IRC logs of #shogun for Sunday, 2011-08-21

« Saturday, 2011-08-20 Index Monday, 2011-08-22 »
--- Log opened Sun Aug 21 00:00:07 2011
--- Day changed Sun Aug 21 2011
@sonney2kblackburn, so then you know what to attack...00:00
@sonney2kbut hey00:00
blackburnsonney2k: could you suggest me parameters for promoter kernel?00:00
blackburn"degree": 4, "shifts": 0, "kernel_cache": 500, "center_offset": 3, "center_pos": 100:00
CIA-87shogun: Soeren Sonnenburg master * re8bff92 / (src/shogun/base/Parallel.cpp src/shogun/base/Parallel.h): don't call library function in header file Parallel.h -> move it to .cpp - http://git.io/kRt-tQ00:00
@sonney2kdegree 2000:00
@sonney2kshifts - the more the better00:01
@sonney2k(expensive though)00:01
blackburnis 10 ok?00:01
@sonney2krest I didn't use just00:01
@sonney2kit will work00:01
blackburnchris gave me some code with some promoter kernel00:02
blackburnas kpca is ready I'll try to apply it00:02
@sonney2kblackburn, all of CMath call library functions from the .h00:03
@sonney2k*sigh*00:07
blackburnbad00:07
blackburnsonney2k: ohhh.. I can't apply kpca to string data00:07
blackburnwe can't change 'domain'00:07
blackburnDAMN00:07
blackburnwell I can but it is a little dirty trick00:07
@sonney2kblackburn, why not?00:07
blackburnbecause it changes feature matrix of features00:07
@sonney2kyou type - yeah right00:07
blackburnthe only way i can see is to delete given features00:07
@sonney2kyou mean00:07
blackburnand replace pointer00:07
@sonney2kno00:07
blackburnwith new simplefeatures instance00:07
@sonney2kthen better add a obtain_from_kpca function to simplefeatures00:08
blackburnmay be apply_to_string_features?00:08
blackburnin kpca00:08
@sonney2kwhy that?00:09
blackburnI don't understand what obtain_from_kpca function should do00:09
blackburnI already did apply_to_distance in MDS and isomap00:09
blackburnit looks convenient for me to do that00:09
@sonney2kblackburn, it would return transformed features based on the preprocessor and its feature object00:10
@sonney2kI really need to sleep now00:10
blackburnsure00:10
blackburnme too00:15
-!- f-x [~user@117.192.222.10] has quit [Read error: Connection reset by peer]00:59
-!- f-x [~user@117.192.222.10] has joined #shogun01:00
blackburnhttp://dl.dropbox.com/u/10139213/shogun/image.png01:01
CIA-87shogun: Sergey Lisitsyn master * r8d8cfd8 / (2 files): Fixed m_eigenvalues handling in MDS - http://git.io/NGs7wQ01:01
CIA-87shogun: Sergey Lisitsyn master * r0bf8b18 / (2 files): Added apply_to_string_features method to kPCA - http://git.io/EnlYGQ01:01
blackburnat least something01:01
-!- blackburn [~blackburn@188.122.238.50] has quit [Quit: Leaving.]01:01
-!- alesis-novik [~alesis@188.74.87.206] has joined #shogun01:05
-!- f-x` [~user@117.192.222.10] has joined #shogun01:42
-!- f-x [~user@117.192.222.10] has quit [Ping timeout: 260 seconds]01:45
-!- f-x` [~user@117.192.222.10] has quit [Remote host closed the connection]02:57
CIA-87shogun: Shashwat Lal Das master * r33ba646 / (5 files): Added exact adaptive norm to VW and corrected train declaration. - http://git.io/-oWTeA06:43
CIA-87shogun: Soeren Sonnenburg master * r23edad1 / (5 files):06:43
CIA-87shogun: Merge pull request #305 from frx/streaming_vw06:43
CIA-87shogun: Added exact adaptive norm to VW and corrected train declaration. - http://git.io/JYBnKw06:43
CIA-87shogun: Soeren Sonnenburg master * rf14ecfb / (7 files in 4 dirs):06:44
CIA-87shogun: Merge pull request #306 from alesis/gmm06:44
CIA-87shogun: Few documentation additions - http://git.io/inyTUw06:44
CIA-87shogun: Soeren Sonnenburg master * red016f4 / src/configure : improve robustness in ruby detection - http://git.io/T9Ph7w07:15
-!- blackburn [~blackburn@188.122.238.50] has joined #shogun11:08
-!- f-x [~user@117.192.207.63] has joined #shogun13:01
-!- f-x [~user@117.192.207.63] has quit [Remote host closed the connection]13:54
CIA-87shogun: Sergey Lisitsyn master * r8631f29 / src/interfaces/modular/Preprocessor.i : Added %newobject for preprocessor routines creating new objects - http://git.io/ruGl6A14:55
-!- sino2 [72bc41d6@gateway/web/freenode/ip.114.188.65.214] has joined #shogun15:20
-!- sino2 [72bc41d6@gateway/web/freenode/ip.114.188.65.214] has quit [Ping timeout: 252 seconds]15:25
-!- f-x [~user@117.192.207.63] has joined #shogun15:29
-!- f-x [~user@117.192.207.63] has quit [Remote host closed the connection]15:48
CIA-87shogun: Justin Patera master * rbbde5a9 / (6 files): syncing my computers, completed more examples - http://git.io/zzRS2g17:10
CIA-87shogun: Justin Patera master * r0f8b2ff / (5 files in 2 dirs): added a 'modshogun.rb' file to do some (ok, eventually a lot) of the ruby start-up stuff, like requiring 'narray' and such. also completed more examples. more examples to come!! - http://git.io/fCAKGg17:10
CIA-87shogun: Justin Patera master * r9e4fe99 / (2 files in 2 dirs): added a fixer program to make life easier for me && modified the modshogun.rb file - http://git.io/eCOtuw17:10
CIA-87shogun: Justin Patera master * refa826b / (136 files): did more work converting files, having some errors and so pushing changes to get errors resolved. - http://git.io/pEIv2w17:10
CIA-87shogun: Soeren Sonnenburg master * r8a410d1 / (144 files in 2 dirs):17:10
CIA-87shogun: Merge pull request #304 from serialhex/master17:10
CIA-87shogun: a few additions to the ruby examples & some other stuff... - http://git.io/CaGxFg17:10
blackburnsonney2k: so are issues reported yesterday fixed now?17:40
CIA-87shogun: Sergey Lisitsyn master * r36776f5 / (9 files): Doc improvements for preprocessors, thread naming fixes - http://git.io/boxxOw18:00
@sonney2kblackburn, I don't know ... seems like the spinlock one at least not19:15
@sonney2kblackburn, btw you did close quite some bugs without writing why?19:15
@sonney2kI don't know is the gsoc ideas really all done now?19:15
blackburnsonney2k: well according to your words yes :D19:43
blackburnwell I'll reopen it if it is not done19:44
blackburnGMM is done by alesis19:45
blackburnoh19:45
blackburnI see19:45
blackburnstructured closed by mistake19:45
blackburnsonney2k: are ruby and java done?19:46
blackburncross-validation too19:46
blackburnvw and online learning are done by f-x19:46
@sonney2kblackburn, ok19:47
@sonney2kjava and ruby should work but we need more examples to run first to be sure19:48
@sonney2kyes19:48
blackburnsonney2k: yeah I should wrote something..19:48
@sonney2kblackburn, yes write something like implemented by f-x19:48
@sonney2kor alesis-novik19:48
@sonney2kotherwise we won't know19:48
@sonney2kwhy you closed the bug19:48
blackburnI'm afraid of spinlock detection still not working19:48
@sonney2kblackburn, I am writing it currently19:49
@sonney2kblackburn, could you please do the change in SGObject for spinlocks?19:49
blackburnsome define should work19:49
blackburnifdef I mean19:49
blackburnyeah sure19:49
blackburnbut a little bit later, I just came home from some jogging and need to take a bath :D19:49
@sonney2kblackburn, do we have posix thread detection?19:50
* sonney2k can smell this from over here19:50
blackburnyes19:50
blackburnin configure19:50
@sonney2k=:-)19:50
@sonney2kok19:50
blackburnline 278319:51
blackburnsonney2k: smell should be better now :D20:05
* sonney2k cleans his nose20:05
* sonney2k takes a deep breath20:05
@sonney2kyeah wonderful20:06
CIA-87shogun: Soeren Sonnenburg master * r15b3364 / src/shogun/base/Parallel.h : should be 0 instead of NULL - http://git.io/I8-zvA20:07
CIA-87shogun: Soeren Sonnenburg master * r15a636a / src/configure : add spinlock detection - http://git.io/wB1g8w20:07
blackburnah yes20:08
blackburnI tried to not make mistake in init20:08
blackburnbut made heh20:08
blackburnokay marked gsoc tasks20:14
serialhexYAY i got commited!!! :P20:15
blackburnsonney2k: I just noticed my headers are like MULTIDIMENSIONALSCALING_H_, but not __CMULTIDIMENSIONALSCALING_H__. does it matter?20:17
@sonney2kblackburn, not really20:20
blackburnokay20:20
@sonney2kblackburn, btw recall that we disabled all update issues in trac for non-authorized users because of spam? considering the latest bugreport with the user not being able to comment / answer - what do we do now?20:22
blackburnwell you could enable it20:23
blackburnI think we will manage to delete spam if necessary20:23
@sonney2kblackburn, ok lets give it a try then20:23
@sonney2kblackburn, ok done20:25
blackburn#define pthread_mutex_init(x)20:27
blackburn#define pthread_mutex_destroy(x)20:27
blackburn#define pthread_mutex_lock(x)20:27
blackburn#define pthread_mutex_unlock(x)20:27
blackburnoh20:27
@sonney2k?20:27
@sonney2kwhat are you doing?20:28
blackburnnot me20:28
@sonney2kme?20:28
blackburnit is in sgobject20:28
@sonney2kwhere?20:28
blackburnline 2220:28
blackburnSGObject.h20:28
blackburnwhat should we do if no pthreads there?20:29
@sonney2kmove this logic into parallel please20:29
blackburnit is not the logic, it is hack20:29
blackburnfor mutex things being compiled20:29
blackburnif no pthreads here20:29
blackburnsonney2k: so what to do if no pthreads20:30
blackburn?20:30
CIA-87shogun: Sergey Lisitsyn master * r385681b / src/shogun/base/Parallel.h : Changed lock routines to pointer usage - http://git.io/5039UQ21:11
CIA-87shogun: Sergey Lisitsyn master * red18465 / (src/shogun/base/SGObject.cpp src/shogun/base/SGObject.h): SGObject mutex to generic lock transition - http://git.io/VMJ_jA21:11
CIA-87shogun: Sergey Lisitsyn master * r85ce83d / (2 files): LTSA and HLLE fix for pthread generic routines - http://git.io/zIQbqQ21:11
@sonney2kblackburn, if no pthreads things should still work21:55
blackburni've guarded it with #ifdefs21:56
@sonney2kso ifdef'ing stuff is best - problem in the SGBase.h case is that one has plenty of ifdefs...21:56
blackburnI will introduce kernel LLE in 20-30 minute :D21:56
@sonney2kbtw. move the library routine based calls into the .cpp21:56
blackburnI became pretty active at the end of gsoc21:56
blackburnhah21:56
@sonney2kblackburn, too bad that you werent before ;-)21:57
blackburnI will move, later21:57
@sonney2kblackburn, move?22:00
blackburn(11:56:44 PM) sonney2k: btw. move the library routine based calls into the .cpp22:00
@sonney2kahh k22:01
CIA-87shogun: Sergey Lisitsyn master * rd7f688d / src/shogun/preprocessor/LocallyLinearEmbedding.cpp : Fix for LLE - http://git.io/f0S9CA22:03
@sonney2kserialhex, do you know where ruby extensions should be installed?22:15
@sonney2kI mean for system wide installation?22:15
@sonney2ksome /usr/local/lib/ruby?22:15
serialhexhmm... not generally no, i'll have to look, but the way i have it now is....22:16
serialhexsomething like that, hd on22:16
@sonney2kk22:16
serialhex/usr/lib/ruby/1.8/i486-linux is where i copy the modshogun files to in order for it to work (on my debian machines)22:19
@sonney2kserialhex, would it work in /usr/local/* too?22:20
serialhexhmm... i dunno22:20
@sonney2kserialhex, seems like:22:22
@sonney2k$ ruby -e 'puts $LOAD_PATH'22:22
@sonney2k/usr/local/lib/site_ruby/1.822:22
@sonney2k/usr/local/lib/site_ruby/1.8/x86_64-linux22:22
@sonney2k/usr/local/lib/site_ruby22:22
@sonney2k/usr/lib/ruby/vendor_ruby/1.822:22
@sonney2k/usr/lib/ruby/vendor_ruby/1.8/x86_64-linux22:22
@sonney2k/usr/lib/ruby/vendor_ruby22:22
@sonney2k/usr/lib/ruby/1.822:22
@sonney2k/usr/lib/ruby/1.8/x86_64-linux22:22
@sonney2kserialhex, so I will hack something up that works on make install22:23
serialhexYES!  i was looking for those!22:23
serialhexit wasn't working in irb for some reason... stupid thing...22:24
@sonney2kI am glad that my old friend google knew the answer :)22:26
serialhexyes!!22:26
serialhexand if you use that it should work when i use RVM or something similar for my ruby22:27
@sonney2kserialhex, problem is that I don't know which of the paths to use when including /linking22:27
serialhexerr... for peoples ruby22:27
@sonney2kinstalling even too22:28
serialhexyeah, i had that problem when i was trying to get thigns to work right in 1.8/1.9 & when using anything other than debian... lemme install rvm & a few rubies on my machine & we can work on this for a bit22:28
serialhex(der it's already installed as my user :P )22:31
serialhexok sonney2k, on line 686 of the configure script, the ruby *.h files should be found there22:37
serialhexerr, by that22:37
serialhex...and the previous 'puts ...' line too22:37
serialhexbut for some reason it dosn't find narray when it's installed as a gem... :(22:38
@sonney2kputs "#{Config::expand(CONFIG['sitelibdir'])}"22:46
@sonney2kthat would be the local install path22:46
@sonney2khmmhh22:47
@sonney2kI don't really know how to select the right one22:47
blackburnsonney2k: what about apply_to_feature_matrix -> apply_to_features?22:49
@sonney2kblackburn, well it doesn't apply to features but only to matrix (if available)22:50
blackburnthrea are already two methods possible to apply to features with no feature_matrix22:51
blackburnkpca and kernel lle22:51
blackburnmore to go22:51
@sonney2kserialhex, in CONFIG - can I somehow change prefix?22:52
@sonney2k$(prefix)22:52
serialhexprefix?? which prefix???22:53
@sonney2k--libexecdir=$(prefix)/lib/ruby1.822:53
@sonney2kis in there22:53
serialhexerm... i'm not sure i fully understand22:54
@sonney2kserialhex, type22:55
@sonney2kCONFIG['libdir']22:55
@sonney2kthere is $(exec_prefix) appearing there22:55
serialhexhuh, ok22:56
@sonney2kirb(main):010:0> CONFIG['libdir']22:56
@sonney2k=> "$(exec_prefix)/lib"22:56
@sonney2kirb(main):011:0> CONFIG['exec_prefix']22:56
@sonney2k=> "$(prefix)"22:56
serialhexyeah i see it now22:56
@sonney2kso if I could change prefix the install path would be correct22:56
serialhexyes, i know there is a better way to do that.. i'm looking22:57
serialhexor maybe a way to do exactly that :P22:58
@sonney2kbecause I have a prefix as arg in ./configure22:58
@sonney2kso I would just need to modify CONFIG['prefix'] with that22:58
@sonney2kand then cna print the exanded path and we are good to go22:58
-!- heiko1 [~heiko@541D5B0C.cm-5-6b.dynamic.ziggo.nl] has joined #shogun23:09
@sonney2kheiko1, so you immigrated to the netherlands?23:13
heiko1hej sonney2k23:13
heiko1no :)23:13
heiko1just a little workshop23:13
heiko1playing around with an quadcopter :)23:13
heiko110 kilometers from the frontier23:13
@sonney2kserialhex, any ideas?23:18
CIA-87shogun: Sergey Lisitsyn master * r0345166 / (5 files in 2 dirs): Introduced kernel LLE - http://git.io/LnQ14Q23:18
@sonney2kheiko1, ok so you can bike home ;-)23:18
serialhexirb(main):003:0> CONFIG["prefix"]23:19
serialhex=> "/usr"23:19
@sonney2kheiko1, so please just continue - I will then start to run the tests and hopefully conversions (give me one example please)23:19
@sonney2kserialhex, yeah but if I do CONFIG['prefix']='/tmp/foo'23:19
@sonney2kand then do puts "#{Config::expand(CONFIG['rubylibdir'])}"23:19
@sonney2kit still says23:19
@sonney2k/usr/lib/ruby/1.823:20
@sonney2keven though23:20
@sonney2kirb(main):003:0> CONFIG['libdir']23:20
@sonney2k=> "$(exec_prefix)/lib"23:20
@sonney2kirb(main):004:0> CONFIG['exec_prefix']23:20
@sonney2k=> "$(prefix)"23:20
@sonney2kirb(main):005:0> CONFIG['prefix']23:20
@sonney2k=> "/usr"23:20
serialhexhmm...23:20
@sonney2kirb(main):002:0> CONFIG['rubylibdir']23:20
@sonney2k=> "$(libdir)/ruby/$(ruby_version)"23:20
heiko1sonney2k, yes, but i dont have a bike here :) stying here23:21
serialhexok, lemme dig deeper...23:21
heiko1sonney2k, yes i am on to it :)23:21
@sonney2kheiko1, great23:21
heiko1its a bit complicated now to write the conversion methods, but i will give you an example as soon as it works23:21
@sonney2kserialhex, thanks - I will have to sleep now but if you manage to modify CONFIG or so in a way that I can just puts of the expanded string - I can immediately use it for configure / make install23:22
@sonney2kheiko1, ok thanks23:22
@sonney2kI took part in some long distance swimming event today (3.5km) - so I now need some sleep :)23:23
@sonney2kcu all23:23
blackburngood luck to you :)23:24
blackburnsee you23:24
heiko1ok have a good rest :)23:25
blackburnheiko1: hey what's up?23:26
heiko1hey blackburn, chilling in the netherlands and hunting bugs, and you? :)23:26
blackburnheh, chilling in russia, implementing more methods :D23:27
heiko1nice :)23:27
serialhexok, np sonney2k23:27
heiko1but i am a bit in a hurry here, hope i finish this soon23:27
heiko1heureca, first time program runs without memory errors, just leaks now :)23:29
blackburnwhat is it you are finishing?23:29
heiko1version conversion of parameters23:33
blackburnah23:33
heiko1to be able to load old serialized instances23:33
blackburnI guess it is pain in ass23:33
heiko1deeply23:33
heiko1but: first version working23:34
heiko1 :)23:34
heiko1only 20 frees missing in my example23:34
heiko1type conversion23:34
heiko1name conversion23:34
heiko1everything possible now23:34
heiko1but its very technical, a bit annoying though23:35
heiko1goind to bed23:43
heiko1good night all23:43
-!- heiko1 [~heiko@541D5B0C.cm-5-6b.dynamic.ziggo.nl] has quit [Quit: Leaving.]23:43
--- Log closed Mon Aug 22 00:00:12 2011
« Saturday, 2011-08-20 Index Monday, 2011-08-22 »

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