ci@ec2-54-235-22-108.compute-1.amazonaws.com] has joined #shogun | 22:50 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7996108 | 22:50 |
-!- travis-ci [~travis-ci@ec2-54-235-22-108.compute-1.amazonaws.com] has left #shogun [] | 22:50 | |
@sonney2k | n-gram is very simple - just n characters | 22:50 |
van51 | at least they are in order in my mind | 22:50 |
@sonney2k | and then you move to the right | 22:50 |
@sonney2k | van51, for that one could use rolling hashes as they are called (so just do the additional character and remove the left hand side one) | 22:51 |
@sonney2k | van51, so for a string of length say 6 you have 4 - 3grams | 22:52 |
@sonney2k | (shifting the 3-gram by one all the time) | 22:53 |
@sonney2k | and bag of words would then take say ' ' etc as tokinizer | 22:53 |
@sonney2k | but you could also consider multiple words. some people also call these n-grams btw (for n == number of words to consider) | 22:54 |
van51 | sonney2k: yeah I've heard about the one | 22:54 |
van51 | and the removal occurs on the bit-level? | 22:55 |
@sonney2k | van51, no char level | 22:55 |
van51 | sonney2k: ok, so it's for the phase when the ngrams are created | 22:56 |
van51 | I mean when they are tokenized | 22:57 |
van51 | but this rolling hash is faster? | 22:57 |
@sonney2k | van51, jsut consider you have to compute the hash for an 8-gram | 22:57 |
@sonney2k | so you do h(x_1...x_8) | 22:58 |
@sonney2k | now to compute the hash for the next 8 gram you would do | 22:58 |
@sonney2k | h(x_2...x_9) | 22:58 |
@sonney2k | for the 'normal' hash function | 22:58 |
@sonney2k | with a rolling hash you can do | 22:58 |
van51 | sonney2k: ok yea | 22:58 |
@sonney2k | h(x_1...x_8) - some_value_computed_on(x_1) + some_value_computed_on(x_9) | 22:59 |
@sonney2k | so it *can* be faster | 22:59 |
van51 | sonney2k: ah I see | 22:59 |
van51 | sounds good to me | 23:00 |
@sonney2k | I used all these tricks for too :) | 23:00 |
van51 | sonney2k: for what? :) | 23:01 |
@sonney2k | van51, to detect splice sites on DNA :) | 23:01 |
@sonney2k | van51, would actually be cool if you could revive the method I wrote back then | 23:01 |
van51 | ahh yea I can see how it could be used there | 23:02 |
@sonney2k | it is maybe 20 lines of python code | 23:02 |
van51 | sonney2k: sure if you share it | 23:02 |
@sonney2k | somewhere on my homepage | 23:02 |
@sonney2k | but for training you need some machine with 64 GB memory :D | 23:02 |
@sonney2k | well of course for smaller subsets of the data / feature space we can use less | 23:03 |
@sonney2k | van51, it is here http://sonnenburgs.de/soeren/item/coffin/ | 23:03 |
@sonney2k | van51, btw one more nice application would be a gender classifier | 23:04 |
van51 | sonney2k: using what as input? | 23:04 |
@sonney2k | van51, http://sonnenburgs.de/soeren/talks/2010-06-23-COFFIN-ICML.pdf | 23:04 |
@sonney2k | have a look at the pictures at the end | 23:04 |
@sonney2k | we could use the face detector from opencv | 23:05 |
@sonney2k | and then based on the face predict the gender of the person | 23:05 |
@sonney2k | or the age | 23:05 |
@sonney2k | as you want | 23:05 |
@sonney2k | the features for that are half way implemented as dotfeatures in shogun but not finished | 23:06 |
@lisitsyn | sonney2k: lbp? | 23:06 |
@sonney2k | we can get vojtechs matlab code though | 23:06 |
@sonney2k | lisitsyn, yes | 23:06 |
@lisitsyn | sonney2k: good it worked so good | 23:06 |
@lisitsyn | I mean lbp is not like the state of the art | 23:06 |
@sonney2k | lisitsyn, but pyramides | 23:07 |
van51 | sonney2k: sounds real nice, I prefer a lot this idea for an application | 23:07 |
@sonney2k | van51, LBPPyrDotFeatures.h | 23:07 |
@lisitsyn | sonney2k: ah so pyramidal yes | 23:07 |
@lisitsyn | that's more trendy | 23:07 |
@sonney2k | as trendy as a 2009 paper can be... | 23:07 |
@lisitsyn | sonney2k: now it is all deep | 23:07 |
@lisitsyn | :D | 23:07 |
@sonney2k | 2010 | 23:07 |
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun | 23:07 | |
van51 | sonney2k: what are these pyramids btw? :) | 23:08 |
van51 | where can I read about it? | 23:08 |
@sonney2k | different scales | 23:08 |
@sonney2k | on my homepage there is a link to the coffin paper | 23:08 |
@lisitsyn | sonney2k: have you seen svm + deep? | 23:09 |
@lisitsyn | sonney2k: http://arxiv.org/pdf/1306.0239.pdf | 23:09 |
-!- iglesiasg [d58f3220@gateway/web/freenode/ip.213.143.50.32] has joined #shogun | 23:09 | |
@sonney2k | lisitsyn, they submitted this to nips | 23:10 |
@sonney2k | I am confused | 23:10 |
@sonney2k | this is ICML!? | 23:10 |
shogun-buildbot | build #465 of ubu1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ubu1%20-%20libshogun/builds/465 blamelist: Soeren Sonnenburg <sonne@debian.org> | 23:10 |
@sonney2k | I think I saw the exact same thing in the NIPS submissions | 23:10 |
@lisitsyn | sonney2k: hah | 23:10 |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 23:10 | |
@lisitsyn | sonney2k: did you get the idea? | 23:10 |
@lisitsyn | I admit I am floating around with no real understanding | 23:11 |
@lisitsyn | just changing objective? | 23:11 |
@sonney2k | I didn't have a look | 23:11 |
@sonney2k | but I like the facial expression recognition data set in there | 23:12 |
@sonney2k | maybe that is sth for van51 :) | 23:12 |
@sonney2k | van51, what is the hardware you have access too? | 23:12 |
@sonney2k | to? | 23:12 |
@sonney2k | van51, there is a data set with ~30000 images of 48x48 for which to predict facial expression | 23:12 |
@sonney2k | you could also do this with dotfeatures and the lbppyr* | 23:13 |
van51 | sonney2k: well atm I have a desktop and a borrowed laptop | 23:13 |
@sonney2k | van51, speed? memory? | 23:14 |
@lisitsyn | good it is not stolen! | 23:14 |
@lisitsyn | ;) | 23:14 |
van51 | but as a student I also have access to a couple machines in some cloud | 23:14 |
van51 | max 4gb ram | 23:14 |
@sonney2k | that is likely not enough | 23:14 |
van51 | sonney2k: I still have access to a machine in my supervisor's lab | 23:15 |
van51 | maybe that could be used :p | 23:15 |
@sonney2k | lisitsyn, should we maybe rent some bigger machine for this gsoc? | 23:15 |
@sonney2k | van51, hehe sounds good! | 23:15 |
van51 | it has 8gb's though | 23:15 |
van51 | hope that is enough | 23:15 |
@lisitsyn | sonney2k: with something that supports opencl ;) | 23:16 |
@sonney2k | lisitsyn, man you and your GPU addiction | 23:16 |
@lisitsyn | sonney2k: haha | 23:16 |
@lisitsyn | sonney2k: I need something fresh I am bored | 23:16 |
@sonney2k | 60 EUR/month for 16GB and some corei5 | 23:16 |
@sonney2k | lisitsyn, turn your road sign recognizer into a nice demo then! | 23:17 |
@lisitsyn | sonney2k: I am bored with it | 23:17 |
@lisitsyn | I am basically bored with everything :D | 23:17 |
@sonney2k | ahh the old problem | 23:17 |
@sonney2k | lisitsyn, how much are the ec2's for that kind of task | 23:18 |
@lisitsyn | sonney2k: uh pretty much I guess | 23:18 |
@lisitsyn | sonney2k: depends on the type of the instance | 23:18 |
@lisitsyn | let me check | 23:19 |
shogun-buildbot | build #1229 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/1229 | 23:20 |
@sonney2k | 1 EUR / hour | 23:21 |
@sonney2k | for a ~32G instance | 23:21 |
@lisitsyn | sonney2k: M3? | 23:22 |
@lisitsyn | omg they have such a monsters | 23:23 |
@lisitsyn | 16 cores 130 gb | 23:23 |
@lisitsyn | :D | 23:23 |
@lisitsyn | sonney2k: how much memory do you need and what for? | 23:23 |
@sonney2k | not really a monster considering that gunnar has a couple of 512GB / 64 core machines | 23:24 |
@sonney2k | and actually the were ~10kEUR 3 years back | 23:24 |
@lisitsyn | sonney2k: well that's organization level money | 23:25 |
@lisitsyn | nobody buys it to watch movies after hard day | 23:25 |
@lisitsyn | :D | 23:25 |
@wiking | letssee what travis says to this | 23:25 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 7f6c46f / / (4 files): https://github.com/shogun-toolbox/shogun/commit/7f6c46fb5bd707bc48e3b89939e741f1abd49e86 | 23:25 |
shogun-notifier- | shogun: An almost ready version of python_modular interface with cmake | 23:25 |
@lisitsyn | wiking: you take it serious man | 23:25 |
@sonney2k | wiking, go go | 23:26 |
@wiking | why? | 23:26 |
@lisitsyn | wiking: so anything I could help with? | 23:26 |
@wiking | lisitsyn: well if this works out | 23:26 |
@sonney2k | I wish heiko didn't finish clone() so quickly | 23:26 |
@wiking | basically we need to refactor a bit the cmakelists.txt in python_modular | 23:26 |
@wiking | so that it works for all modular | 23:26 |
@wiking | i.e. i just need to cut that file in half | 23:26 |
@wiking | put one part in interface/CMakelists.txt the rest in the *_modular/Cmakelists.txt and include interface/CMakelists.txt from those | 23:27 |
@wiking | and then i think the *_modular are done | 23:27 |
@sonney2k | wiking, once you finish you give us a tour right? | 23:27 |
@wiking | lisitsyn: you could refactor a bit the shogun/Cmakefiles.txt | 23:27 |
@sonney2k | I won't be able to parse Cmake* and do additions | 23:28 |
@wiking | lisitsyn: 3 main things are missing there | 23:28 |
@lisitsyn | wiking: what's wrong with it? | 23:28 |
@lisitsyn | sonney2k: that's easy just get some vodka | 23:28 |
@wiking | 1) for example all the fancy cpu optimization flags are missing, i.e. cpuid detection etc. | 23:28 |
@wiking | 2) spinlock detection is missing | 23:28 |
@wiking | 3) the autodetection of the possible interfaces and dying if for example swig is not present and somebody wants to build a modular interface | 23:29 |
@lisitsyn | well I am considering a case without all that cpuid stuff | 23:29 |
@lisitsyn | sonney2k: do we really need it? | 23:29 |
@wiking | for 3) almost everything is there in shogun/CMakeLists.txt | 23:29 |
@wiking | we just need to have the right if,else shit | 23:29 |
@lisitsyn | wiking: got it I'll do tomorrow morning | 23:29 |
@wiking | spinlock detection can be taken straight from ./configure | 23:30 |
@wiking | and put it into a separate cmake/DetectSpinlock.cmake file | 23:30 |
@lisitsyn | wiking: yes I understand | 23:30 |
@wiking | and put the cmake syntax aroudn that cc test | 23:30 |
@wiking | ah yeah | 23:30 |
@wiking | tests target is missing | 23:31 |
@lisitsyn | can do that too | 23:31 |
@wiking | i.e. running the examples and integration test | 23:31 |
@wiking | shouldn't be that hard | 23:31 |
@lisitsyn | unit-tests? | 23:31 |
@wiking | that's done | 23:31 |
@lisitsyn | ahh I see | 23:31 |
@lisitsyn | cool | 23:31 |
@wiking | as you can see from the previous build on travis | 23:31 |
@wiking | the simple libshogun interface build runs | 23:31 |
@wiking | https://travis-ci.org/shogun-toolbox/shogun/jobs/7998167 | 23:32 |
@lisitsyn | wiking: so now using ctest | 23:32 |
@lisitsyn | we can get prettier output | 23:32 |
@wiking | actually no | 23:32 |
@wiking | but yeah we could use ctest as well | 23:32 |
@wiking | ;) | 23:32 |
@wiking | i thought to add all the tests into the ctest target | 23:32 |
@wiking | but then i didn't realise how i can have the test outputs more verbose | 23:33 |
@lisitsyn | wiking: ctest -VV | 23:33 |
@wiking | lisitsyn: yep | 23:33 |
@lisitsyn | wiking: it is like when you shout verbose | 23:33 |
@lisitsyn | two times | 23:33 |
@lisitsyn | :) | 23:33 |
@wiking | but how do you translet that to make test ? :) | 23:33 |
@lisitsyn | wiking: well in tapkee I kept the makefile | 23:34 |
@sonney2k | lisitsyn, cpu detection is the least of our problems. sse might be worth it but -march=native should do it | 23:34 |
@sonney2k | (nowadays) | 23:34 |
@lisitsyn | sonney2k: yeah modern compilers are modern enough | 23:34 |
@lisitsyn | to do that | 23:34 |
@lisitsyn | wiking: and added a target for that | 23:34 |
@lisitsyn | wiking: so make test runs it too right? | 23:35 |
@lisitsyn | builds then runs | 23:36 |
@lisitsyn | or what? | 23:36 |
@wiking | add_test(unit-tests shogun-unit-test) | 23:37 |
@wiking | added to make test | 23:37 |
@wiking | but you still have the old target | 23:37 |
@wiking | make unit-tests | 23:37 |
@wiking | which basically builds + runs the binary | 23:37 |
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has joined #shogun | 23:37 | |
@lisitsyn | wiking: yeah I am not sure how to pass -VV here | 23:37 |
@wiking | lol | 23:37 |
@wiking | this error is funny | 23:38 |
@wiking | https://travis-ci.org/shogun-toolbox/shogun/jobs/7998168 | 23:38 |
@wiking | i have basically none of these on my machine :) | 23:38 |
@wiking | let's see on an ubuntu | 23:38 |
@wiking | with verbose mode wtf | 23:38 |
@lisitsyn | let me sparsatrix? | 23:38 |
@lisitsyn | ehmm | 23:38 |
@lisitsyn | remove let me it was from another message | 23:39 |
@lisitsyn | what is sparsatrix | 23:39 |
-!- travis-ci [~travis-ci@ec2-23-23-2-43.compute-1.amazonaws.com] has joined #shogun | 23:39 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7998165 | 23:39 |
-!- travis-ci [~travis-ci@ec2-23-23-2-43.compute-1.amazonaws.com] has left #shogun [] | 23:39 | |
@lisitsyn | wiking: am I too tired or? | 23:40 |
@wiking | dunno | 23:40 |
@wiking | really no clue about what's happening there on travis | 23:40 |
@lisitsyn | /home/travis/build/shogun-toolbox/shogun/src/shogun/lib/SGSparsatrix.h:23:26: | 23:40 |
@lisitsyn | Sparsatrix? | 23:40 |
@lisitsyn | :D | 23:40 |
@lisitsyn | how did that happen | 23:41 |
@wiking | :D | 23:41 |
@lisitsyn | Spars_eM_atrix | 23:41 |
@lisitsyn | what makes eM be deleted? | 23:41 |
@wiking | yeah no clue :( | 23:41 |
@lisitsyn | wiking: even file was renamed | 23:42 |
@lisitsyn | well let me build it | 23:42 |
@wiking | sonney2k: what does .scrub_docstrings.py do actually? | 23:42 |
@lisitsyn | why there is no Sparsector | 23:42 |
@lisitsyn | sparsector would be nice too | 23:43 |
@wiking | ah i think this is the problem | 23:43 |
@lisitsyn | wiking: scrub docstrings? | 23:43 |
@wiking | i did not ran .scrub_docstrings.py on the _wrap.cxx file ;) | 23:43 |
@wiking | well here we go another target :) | 23:44 |
@lisitsyn | oh I am powering off | 23:45 |
@sonney2k | lisitsyn, sparsatian | 23:46 |
@lisitsyn | wiking: I'll fix these three issues tomorrow then | 23:46 |
@wiking | lisitsyn: cool thnx | 23:46 |
@lisitsyn | sonney2k: yes that would be nice too | 23:46 |
@wiking | hopefully i'll have atleast the python_modular done by tomorrow morning | 23:46 |
@lisitsyn | sonney2k: stringist | 23:46 |
@lisitsyn | sonney2k: deneatures | 23:46 |
@wiking | ah yeah libshogun static target is missing as well | 23:46 |
@sonney2k | that ec2 node is kaboom | 23:47 |
@sonney2k | memory error | 23:47 |
@lisitsyn | sonney2k: ? | 23:47 |
@lisitsyn | did you crash it? | 23:49 |
--- Log closed Wed Jun 12 00:00:41 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!