Open in new window / Try shogun cloud
--- Log opened Fri Mar 14 00:00:08 2014
--- Day changed Fri Mar 14 2014
-!- fs_ [] has quit [Remote host closed the connection]00:00
shogun-buildbot_build #501 of precise - libshogun is complete: Success [build successful]  Build details are at
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has joined #shogun00:10
-!- zxtx [] has joined #shogun00:21
-!- sonne|osx [] has quit [Quit: sonne|osx]00:24
-!- PirosB3 [~pirosb3@] has joined #shogun00:33
PirosB3hey wiking did you have a look at my answer on the ML?00:33
PirosB3did you hear from Mark at all?00:33
shogun-buildbot_build #221 of clang34 - undefined behaviour analysis is complete: Failure [failed compile test]  Build details are at  blamelist: Thoralf Klein <>, Thoralf Klein <>01:06
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has quit [Ping timeout: 240 seconds]01:30
-!- zxtx [] has quit [Ping timeout: 255 seconds]01:40
-!- _flatline_ [] has joined #shogun01:42
-!- zxtx [] has joined #shogun01:43
-!- zxtx [] has quit [Ping timeout: 255 seconds]01:48
-!- PirosB3 [~pirosb3@] has quit [Ping timeout: 240 seconds]01:50
-!- lisitsyn [~lisitsyn@] has quit [Ping timeout: 240 seconds]01:50
-!- zxtx [] has joined #shogun01:56
-!- shogun-notifier- [] has quit [Quit: transmission timeout]02:19
-!- zxtx_ [] has joined #shogun02:25
-!- zxtx [] has quit [Ping timeout: 240 seconds]02:31
shogun-buildbot_build #221 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at  blamelist: Heiko Strathmann <>, Thoralf Klein <>, Thoralf Klein <>03:40
-!- soumyaC [uid15286@gateway/web/] has quit [Quit: Connection closed for inactivity]03:45
-!- Saurabh7 [~Saurabh7@] has joined #shogun04:03
-!- Saurabh7 [~Saurabh7@] has quit [Quit: Leaving]05:05
-!- _flatline_ [] has quit [Ping timeout: 264 seconds]05:52
-!- abinash_panda [0e8b2997@gateway/web/freenode/ip.] has joined #shogun06:07
-!- kprah [67157d4c@gateway/web/freenode/ip.] has joined #shogun06:09
-!- witness___ [uid10044@gateway/web/] has joined #shogun06:14
-!- kprah [67157d4c@gateway/web/freenode/ip.] has quit [Quit: Page closed]06:25
-!- zxtx_ [] has quit [Ping timeout: 255 seconds]06:31
-!- abinash_panda [0e8b2997@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]06:38
-!- zxtx_ [] has joined #shogun07:19
-!- fs_ [] has joined #shogun07:31
-!- [paladin] [~udit@] has joined #shogun08:41
-!- fs_ [] has quit [Remote host closed the connection]09:03
-!- HeikoS1 [~heiko@] has joined #shogun09:54
-!- HeikoS1 [~heiko@] has quit [Client Quit]09:54
-!- HeikoS1 [~heiko@] has joined #shogun10:00
-!- soumyaC [uid15286@gateway/web/] has joined #shogun10:01
-!- HeikoS1 [~heiko@] has quit [Quit: Leaving.]10:05
-!- HeikoS1 [~heiko@] has joined #shogun10:07
-!- HeikoS1 [~heiko@] has quit [Ping timeout: 252 seconds]10:20
-!- HeikoS1 [~heiko@] has joined #shogun10:23
-!- HeikoS1 [~heiko@] has quit [Ping timeout: 252 seconds]10:40
-!- travis-ci [] has joined #shogun10:43
travis-ci[travis-ci] it's Thoralf Klein's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []10:43
-!- lisitsyn [] has joined #shogun10:46
-!- khalednasr [2945be63@gateway/web/freenode/ip.] has joined #shogun12:04
-!- lambday [67157f54@gateway/web/freenode/ip.] has joined #shogun12:06
-!- HeikoS1 [] has joined #shogun12:07
HeikoS1[paladin]: hey there12:10
-!- blaze [7aacf93e@gateway/web/freenode/ip.] has joined #shogun12:13
-!- blaze is now known as Guest6493712:14
-!- PirosB3 [~pirosb3@] has joined #shogun12:20
Guest64937hey everyone.. I am a beginner.. I have setup shogun with DCMAKE_BUILD_TYPE=Debug -DENABLE_TESTING=ON flags. whenever i make any changes to library class say KernelPCA , do I have to rebuild the project for the changes to reflect..?12:25
@wikingGuest64937: obviously as any other cpp project12:26
lambdayGuest64937: yeah12:26
Guest64937wiking: ok thanks..12:27
PirosB3hi guys, do I need a specific flag on cmake to compile with OpenMP?12:27
lambdayPirosB3: I think cmake detects it automatically and compiles it with it whenever present12:28
lambdaywiking: ^ ?12:28
PirosB3lambday: is there any way I can make sure it has been compiled?12:29
PirosB3other than recompiling :)12:29
lambdayPirosB3: umm...12:29
lambdayPirosB3: I just did a grep -rin openmp in my build directory - it shows a lot of results12:31
PirosB3I will do that now12:31
lambdayyou should grep for "-fopenmp" :P12:31
lambdayPirosB3: I guess ./src/shogun/CMakeFiles/shogun.dir/link.txt should contain -fopenmp12:32
lambdaywhere . is you build dir12:32
PirosB3thanks lambday12:32
PirosB3also, are Python binding necessary for most of the Shogun development? I am STILL having issues in compiling the modular binding..12:33
PirosB3it's really giving me a hard time. And I cannot upgrade mac just now (10.7.5 currently)12:33
lambdayPirosB3: well, for GSoC its sort of necessary12:33
PirosB3especially for my project. I really do not know what to do..12:34
lambdayPirosB3: can you get hold of any linux box? (bad solution I agree)12:34
PirosB3I tried installing latest clang from homebrew, but there are other incompatibility issues over there12:34
PirosB3lambday: I can, but the emulated environment is tooo slow12:34
lambdayPirosB3: ought to be...12:34
lambdayno I mean dedicated linux box... not virtual box12:35
-!- newlido [~quassel@] has joined #shogun12:35
PirosB3lambday: unfortunately not really. I can when I am working on GSoC because I will be at home12:35
lambdayPirosB3: no worries then12:35
PirosB3really, I should just upgrade my mac12:35
PirosB3the big mess is that ALL my coruseworks are in when GSoC closes!12:36
PirosB3I have thought of this, and it's too risky to update now12:36
PirosB3I am trying to find workarounds12:36
jiaolongHi, one question about the shogun compiling, after updating the code, it seems the python module still compiles with the old data12:36
jiaolongmake[2]: *** [src/interfaces/python_modular/CMakeFiles/_python_modular.dir/modshogunPYTHON_wrap.cxx.o] Error 112:36
lambdayPirosB3: no worries12:36
@wikingPirosB3: autodetected (openmp)12:36
PirosB3can anyone give me any suggestions regarding this?
lambdayjiaolong: old data??12:37
PirosB3jiaolong: oh no you too! what what error do you get?12:37
jiaolongYes, I mean, I wrote a function in the old file12:37
@wikingPirosB3: not necessary but would be good on the end to have modular interfaces for the stuff you've created...12:37
jiaolongI always compile the interfeces12:38
@wikingjiaolong: what's old that?12:38
PirosB3wiking: to debug the memory issues I need to have modular binding12:39
PirosB3so I MUST get it solved12:39
@wikingPirosB3: you've already fired this as an issue... so i guess you should get an answer there...12:39
@wikingPirosB3: what?12:39
@wikingPirosB3: why would you need to have modular binding if u wanna resolve a memory issue?12:39
PirosB3for my ticket, regarding the python script12:39
@wikingPirosB3: unless you want to debug the modular binding itself...12:39
PirosB3because the issue is related to SWIG I assume12:40
@wikingPirosB3: a virtualbox could help you out until you dont have a solution for your issue on osx12:40
jiaolongI switch to the develop branch and compiling reports an error which points to my old branch12:40
lambdayjiaolong: you did do a rm -rf * and fresh cmake, right?12:41
@wikingjiaolong: the best in that case is simple recompile it from scratch....12:41
@wikingjiaolong: see lambday's comment12:41
lambdayjiaolong: could you please try that? should work :)12:41
@wikingPirosB3: anyways debugging memory with valgrind on osx is not working...12:42
jiaolonghow to do that? rm -rf * in build folder?12:42
@wikingjiaolong: yeah or rm -rf build12:42
lambdayjiaolong: yeah ... delete everything inside build12:42
jiaolonghaha, oh my god12:42
@wikingand then mkdir build12:42
@wikingwhatever fits you better12:42
lambdaylol :D12:42
jiaolonga lot of time, :(12:42
@wikingjiaolong: what?12:43
jiaolongI am sure that would work12:43
@wikingjiaolong: compiling?12:43
jiaolongno better way for re-compiling12:43
@wikingjiaolong: install ccache12:43
PirosB3ccache is amazing12:43
jiaolongIt seems I have installed ccache12:43
@wikingjiaolong: well then just remove the build directory re-run cmake and do make12:44
@wikingthat should be much faster than it was for the first time12:44
@wikingit ccache was used12:44
jiaolongOK, thanks12:44
PirosB3I just wonder, for hdf5 support on Ubuntu. do I need to install anything?12:44
lambdaywiking: you mean, instead doing rm -rf * inside build, if one removes and recreates build then it would be faster?12:45
@wikinglambday: it's the same :)12:46
@wikingPirosB3: yeah the -dev packages12:46
lambdaywiking: hmm yeah I thought I was missing something :)12:46
lambdayHeikoS1: hi!12:47
HeikoS1lambday: how are things?12:48
-!- kislay [~kislay@] has joined #shogun12:48
PirosB3do I also need libhdf5-openmpi-dev?12:48
lambdayHeikoS1: utterly confused :(12:48
HeikoS1lambday: I just saw your mail ...12:48
lambdayHeikoS1: weird that they say something in the paper but have something else in the implementation :(12:48
HeikoS1lambday: yeah well12:48
HeikoS1lambday: it is not *that* important acutally12:48
HeikoS1lambday: just important to do things right, but there are multiple ways12:48
HeikoS1lambday: I will actually talk to Dino in a bit (I just got back from Germany)12:49
lambdayHeikoS1: but they are not actually computing h(z_a, z_b)12:49
HeikoS1so we will write a response soon12:49
lambdayHeikoS1: oh great! How was it :)12:49
HeikoS1lambday: very nice!12:49
HeikoS1lambday: maybe its even a bug ;) but who knows .....12:49
lambdayHeikoS1: in their code?12:50
HeikoS1lambday: are you idling without that thing being clear? then I could discuss this now. otherwise this afternoon12:50
lambdayHeikoS1: nono... I am reading papers :D12:50
HeikoS1lambday: ok!12:51
HeikoS1lambday: dino also set up some other things that are good for the beginning12:51
-!- Guest64937 [7aacf93e@gateway/web/freenode/ip.] has quit [Quit: Page closed]12:51
HeikoS1see his issues12:51
HeikoS1but yeah, well get back to you later today12:52
lambdayHeikoS1: alright... I am checking12:52
-!- PirosB3 [~pirosb3@] has quit [Quit: PirosB3]13:19
[paladin]HeikoS1: Hey. Sorry for the delay.13:22
HeikoS1[paladin]: no worries, but lets talk later, I have to go now for a couple of hours, back later today13:23
HeikoS1see you13:23
[paladin]HeikoS1: okay, I've sent a mail, anyways.13:23
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has joined #shogun13:29
-!- [paladin] [~udit@] has quit [Quit: leaving]13:41
-!- kislay [~kislay@] has quit [Read error: Connection reset by peer]14:01
-!- vperic [~vperic@] has joined #shogun14:08
-!- kislay [~kislay@] has joined #shogun14:13
-!- PirosB3 [~pirosb3@] has joined #shogun14:17
-!- fs_ [] has joined #shogun14:22
-!- khalednasr92 [~k.nasr92@] has joined #shogun14:30
-!- khalednasr92 [~k.nasr92@] has quit [Remote host closed the connection]14:37
-!- khalednasr [2945be63@gateway/web/freenode/ip.] has quit [Quit: Page closed]14:38
-!- khalednasr [~k.nasr92@] has joined #shogun14:39
-!- dhruv13 [~dhruv13@] has joined #shogun14:54
jiaolongHi, what is the 'push_back' function in SGVector? Any suggestions to get familiar with libshogun  faster? What I am doing is to check the online help document again and again. Sometimes, I have to try different functions for many times.14:54
PirosB3hi guys14:56
kislayhey, I was working with the idea on OpenCV integration.  Any starting point on how to work on adding an optional dependency for opencv in cmake.15:24
kislayI couldn't comprehend it fully :(15:25
-!- wavelander [~udit@] has joined #shogun15:29
-!- zxtx_ [] has quit [Ping timeout: 240 seconds]15:30
-!- wavelander [~udit@] has quit [Quit: leaving]15:37
-!- wavelander [~udit@] has joined #shogun15:37
-!- wavelander [~udit@] has left #shogun []15:40
-!- wavelander [~udit@] has joined #shogun15:40
-!- lisitsyn [] has quit [Quit: Leaving.]15:41
kislayHey lambday , could you help me with the issue.15:41
-!- wavelander [~udit@] has quit [Quit: leaving]15:48
-!- witness___ [uid10044@gateway/web/] has quit [Quit: Connection closed for inactivity]16:04
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has joined #shogun16:32
-!- vperic [~vperic@] has quit [Quit: Konversation terminated!]16:42
PirosB3kislay: what issue do you have16:46
-!- dhruv13 [~dhruv13@] has quit [Ping timeout: 252 seconds]16:47
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:50
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has joined #shogun16:51
-!- khalednasr [~k.nasr92@] has quit [Quit: Leaving]17:05
-!- PirosB3 [~pirosb3@] has quit [Ping timeout: 252 seconds]17:16
lambdaykislay: hey.. sorry I was away17:43
lambdaykislay: I am afraid I won't be of much help regarding cmake stuff :(17:43
lambdaykislay: may be try to get how existing things work?17:44
-!- PirosB3 [] has joined #shogun17:54
PirosB3hey guys, can someone explain to me what the parameter framework is used for?17:55
PirosB3it's for serialization purposes right?17:55
-!- dhruv13_ [~dhruv13@] has joined #shogun17:55
@sonney2kPirosB3, yes and for model selection / clone()18:00
PirosB3why can I see lots of castings to (SGString<float64_t>**)18:01
PirosB3string of type float64?18:01
kislaylambday, oh. I was away!18:06
kislayno, openCV is not integrated right now with shogun.18:07
kislayI just want to know what its meant by "adding an optional dependency for opencv in cmake"18:08
-!- srichakradhar [~srichakra@] has joined #shogun18:12
kislayPirosB3, hey sorry... I was away18:14
kislayI was trying to work on the issue regarding  #1700, where some preliminary steps were listed18:15
-!- zxtx_ [] has joined #shogun18:24
-!- dhruv13_ [~dhruv13@] has quit [Ping timeout: 240 seconds]18:33
-!- kislay [~kislay@] has quit [Quit: Leaving]18:47
PirosB3sonney2k: with regards to ticket: I should:18:51
PirosB31) add a new datatype for vector18:51
-!- dhruv13 [~dhruv13@] has joined #shogun18:51
PirosB32) Implement new methods for SGVector18:51
PirosB33) modify the save, load method18:52
PirosB3is that correct?18:52
cameo54321HeikoS1: Hi! I looked into how random number generation happens in shogun18:55
HeikoS1cameo54321: yeah?18:56
cameo54321HeikoS1: I also looked to random number generation pointed out by wiking18:57
cameo54321It turns out that shogun uses sfmt18:57
cameo54321for random number generation18:57
cameo54321but the link i mentioned above has a lot of implementations18:57
cameo54321directly from cplusplus18:57
cameo54321including the mersene twister18:58
HeikoS1cameo54321: wow I see18:59
HeikoS1cameo54321: please open an issue to discuss that and reference me and wiking with some suggestions how to fix that then. Could be a very nice intro task18:59
cameo54321One way is to keep the legacy way of how random number generation is done in shogun18:59
cameo54321The other way is to directly take the distributions from random number generation methods already mentioned in the link above19:00
cameo54321using sfmt and then eigen19:00
cameo54321A new issue? or discussion on the existing 1929?19:00
PirosB3HeikoS1: when you have finished with cameo54321, can I ask you a question regarding Parameter framework?19:01
HeikoS1PirosB3: sure shoot19:03
HeikoS1cameo54321: yeah, taking some external implementation might be a nice idea, but lets discuss that in the issue, feel free to guide the discussion, point out alternatives, their implementations etc19:03
PirosB3I am trying to understand more regarding From what I understand, I should:19:03
PirosB3 add a new datatype for vector (DataType)19:03
PirosB3implement the add_vector function with type SGVector19:04
PirosB3and then, override some methods (save, load) in Parameter19:04
PirosB3is that correct?19:04
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:05
HeikoS1I am not so sure there, let me read the issue19:05
PirosB3thx :)19:07
-!- lisitsyn [~lisitsyn@] has joined #shogun19:12
HeikoS1PirosB3: we want a method similar to19:13
HeikoS1void add_vector(SGSparseVector<bool>** param, index_t* length,19:13
HeikoS1const char* name, const char* description="")19:13
HeikoS1just for SGVector<T>19:14
HeikoS1a vector of SGVectors19:14
HeikoS1have a look at the above method to see how its done, should be easy19:14
PirosB3yep, but that not all of it19:14
-!- fs_ [] has quit [Ping timeout: 240 seconds]19:14
PirosB3you need to also add the vector to the switch statement19:15
PirosB3struct type19:15
HeikoS1file, line?19:15
-!- zxtx_ [] has quit [Read error: Operation timed out]19:16
PirosB3also, shouldn't I need to add another EStructType?19:16
dhruv13HeikoS1: wiking: Is there a clean way to access the idx'th feature's dim'th value cleanly?19:18
-!- wavelander [~udit@] has joined #shogun19:18
HeikoS1PirosB3: I see19:19
HeikoS1yes thats true19:19
HeikoS1PirosB3: than the task is not that easy in fact19:19
HeikoS1PirosB3: but first step is being able to register19:19
PirosB3HeikoS1: I think the best way of approaching this problem is stating with a unit test19:19
dhruv13HeikoS1: wiking: I've had to assume a CDenseFeature<float64_t> to do it19:19
dhruv13HeikoS1: wiking: can i be more generic?19:20
PirosB3but as I have never used the parameter framework before, i would need an example19:20
HeikoS1afk for 5 mins ....19:20
PirosB3hey dhruv13 hows it goin19:20
PirosB3are you doing id3?19:20
dhruv13PirosB3: okay, a little puzzled by the myriad of Feature type classed19:20
dhruv13PirosB3: id3?19:21
PirosB3it's crazy isn't it19:21
PirosB3decision tree19:21
dhruv13PirosB3: yeah! i found a way to do what i need to do, but it seems dirty19:21
dhruv13PirosB3: nope, still on kdtree19:21
-!- PirosB3 [] has quit [Quit: PirosB3]19:21
dhruv13besser82: hey!19:21
@besser82dhruv13, ho!  :D19:22
-!- PirosB3 [] has joined #shogun19:22
PirosB3yeah, really little progress here too19:22
dhruv13PirosB3: we need to keep at it! Gambatte!!19:23
dhruv13i was wondering, is there a reason this project is called shogun?19:24
HeikoS1PirosB3: yes unit test is the best start19:25
PirosB3HeikoS1: can you help me formulate a unit test to begin with?19:25
HeikoS1PirosB3: have a look at the existing unit tests for the ST_SPARSE things19:26
HeikoS1PirosB3: both implementation and testing should be very similar19:27
HeikoS1PirosB3: but careful, this is involved stuff, not sure whether this is a good entrance task19:27
PirosB3do you think I should document the ticket with some steps?19:27
PirosB3in this way, if I am doing something dumb, people could tell me so19:28
HeikoS1PirosB3: nobody except me really knows this stuff unfortunately, so you wont get a lot of feedback19:28
PirosB3good to know!19:28
PirosB3anyways thanks for your help HeikoS119:28
HeikoS1PirosB3: I in fact suggest to try something else. If you still want to do it, read code first to get a feeling for what is to do19:28
HeikoS1PirosB3: once you understand that, it should not be too hard.19:29
PirosB3HeikoS1: do you suggest any task related to the Shogun SaaS?19:29
-!- khalednasr [~k.nasr92@] has joined #shogun19:29
HeikoS1PirosB3: you need to do it very systematically, increasing step by step with unit tests, thats the best way to go in my experience for those things19:29
dhruv13HeikoS1: Hey! can you help me out with something?19:29
HeikoS1PirosB3: I cant really say something about SaaS, dont know that19:29
HeikoS1dhruv13: yes, sorry19:29
dhruv13HeikoS1: Please take a look at this:
HeikoS1dhruv13: accessing individual features is hard for non CDenseFeatures (I think)19:30
dhruv13HeikoS1: could the kdtree_get_pt() be any cleaner?19:30
HeikoS1dont reaturn NaN, throw an error, people should only askig for valid features19:31
HeikoS1why do you cast to float64_t19:31
HeikoS1I think this can be done for any dense features19:31
HeikoS1Oh and please use the get_feature_vector() method which returns SGVector19:32
HeikoS1cleaner to read19:32
HeikoS1dhruv13: String features also support get_feature_vector btw19:33
HeikoS1but its all discrete19:33
HeikoS1dhruv13: does that help?19:34
dhruv13HeikoS1: umm... a lot of questions...19:34
HeikoS1dhruv13: ok go19:34
HeikoS1dhruv13: dont have loads of time but a little bit :)19:34
dhruv13HeikoS1: 1) Since KDTrees make sense with number type features only, do i keep the CDenseFeatures assumption?19:35
HeikoS1dhruv13: yeah, but maybe do a generalisation to any type?19:35
HeikoS1dhruv13: or do you only want to be float64?19:35
dhruv13HeikoS1: 2) Since I will integrate this with KNN.cpp, I think I can get away with the nan19:35
HeikoS1dhruv13: but why do you want to do that?19:36
dhruv13HeikoS1: the same assumption was made in EuclideanDistance::distance()19:36
HeikoS1dhruv13: well, ok then19:36
HeikoS1although I dont like it19:36
HeikoS1since its easy to check before19:36
dhruv13HeikoS1: oh, the nan i can change19:37
HeikoS1Thats the same as checking for NaN19:37
HeikoS1dhruv13: ok then, stay 64 bit19:37
dhruv13I meant the float64_t assumption19:37
HeikoS1otherwise its also annoying for the modular interfaces19:37
HeikoS1RealEuclideanDistance, ShortEuclideanDistance, thats ugly19:37
dhruv13so i will change the nan to throwing errors, and keep the CDenseFeatures<float64_t>19:38
dhruv13is that fine?19:38
HeikoS1yeah, thats good I think19:38
dhruv13cool! I was worried that i was not being generic enough19:38
HeikoS1dhruv13: maybe talk to somebody else about this, to get a second opinion19:38
dhruv13was stuck with that for a couple of days :-(19:38
dhruv13HeikoS1: okay... I' mostly looking at iglesiasg's implementation for Covertrees for inspiration19:39
dhruv13HeikoS1: most of my interface assumptions come from there... I will test this and get a PR ready soon19:40
dhruv13HeikoS1: Thanks a lot!!19:40
PirosB3hey guys how do I run the unit test framework=?19:43
PirosB3I do make test, but says "no target found"!19:43
dhruv13PirosB3: go to $BUILD_DIR/tests/unit/19:44
dhruv13PirosB3: and run ./shogun_unit_test19:44
PirosB3oh nice thx19:44
-!- wavelander is now known as [paladin]19:44
dhruv13PirosB3: do ./shogun_unit_test --help to check options19:44
dhruv13PirosB3: you can filter the tests you want to run19:45
-!- dhruv13__ [~dhruv13@] has joined #shogun19:54
-!- dhruv13 [~dhruv13@] has quit [Ping timeout: 240 seconds]19:56
-!- dhruv13__ [~dhruv13@] has quit [Ping timeout: 240 seconds]20:02
-!- dhruv13__ [~dhruv13@] has joined #shogun20:15
-!- fs_ [] has joined #shogun20:17
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has joined #shogun20:18
-!- sonne|osx [] has joined #shogun20:21
dhruv13__wiking: are you there?20:37
dhruv13__HeikoS1: are you free to answer something?20:39
-!- thoralf [] has joined #shogun20:46
thoralfHey everybody.20:46
thoralfPirosB3: Hey!20:47
PirosB3hey thoralf :)20:47
thoralfPirosB3: I'll be here for appox. 30 minutes.  If you want me to help, then now would be a good time. ;)20:47
PirosB3so basically I was looking at the Parameters ticket20:49
PirosB3i'll link it20:49
PirosB3and I was trying to understand the best way to handle this20:50
PirosB3I would need to add a DataType20:50
PirosB3I would also need to modify Parameters add_vector to be compatible with SGVector20:51
-!- cameo54321 [27290613@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:52
PirosB3and then do some changes in copy()20:52
PirosB3and save20:52
PirosB3is that the correct way of doing it?20:52
-!- sonne|osx [] has quit [Quit: sonne|osx]20:54
dhruv13__thoralf: can you help me out with something?20:54
-!- srichakradhar [~srichakra@] has quit [Ping timeout: 240 seconds]20:54
thoralfPirosB3: Don't you want to debug the memory issues in the other ticket first? ;)20:57
thoralfdhruv13__: Depends.20:57
thoralfdhruv13__: Just ask. ;)20:57
PirosB3oh sure! thoralf did you pull down my branch20:57
-!- dhruv13_ [~dhruv13@] has joined #shogun20:57
PirosB3git remote add pirosb3
PirosB3git fetch pirosb3 feature/186820:59
-!- dhruv13__ [~dhruv13@] has quit [Ping timeout: 240 seconds]20:59
PirosB3git checkout pirosb3/feature/186820:59
thoralfPirosB3: No, I will not pull your branch.  Can you simply just try to run gdb or valgrind on your script and tell, if you can gather more information?20:59
thoralfNo time to pull and compile everything, sorry.21:00
PirosB3thoralf: I am having some issues with valgrind and mac21:00
PirosB3I will try to virtualize it21:01
PirosB3that's the only thing I can do21:01
thoralfPirosB3: Having proper tooling is important... isn't there a package for valgrind?21:01
PirosB3yes, but it's full of issues on mac 10.7 unfortunately21:01
PirosB3the idea is that I will virtualize everything21:02
PirosB3thoralf: in get_confusion_matrix21:04
PirosB3there is a memset21:04
PirosB3where is that freed? is it up to us to free it?21:05
thoralfThe freeing happens automatically.21:06
lisitsynwhat memset has to do with free'ing ;)21:06
PirosB3in fact21:07
thoralfWe have a framework for this - it's done by SGRefObject.21:07
PirosB3lol :)21:07
thoralflisitsyn: :)21:07
PirosB3shogun has it's own memory management correct?21:07
thoralfPirosB3: Calling SG_REF/SG_UNREF when we want to in-/decrement usage counters21:07
PirosB3could that interfere with SWIG?21:07
thoralfWhen it's 0, then it will be freed automatically.21:07
thoralfPirosB3: I think SWIG handles this.21:08
lisitsynwell we tell swig to call our ref/unref21:09
PirosB3thoralf: well then there is free called twice21:10
PirosB3where is it!21:10
lisitsyn 5 %feature("ref")   shogun::SGRefObject "SG_REF($this);"21:12
lisitsyn 6 %feature("unref") shogun::SGRefObject "SG_UNREF($this);"21:12
thoralfthat't why we need this stacktrace. ;)21:12
lisitsynPirosB3: src/interfaces/modular/SGBase.i21:12
PirosB3you know what's funny lisitsyn, if I put a breakpoint just before get_confusion_matrix21:13
PirosB3I can call it once (from debugger)21:13
lisitsynwhat's going on?21:13
PirosB3then I call it again, at I have the issue21:13
lisitsynI do not understand the issue21:13
PirosB3lisitsyn: the issue is that there is a double free21:14
PirosB3when using the Python binding21:15
lisitsynPirosB3: uhm interesting21:18
PirosB3also, is there no way to init_shogun in the modular bindings? it would be good to redirect the output21:18
lisitsynPirosB3: not sure about that21:18
lisitsynPirosB3: okay to resolve that please compile with debug and run valgrind over it21:19
lisitsynit will be pretty clear21:19
lisitsynI glanced over this method and can't see the error21:20
lisitsynI wouldn't expect it is something with typemaps21:21
lisitsynthey are pretty good already so the error is somewhere else21:21
PirosB3now it works21:21
thoralfHeisenbugs :)21:21
PirosB3yeah but it's not stable21:21
PirosB3how can this be stable21:22
lisitsynrace condition or whatever like that21:22
PirosB3woohooo confusioon matrix works!21:22
thoralfBebugging has side effects.21:22
PirosB3lisitsyn: py is single threaded tho21:22
PirosB3thats what I thought21:22
lisitsynpython is but not shogun21:23
PirosB3do you want to see my confusion matrix!21:23
PirosB3[[ 34  21   1   0   0]21:23
PirosB3 [ 28 438 159   0   0]21:23
PirosB3 [  3 253 844  58   0]21:23
PirosB3 [  0   0   3  63   9]21:23
PirosB3 [  0   0   0   6   2]]21:23
PirosB3:) sorry21:23
lisitsyn5th class is pretty bad :D21:23
PirosB3anyways, lisitsyn is shogun all OpenMP?21:23
lisitsynmostly pthread21:24
PirosB3can I ask you a question regarding a task?21:24
PirosB3i'll just ask21:24
thoralfCan I answer my questions myself? ;)21:24
PirosB3what is Q21:27
PirosB3(q statistic)21:27
PirosB3(q test?)21:28
lisitsynI guess that's some estimate on variance21:28
PirosB3I think I really need to pick up a couple of books21:29
-!- [paladin] [~udit@] has quit [Quit: leaving]21:29
thoralfPirosB3: Okay, since the memory problem is away now, please rename `` to `` (or ``).  And it should be done.  But in order to try it myself, you can make a small comment on how to invoke and which shogun-data files to use as an example.21:29
thoralfThen it should be done.21:29
PirosB3I'll add it into the ticket21:30
lisitsynPirosB3: that's pretty narrow thing on these MMD deals21:30
thoralfPirosB3: If the memory issue comes back, I'll visit you at home and force you to fix it. :D21:30
lisitsynso you probably should not care that much ;)21:30
PirosB3ahaha, are you sure? it's cold over here!21:31
thoralfPirosB3: Where are you?21:31
PirosB3Berlin is so much better21:31
lisitsynwhat do you know about cold21:31
thoralflisitsyn: +10 degree is pretty cold! :)21:32
PirosB3lisitsyn: you probably know more about cold, but I know more about rain + cold + sudden change in temperature + humidity (all together)21:32
thoralfHaha.  Rainman. :)21:32
lisitsynyeah that's true21:32
lisitsynhere I just have it from -30 to +3021:32
dhruv13_india at this time of the year is +34... and don't get me started on humidity :-P21:34
dhruv13_i really don't know cold :-(21:35
PirosB3thoralf: pushed21:35
PirosB3dhruv13_: I want to go to india so much21:36
dhruv13_PirosB3: what parts?21:36
dhruv13_PirosB3: india is huge!21:36
PirosB3I have a really good friend in Punjab21:36
lisitsynwhat do you know about huge21:36
PirosB3I have an ex colleague from Kerela21:36
dhruv13_lisitsyn: ha ha21:36
PirosB3and I want to go to Ladakh21:36
dhruv13_PirosB3: cool places! I want to go to ladakh too :-P21:37
PirosB3it must be amazing..21:37
dhruv13_lisitsyn: where are you from?21:37
PirosB3thoralf: do you want the instructions on a gist, or shall i just add them to the ticket?21:37
lisitsyndhruv13_: from really huge country ;)21:38
dhruv13_lisitsyn: Russia?21:38
lisitsynno idea who needs that huge area though :D21:38
dhruv13_we seem to need every inch of ours :-P21:39
PirosB3ooh I did NYE with 3 russians21:39
PirosB3that was cool21:39
PirosB3actually 1 from belarus21:40
lisitsynwell he still could be russian21:41
dhruv13_CDistance::get_lhs() should really be CDistance::get_lhs() const21:41
lisitsyndhruv13_: actually yes21:41
lisitsynwe have this issue for well a long :)21:41
PirosB3what I do not get21:42
PirosB3is that these structures should be immutable21:42
dhruv13_lisitsyn: causing a lot of problems for my adaptor...21:42
PirosB3but then you have methods like Distance::init(CFeatures* l, CFeatures* r)21:43
PirosB3shouldn't it return a new copy?21:43
lisitsynthey are not immutable21:43
dhruv13_lisitsyn: is there a clean way to access feature vectors from distance objects? right now i have to assume i'm dealing with EuclideanDistance21:44
PirosB3lisitsyn: shouldn't they be?21:44
lisitsynPirosB3: they never intended to be ;)21:44
lisitsyndhruv13_: you may get features from distance21:44
lisitsynthen get feature vector from features21:45
dhruv13_lisitsyn: i mean feature idx's dim'th element21:45
lisitsynPirosB3: immutability would be like a great change for whole library21:46
lisitsyna bit too much :D21:46
dhruv13_i'm doing distance->get_lhs()->get_feature_vector() in an inline function21:46
lisitsyndhruv13_: oh I see21:46
dhruv13_oh, do i have to unref the sgvector?21:46
lisitsynwhy do you need to access some feature vector element?21:46
lisitsyndhruv13_: on21:46
dhruv13_nanoflann integration, trying to crate an adaptor21:47
lisitsynsgvector acts like a shared_ptr21:47
lisitsynsorry I am off for today ;)21:48
lisitsyndhruv13_: what I can suggest is to try to avoid accessing that frequently21:48
PirosB3ok see you21:48
lisitsynso that you just get whole vector instead of accessing this vector in a loop21:49
dhruv13_lisitsyn: bye! but I can't seem to find a work around21:49
-!- HeikoS1 [] has quit [Quit: Leaving.]21:50
dhruv13_lisitsyn: yeah, it's really bad right now...21:50
thoralfPirosB3: Just put the instruction on the ticket.21:50
thoralfPirosB3: Should be only 3 lines, right?21:51
PirosB3train, predict, evaluate21:52
thoralfIs there a suitable example multilabel file in data?21:54
-!- dhruv13 [~dhruv13@] has joined #shogun21:54
dhruv13lisitsyn: hey! went offline for a while... can you think of a workaround? nanoflann calls this function quite often21:55
-!- dhruv13_ [~dhruv13@] has quit [Ping timeout: 252 seconds]21:56
dhruv13thoralf: any ideas?21:58
thoralfdhruv13: Don't even know what this is about, sorry.22:00
dhruv13thoralf: oh, sorry; i'm trying to access individual feature vectors from a distance object in an inline function22:01
thoralfdhruv13: No, not my domain, sorry.22:01
dhruv13thoralf: okay :-)22:01
-!- zxtx [] has joined #shogun22:04
-!- soumyaC [uid15286@gateway/web/] has quit [Quit: Connection closed for inactivity]22:05
-!- newlido [~quassel@] has quit [Remote host closed the connection]22:13
-!- kislay [~kislay@] has joined #shogun23:05
-!- thoralf [] has quit [Ping timeout: 240 seconds]23:21
-!- thoralf [] has joined #shogun23:21
--- Log closed Sat Mar 15 00:00:25 2014