IRC logs of #shogun for Wednesday, 2011-11-23

--- Log opened Wed Nov 23 00:00:51 2011
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun02:25
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 260 seconds]04:16
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun05:00
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 240 seconds]05:19
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun05:46
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 252 seconds]05:53
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun06:37
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 240 seconds]06:43
CIA-37shogun: Soeren Sonnenburg master * recdb955 / (3 files in 3 dirs): rb_require narray on modshogun module load - http://git.io/JHWmTw08:53
-!- policecar [5ce7be6e@gateway/web/freenode/ip.92.231.190.110] has joined #shogun10:24
policecarhey hey, i'm manually installing shogun 1.0.0 on snowleopard and trying to make it work with a macports installed python2.6. do you happen to know where i need to put the shogun binaries for this python to be able to see the module? basically i'm looking for a macport not fink path version of /usr/local/lib/python2.6/dist-packages/ ...10:33
-!- policecar_ [~policecar@g231190110.adsl.alicedsl.de] has joined #shogun10:36
-!- policecar [5ce7be6e@gateway/web/freenode/ip.92.231.190.110] has quit [Quit: Page closed]10:37
-!- policecar_ is now known as policecar10:37
sonne|workpolicecar: please try shogun from git11:52
sonne|workor did it compile OK already11:52
sonne|work(I did a lot of fixes recently)11:53
policecarit compiled just fine. but thanks for the hint.11:53
sonne|workok then11:53
sonne|workjust search for e.g. modshogun.py11:53
policecari just solved the problem by adding /usr/local/lib/python2.6/dist-packages to the pythonpath a moment ago11:53
policecarwhich is not nice but it works for now11:53
sonne|workexactly what you should be doing :)11:53
policecarmaybe i missed the instruction11:53
policecarthanks for getting back to me!11:54
sonne|workpolicecar: well no - in principle it should work out of the box but everyone is cooking his own soup11:54
policecaryep11:55
-!- policecar [~policecar@g231190110.adsl.alicedsl.de] has quit [Quit: policecar]14:29
-!- policecar [~policecar@g231190110.adsl.alicedsl.de] has joined #shogun14:40
-!- policecar [~policecar@g231190110.adsl.alicedsl.de] has quit [Client Quit]14:43
-!- policecar [~policecar@g231190110.adsl.alicedsl.de] has joined #shogun15:05
-!- policecar [~policecar@g231190110.adsl.alicedsl.de] has left #shogun []15:06
-!- blackburn [~blackburn@188.168.4.253] has joined #shogun15:12
blackburnsonney2k: so is GLPK the only way to make MKL regression work?20:03
CIA-37shogun: Soeren Sonnenburg master * rbd6dcde / (5 files in 3 dirs):20:07
CIA-37shogun: get ruby back to work20:07
CIA-37shogun: - add configure option --ruby to be able to specify the ruby version to be used20:07
CIA-37shogun: - use rb_cObject instead of ruby_class to fix undefined symbol error20:07
CIA-37shogun: - beautify example - http://git.io/5PAJog20:07
@sonney2kblackburn, no - we have mkl_regression_3sine_waves.m for example20:08
@sonney2kso it works with regression too20:08
blackburnsonney2k: what is the problem then?20:08
@sonney2kI need to reboot again to check if things work under osx20:08
@sonney2kcu20:08
blackburnok20:08
@sonney2kblackburn, yes works all nice now20:19
@sonney2k:D20:19
blackburnsonney2k: ruby?20:20
@sonney2kyes20:20
blackburnnice20:20
@sonney2kreally weird hack I am using...20:22
@sonney2kwe no longer link to narray.so20:22
@sonney2kbut instead open it with dlopen20:22
@sonney2kand then extract the few needed functions20:22
@sonney2kwith dlsym20:22
@sonney2kregarding mkl + regression I suspect kernels are not properly normalized20:22
blackburnsonney2k: similar to java I would say?20:23
blackburnsonney2k: have you seen tikz? completely crazy stuff http://www.texample.net/tikz/examples/seismic-focal-mechanism-in-3d-view/20:24
@sonney2kone potential fix would be to use the sqrtdiag kernel normalizer20:25
@sonney2klatex wizardry...20:26
blackburnsonney2k: the problem I have now is after changing script there is an index error..20:27
blackburni.e. before train one kernel is used20:28
@sonney2kI am not sure what you are saying20:28
blackburnsonney2k: mkl is trained with one combined kernel20:28
blackburnthen another is set20:28
@sonney2kyes20:28
blackburnand apply is called20:28
@sonney2kyes20:28
blackburnand I receive index errors20:28
blackburnin apply20:28
@sonney2kin the example I pushed?20:28
blackburnno20:28
@sonney2kin which?20:29
blackburnin the example I sent to you with dropbox20:29
blackburnif you have time please take a look20:29
blackburnhmm20:30
blackburnthen if I change this line:20:30
blackburncols,rows = numpy.meshgrid(test_idxs2,train_idxs2)20:30
blackburnit segfaults..20:30
@sonney2kwhich line?20:30
@sonney2kwhich file20:31
blackburnmkl_regression_cv.py20:31
blackburn7720:31
@sonney2kdamn complex example20:32
blackburnsonney2k: sorry, it is a port of example madeleine uses20:32
blackburnin java it was even worse20:32
blackburnsonney2k: I rewrote whole example trying to solve her issue20:33
blackburnand it leaded to another issues..20:33
@sonney2kGLPK can only be used with mkl_norm=120:33
@sonney2kgreat20:33
@sonney2kI guess this is kind of spaghetti research code20:33
blackburnsonney2k: it is CV in fact20:34
@sonney2kblackburn, why are you using meshgrid20:34
blackburnsonney2k: what to use?20:34
@sonney2kjust index oeprations20:34
@sonney2koperations20:35
blackburnehmm won't it be diags?20:35
@sonney2klike m[idx][:,idx] ?20:35
shogun-buildbotbuild #235 of ruby_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/23520:36
blackburnsonney2k: is it a share file? did you file changed anyhow?20:36
blackburnyour*20:36
@sonney2ksry but I don't have time now - just ask I will answer late20:37
@sonney2kare20:37
blackburnsonney2k: ok sure20:38
@sonney2kblackburn, any news?23:49
blackburnsonney2k: no, I'm still pretty busy with my semester work :(23:52
blackburnsonney2k: ah yes, I've got with segfault labels23:53
blackburnwith this example23:53
blackburn==15121== Invalid read of size 123:54
blackburn==15121==    at 0x8B2F49F: make_contiguous(_object*, int*, int, int, bool) (modshogun_wrap.cxx:4177)23:54
blackburn==15121==    by 0x8E56143: _wrap_new_Labels__SWIG_2 (modshogun_wrap.cxx:261084)23:54
blackburn==15121==    by 0x8E5671C: _wrap_new_Labels (modshogun_wrap.cxx:261181)23:54
@sonney2kimpressive that you always find sth new...23:56
@sonney2kno way to do anything about it w/o seeing line 417723:56
blackburnsonney2k: I guess it is free flag issue23:57
blackburn  4177     if (PyArray_ISFARRAY(ary) && !force_copy)23:58
blackburn  4178     {23:58
blackburn  4179         array = ary;23:58
blackburn  4180         *is_new_object = 0;23:58
blackburn  4181     }23:58
blackburn  4173 static PyObject* make_contiguous(PyObject* ary, int* is_new_object,23:59
blackburn  4174                                int dims, int typecode, bool force_copy=false)23:59
blackburn  4175 {23:59
blackburn  4176     PyObject* array;23:59
blackburnsonney2k: not really better I would say :)23:59
--- Log closed Thu Nov 24 00:00:13 2011

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