Open in new window / Try shogun cloud
--- Log opened Thu Mar 21 00:00:15 2013
-!- zxtx [zv@conference/pycon/x-myicjgrlhekelmvl] has quit [Ping timeout: 246 seconds]00:03
-!- SCV__ [] has joined #shogun00:09
-!- FSCV_ [~FSCV@] has quit [Ping timeout: 264 seconds]00:10
-!- FSCV_ [~FSCV@] has joined #shogun00:16
-!- SCV__ [] has quit [Ping timeout: 264 seconds]00:18
-!- zxtx [zv@conference/pycon/x-wxlbiewpetopbeko] has joined #shogun00:26
-!- naywhaya1e [] has joined #shogun00:30
-!- naywhayare [~ryan@] has quit [Ping timeout: 272 seconds]00:32
-!- zxtx [zv@conference/pycon/x-wxlbiewpetopbeko] has quit [Read error: Operation timed out]00:58
-!- zxtx [zv@conference/pycon/x-zkmfodkgegyjypci] has joined #shogun01:13
-!- zxtx [zv@conference/pycon/x-zkmfodkgegyjypci] has quit [Ping timeout: 264 seconds]01:47
-!- zxtx [zv@conference/pycon/x-zlefkqqgajgfzmmu] has joined #shogun01:59
-!- zxtx [zv@conference/pycon/x-zlefkqqgajgfzmmu] has quit [Ping timeout: 245 seconds]02:09
-!- FSCV_ [~FSCV@] has quit [Read error: Connection reset by peer]02:14
shogun-buildbot_build #328 of nightly_default is complete: Failure [failed test]  Build details are at
-!- zxtx [zv@conference/pycon/x-fuobffokswdzgwuc] has joined #shogun03:57
-!- zxtx [zv@conference/pycon/x-fuobffokswdzgwuc] has quit [Ping timeout: 258 seconds]04:24
-!- naywhaya1e is now known as naywhayare05:31
-!- zxtx [] has joined #shogun07:31
-!- blackburn [~blackburn@] has left #shogun []07:54
-!- blackburn [] has joined #shogun08:45
-!- blackburn [] has quit [Ping timeout: 260 seconds]08:54
-!- n4nd0 [] has joined #shogun08:59
-!- blackburn [] has joined #shogun09:06
sonne|workblackburn: I've tried to fix the mkl xval example09:17
sonne|workit is crashing when reading streaming features09:17
sonne|workI know the reason why but I have no idea how to fix it09:18
blackburnsonne|work: yeah streaming is not in a good state09:18
sonne|workwell reason is sgvector now is auto referenced09:18
sonne|workand that doesn't work with the way streaming is done09:18
blackburnyeah I remember that thing09:18
sonne|workI think we have to consider using row pointers there09:18
sonne|workblackburn: could you please register at gsoc - melange site?09:20
blackburnsonne|work: register mysefl? already did09:20
sonne|workblackburn: I would rather now start workign on the proposal09:20
sonne|workahh cool09:20
sonne|workwhat is your short name?09:20
blackburnsonne|work: let me check09:20
blackburnsonne|work: sergey_lisitsyn09:21
sonne|workI will put you as backup admin09:22
sonne|work(needed to register an org)09:22
blackburnsonne|work: I know09:22
blackburnsonne|work: I put some things to proposal before09:23
sonne|workblackburn: where?09:23
blackburnsonne|work: gdocs, shared with you09:24
sonne|workblackburn: btw olivier sent an abstract :)09:26
blackburnsonne|work: forward it to me, I'll polish09:26
blackburnsonne|work: what needs most care is 1) 2) and 4)09:27
n4nd0blackburn, sonne|work : what is his project about?09:29
blackburnn4nd0: hashing I think09:29
n4nd0nvm I can see the google docs too09:29
sonne|workn4nd0: the hashing trick, blackburn: I've put it in the ideas list09:29
n4nd0ah no the google docs is other thing hehe09:30
blackburnn4nd0: ?09:30
n4nd0blackburn: I thought the google docs was going to be the Olivier's abstract09:31
blackburnsonne|work: what about putting a shogun mentor to each non-shogun mentor?09:35
sonne|workblackburn: shogun mentor is you me heiko wiking?09:36
blackburnsonne|work: yes09:37
blackburnsonne|work: I'll co-mentor compressive sensing stuff09:37
blackburnsonne|work: can co-mentor LMNN by georg too09:38
blackburnsonne|work: anything you'd like to co-mentor?09:38
n4nd0blackburn: wiking could take a look to an hypothetical structured learning project, I think he knows about that stuff09:38
blackburnlet's put it as 'Shogun mentor'09:38
sonne|workI am very familiar with d3js and the hashing trick09:38
blackburnsonne|work: you familiar with d3js?09:39
blackburnsonne|work: we should knew that :D09:39
sonne|worksome SO too sure09:39
blackburnsonne|work: ICA?09:39
sonne|workwell blackburn ICA is close to PCA in some sense09:40
blackburnsonne|work: and that means?09:40
sonne|work(andreas would kill me for saying that)09:40
sonne|workyou should :)09:40
sonne|workI mean in the end it gives you a new basis (not necessarily fewer dims though)09:40
blackburnsonne|work: I'd put it in tapkee but not sure you'd like that :D09:41
sonne|workin this sense it is different then :P09:43
blackburnsonne|work: tapkee finally received t-SNE, one adapter and it would be in shogun09:44
blackburnsonne|work: *very* cool method for visualization09:44
blackburnunbelievably intelligent comparing to other methods09:44
blackburnwhich reminds me to put this task to entrance09:47
-!- Yanglittle [deb20afd@gateway/web/freenode/ip.] has joined #shogun10:15
YanglittleI have a problem when installing shogun, I run ./configure,but it can't Matlab I've installed.10:16
n4nd0Yanglittle: hey, so it doesn't recognize your matlab?10:16
Yanglittleany suggestions?10:17
n4nd0Yanglittle: what is your matlab version?10:18
n4nd0Yanglittle: so you are giving the flag to compile for the interface matlab static?10:19
YanglittleI installed all interface  .10:20
Yanglittleit can't run sg in matlab, so I run the instructions again ,and it says no matlab..10:21
n4nd0Yanglittle: are you in linux?10:22
Yanglittleyes, ubuntu 12.1010:22
n4nd0what does matlab tell you if you execute this10:22
n4nd0in your shell10:22
n4nd0matlab -nojvm -nodisplay -r 'version;quit'10:23
YanglittleK>> divide_by_std=false; K>> sg('add_preproc', 'PRUNEVARSUBMEAN', divide_by_std); Undefined function 'sg' for input arguments of type 'char'.10:23
n4nd0mm that is weird10:24
n4nd0I am trying to check if you can run matlab at all from command line10:24
n4nd0I don't understand why does it appear any sg there10:24
YanglittleI don't understand..10:25
n4nd0Yanglittle: mmm ok :)10:25
n4nd0Yanglittle: so have you executed this: matlab -nojvm -nodisplay -r 'version;quit' in your terminal?10:26
YanglittleI runned just now, it says no matlab command10:27
n4nd0Yanglittle: ok, that's why it is not found by configure script10:27
n4nd0you have to modify your path so matlab can be executed from terminal10:27
Yanglittleok  wait a moment10:28
n4nd0or you could also add a link to your matlab executable in a directory of your path such as /usr/local/bin10:28
n4nd0in any case I think this should be done automatically when you install matlab10:28
Yanglittle< M A T L A B (R) >                   Copyright 1984-2012 The MathWorks, Inc.                     R2012b ( 64-bit (glnxa64)                               August 22, 2012    To get started, type one of these: helpwin, helpdesk, or demo. For product information, visit
n4nd0Yanglittle: try ./configure --interfaces=matlab_static now10:30
YanglittleI can't make a soft link to matlab10:30
n4nd0Yanglittle: why not?10:30
n4nd0anyway if it is recognized now, it does not matter10:30
n4nd0I guess maybe you are installing in a system without permissions?10:31
Yanglittlestill can't find matlab !!10:32
YanglittleChecking for libshogun ... yes (internal in source tree) Checking for Matlab ... no10:33
Yanglittleit's boring, shoud I reinstall matlab ?10:34
n4nd0Yanglittle: how did you install it?10:34
n4nd0the configure script is looking for other things and if they are not found in the directories where they should be then it won't be recognized10:35
YanglittleI installed at the default position10:36
n4nd0Yanglittle: ok let's check where it can be failing then10:37
n4nd0Yanglittle: take a look to your configure.log file please10:38
n4nd0Yanglittle: there should be a section with == Checking for Matlab ==10:39
n4nd0and something should be said down there10:39
Yanglittle============ Checking for Matlab ============ Result is: no10:41
n4nd0hehe that's not very informative indeed :)10:41
Yanglittleand then nothing .10:42
n4nd0it looks to me like the configure script is still failing for you in matlab -nojvm -nodisplay -r 'version;quit'10:42
sonne|workYanglittle: yeah just run matlab -nojvm -nodisplay -r 'version;quit'10:43
n4nd0Yanglittle: you told me it worked though10:43
Yanglittleit can't find my matlab .10:43
sonne|workon your cmdline10:43
sonne|workn4nd0: oh you told him but still...10:43
n4nd0Yanglittle: I think it might be that you added in a terminal window to your path a directory so matlab could be found10:43
n4nd0and you are executing ./configure from another terminal10:44
n4nd0sonne|work: btw the thing of the SGVector constructor for the modular interfaces that accepts an SGVector without const10:45
n4nd0sonne|work: it is not working out so well10:45
YanglittleChecking for Posix thread spinlock support ... yes Checking for libshogun ... yes (internal in source tree) Checking for Readline support ... no Checking for Python numpy support ... 1.6.2 Checking for Python/Python Numpy Developer Files ... yes Checking for Octave ... no Checking for R ... no Checking for Matlab ... no Checking for Java ... not detected Checking for Mono C# ... not detected Checking for Ruby ... 1.9.3p194 Checki10:45
n4nd0Yanglittle: can you please check it you can run matlab from the same command line where you are running configure?10:46
sonne|workYanglittle: run matlab -nojvm -nodisplay -r 'version;quit' in the exact same terminal you ran configure10:46
sonne|workn4nd0: what happens?10:46
n4nd0sonne|work: the const was not the problem it seems, but the & - that it accepts an SGVector&10:46
Yanglittlecan't find matlab command10:47
n4nd0Yanglittle: you have to make it find matlab and run configure from the same terminal window...10:47
n4nd0Yanglittle: if you have changed your path you have to export that so the change is visible from all the system10:47
Yanglittlewhen run matlab command,does it need to change the path?10:48
n4nd0Yanglittle: no10:48
n4nd0the matlab binary has to be visible from somewhere in the path10:48
Yanglittleok i'll check the system path10:49
n4nd0Yanglittle: you said you get it work before from the command line right? to run matlab -nojvm -nodisplay -r 'version;quit'10:49
Yanglittlei changed the path to the matlab ,sorry.10:50
Yanglittleso  it works  ~10:50
n4nd0mm I don't really understand what you mean10:50
n4nd0not sure if you changed the PATH env variable10:50
n4nd0or cd to the matlab dir :)10:51
Yanglittleit works well under matlab dir, but other  fails10:51
n4nd0hehe sure10:51
n4nd0export PATH=$PATH:/your/matlab/dir10:52
sonne|workn4nd0: yes sure the & is the problem10:52
n4nd0where /your/matlab/dir has to substituted for your actual dir10:52
n4nd0sonne|work: I see, then the issue is that one cannot make a constructor SGVector(SGVector orig)10:52
sonne|workwhy not?10:52
n4nd0sonne|work: I don't really know why TBH but the compiler tells me no way10:53
sonne|workwhat does it say?10:53
n4nd0../shogun/lib/SGVector.h:41:28: error: invalid constructor; you probably meant 'shogun::SGVector<T> (const shogun::SGVector<T>&)'10:53
n4nd0I tried putting SGVector<T> too10:53
sonne|workn4nd0: code?10:53
n4nd0sonne|work: a second10:53
n4nd0sonne|work: basically that: SGVector(SGVector<T> orig);10:54
Yanglittleexport PATH=$PATH:/your/matlab/dir    where to add?  i'm dizzy....10:54
n4nd0sonne|work: it crashes when including the header from another file10:54
n4nd0Yanglittle: you should execute that in your command line10:55
n4nd0Yanglittle: However, it might be a good idea to do some reading first: take a look to this post to understand what you are doing
YanglittleI will  - -10:56
sonne|workn4nd0: no that should work10:56
n4nd0sonne|work: mmm ok, I may be doing an stupid error10:56
n4nd0Yanglittle: but just execute export PATH=$PATH:/your/matlab/dir from a command line10:57
n4nd0Yanglittle: and try out that it worked by executing matlab from any dir10:57
sonne|workn4nd0: show more code then :)10:58
n4nd0sonne|work: the c++ code?10:58
n4nd0so this is the header10:59
n4nd0SGVector(SGVector<T> orig);10:59
n4nd0and this is the impl10:59
n4nd0template<class T> SGVector<T>::SGVector(SGVector<T> orig) : SGReferencedData(orig)  {  copy_data(orig);  }11:00
n4nd0but the compiler is not getting there, it stays in the header11:01
n4nd0I am going to make clean and try again just in case..11:01
-!- Yanglittle [deb20afd@gateway/web/freenode/ip.] has quit [Quit: Page closed]11:03
sonne|workn4nd0: why do you do SGReferenceData(orig)? and why do you copy any data?11:05
sonne|workmaybe right though11:05
n4nd0sonne|work: because it is the constructor11:05
n4nd0I thought that was the right way :)11:06
n4nd0sonne|work: anyway, the error is still there11:06
n4nd0I believe one cannot have a SGVector(const SGVector &orig) constructor11:07
n4nd0and a SGVector(SGVector orig) one11:07
n4nd0sonne|work: aaaaah look a reasonable answer11:08
sonne|workn4nd0: hehe that makes sense11:09
sonne|workn4nd0: maybe you implement a set() method then?11:10
n4nd0sonne|work: yeah, I was thinking of that11:10
sonne|workthat could then just call the copy constructor11:10
sonne|workand well we could enhance typemaps to also deal with & stuff11:10
sonne|work(easy too)11:11
n4nd0sonne|work: the set method is easy11:11
n4nd0sonne|work: I have checked the typemaps file and the SGVector_protocols11:11
n4nd0that is a whole new world :)11:11
-!- blackburn [] has quit [Quit: Leaving.]11:14
-!- blackburn [] has joined #shogun11:16
sonne|workn4nd0: yeah set is certainly sufficient :)11:17
sonne|workfrom #c-base:11:26
sonne|work1987 - Larry Wall falls asleep and hits Larry Wall's forehead on the keyboard. Upon waking Larry Wall decides that the string of characters on Larry Wall's monitor isn't random but an example program in a programming language that God wants His prophet, Larry Wall, to design. Perl is born.11:26
-!- heiko [] has joined #shogun11:29
sonne|workheiko: I've finally understood why the streaming stuff crashes11:30
sonne|workwhich doesn't help me at all - it is non-trivial to fix :/11:30
sonne|workreason is simply that sgvectors are used in there and back in the days we didn't have refcounting but these were just ptrs11:31
sonne|workso kaboom11:31
sonne|workyour ass11:31
n4nd0Larry Wall thing haha11:31
n4nd0sonne|work: complete history :D11:33
sonne|workn4nd0: seen the mailinglist?11:33
n4nd0sonne|work: yep11:33
sonne|workwhy is this guy saying june?11:33
sonne|work July 12th!11:33
n4nd0sonne|work: idk11:33
n4nd0but let's host a second one in Stockholm :D11:33
sonne|workjust organize it :D11:34
blackburnsonne|work: welcome to russia!11:37
blackburnn4nd0: lovely isn't it?11:39
n4nd0indeed hehe11:39
blackburnhah this one is better11:40
n4nd0blackburn: who is the short one?11:41
blackburnn4nd0: old president11:41
blackburnand old new president11:41
blackburnoh you know we have only two FOR EVER11:42
n4nd0yeah I recgonized Putin11:42
n4nd0aah so they take turns??11:42
heikosonne|work: ok, well I reckoned that :?)11:43
blackburnn4nd0: yes 2 times of la poutine and one time of mjedwedjew and 2 times of la poutine11:43
blackburnand so on11:44
sonne|workheiko: so it is no wonder that this all breaks. It needs a large refactoring11:44
blackburnokay actually that was a last time and putin is forever11:44
blackburnsonne|work: one more joke on perl11:44
heikosonne|work, yes, this is also non-trivial since sometimes he combines things in a strange way11:45
heikoso one has to think when refactoring11:45
n4nd0blackburn: the wat video, that one is awesome11:45
blackburnperl is the only language which sources do not change after RSA encryption11:45
sonne|workblackburn: *lol*11:45
heikosonne|work, btw at some point I tried to do this, but gave up after 1 day and did never commit11:45
blackburnheiko: you were to commit suicide though I'd expect11:45
blackburnrefactoring that shit is tough11:45
blackburnheiko: we need a great powerful lord of refactoring11:48
blackburnthe mighty specialist11:48
heikoblackburn: yes, you exactly got it :)11:48
blackburnsonne|work: heiko: so I offer that guy to visit hell?11:49
blackburnerr russia I mean11:49
heikoblackburn: sonne|work problem is again that you dont notice when you break modules of the code, so tests would help there11:49
heikoblackburn:  haha :)11:49
heikowhy do all the Russians tell people not to go to Russia? :D11:49
heikocant be that hellishj11:49
blackburnheiko: oh well11:49
blackburnheiko: one example - spring came and roads have vanished :D11:50
heikohaha :)11:50
blackburndirt everywhere lol11:50
blackburnheiko: do other russians tell you the same?11:50
heikoblackburn: yes and polish people11:52
heikoI have to been to Warsaw recently11:52
heikoand a local guy told me11:52
heikoRussia is a wild wild country - you should not go there11:52
blackburnheiko: poland should be kinda heaven11:52
heikowe should shoot it to the moon :)11:52
blackburncomparing to russia11:53
blackburnheiko: polish people are freaked out about russia though11:53
blackburnheiko: so they are not whom to believe but believe me :D11:53
heikohehe ok :)11:53
heikoblackburn:  I would really like to visit at some point11:53
blackburnheiko: moscow should be ok11:54
heikoblackburn: yes!11:54
heikoor st. petersburg11:54
blackburnheiko: but not syzran or saransk11:54
heikoI not even ever heard those names11:55
blackburnheiko: well I have not known about heidelberg before yesterday11:55
blackburnheiko: so that's normal11:55
heikohaha :)11:55
heikowhy have you heard about it?11:55
heikothe famous neuroscientists from there fold you?11:55
blackburnheiko: heidelberg?11:56
blackburnheiko: georg zeller is located there11:56
blackburnheiko: vapnik is originally from uzbekistan11:56
blackburnheiko: that's hell11:56
blackburneven more dangerous :D11:56
heikoblackburn: well thats why he came up with large margins to keep stuff away :D11:57
blackburnheiko: hahah11:57
blackburnheiko: well when it was unite - USSR - it was ok11:57
blackburnheiko: but now they don't like russians a little and as every european guy would look russian I'd not visit it11:58
heikook :)11:58
heikoI dont really consider myself as Russian looking, but Ill stay away anyways11:58
blackburnheiko: what is russian looking? we looks like other europeans mostly11:59
heikoblackburn: I would say no, its possible to spot them11:59
heikowell, I dont know how you look like blackburn11:59
blackburnheiko: yes but not really by looking but behaviour I think11:59
blackburnheiko: see you in berlin :D12:00
heikoindeed :)12:00
blackburnheiko: sight is discriminative feature I think12:00
blackburnheiko: ah yes and russians never smile12:00
blackburnheiko: I bet you use it as a feature too12:01
blackburnheiko: hmm this guy is moody12:01
blackburnand classifier outputs 95%12:01
blackburnof being russian12:01
blackburnheiko: in moscow subway you could met some gentle and polite people that would say where should you go12:01
heikovapnik would like that12:01
blackburnthey could say who are you also12:02
blackburnactually one could learn whole russian obscene dictionary being slow in subway12:03
heikosame here blackburn12:03
blackburnheiko: oh then it is common12:05
blackburnheiko: what do you think about forcing all algebra be written in eigen3?12:09
blackburnfor new code12:09
heikoblackburn: good!12:10
heikoblackburn: in particular, not use lapacl12:10
heikothis api is a hell to debug12:10
heikoand eigen3 got everything we need12:10
blackburnheiko: lapack is better than loops12:10
heikoI also tried many of their solvers12:10
heikothat is good work12:10
blackburneigen3 is better than lapack12:10
blackburneigen4 is better than eigen312:10
blackburnbut we get no eigen4 yet12:11
heikoblackburn: how do the sparse impleentations work with shoguns sparese representation?12:11
blackburnheiko: well would require copy..12:11
heikoblackburn: I see, thats bad12:11
heikowhy is that?12:11
blackburnheiko: well the same reason we copy from sgmatrix :)12:12
blackburnheiko: if we made sgmatrix an adapter for eigen3 matrix and sgsparsematrix an adpater for eigen3 sparse matrix..12:13
blackburnheiko: but copying sparse matrix is not a big issue, is it?12:13
heikoblackburn: sparse=large so yet it is12:14
heikoblackburn: well12:14
heikoI guess not12:14
heikosince solving systems with these matrices is much more significant in terms of costs12:14
heikoas long as we dont copy in the most inner loop of something, should be fine12:14
blackburnheiko: I don't get it - could you explain?12:15
heikoblackburn: so you have a very large sparse matrix12:15
heikoyou copy it12:15
blackburnheiko: is that a feature matrix?12:15
heikotakes some time12:15
blackburnheiko: 0.00001ms may be12:16
heikocovariance matrix for example12:16
heikonow, you want to solve a system with the matrix12:16
heikotakes much longer12:16
heikoso copying is not a big thing in this case12:16
blackburnheiko: yes12:16
heiko(if the matrix is very large)12:16
blackburnheiko: it is just a memory efficiency12:16
blackburnheiko: I'd say solving is always slower than copying12:16
heikoblackburn: yes12:17
heikoblackburn: I am heavily using sparse matrices in matlab currently12:17
heikoand matlab always copies12:17
blackburnheiko: C++11 avoids some copies but in general copying happens all the time12:17
heikobut what takes time are the solving operations12:17
blackburnheiko: have you heard about rvalues btw?12:17
blackburnheiko: in C++ move semantics appeared12:18
blackburnheiko: imagine you do method(get_parameter())12:18
blackburnheiko: where get_parameter returns some class12:18
blackburnheiko: in this case copy would be required12:18
blackburnif you define method(class&&) - copying won't happen12:19
blackburnheiko: instead of const &12:21
heikowhats the difference to a reference then?12:21
blackburnheiko: temporary objects could be detected only with &&12:21
heikoI see12:22
heikonice idea12:22
heikoI  am not really following the c++ development12:22
heikotoo annoyed by the language :D12:22
blackburnheiko: that's the language you write in ;)12:22
blackburnheiko: language is awesome12:22
heikoI know, just kidding :)12:22
blackburnheiko: you just have to restict yourself12:22
blackburnheiko: that's the difference with other langs12:22
blackburnheiko: in C++ one should rather consruct some thing to work on top of12:23
heikoblackburn: I tend to appreciate higher level languages more and more since my time seems to run lower and lower :)12:23
blackburnheiko: C++ is high-level12:23
heikoblackburn:  I mean type free languages with memory managment and stuff12:23
heikolike python (which I like a lot)12:24
heikobut I also like c12:24
blackburnheiko: types are necessary to write reliable code you like :)12:24
heikoits just a difficult relationship :)12:24
heikoI know12:24
heikoI prefer C for backend stuff like shogun core12:24
heikoblackburn: I should do some work, see you later today :)12:24
blackburnheiko: I don't like writing loops so I can't say the same12:24
n4nd0sonne|work: have you seen this vid?
blackburnheiko: where do you plan to find that genius?12:28
blackburnI am still scaried with your task :D12:28
n4nd0blackburn: you said neither you nor sonne|work had all the reqs12:30
blackburnn4nd0: yeah12:35
-!- Yanglittle [deb20afd@gateway/web/freenode/ip.] has joined #shogun12:43
Yanglittleeee,another problem,12:44
n4nd0Yanglittle: hey, so did you manage to solve the previous?12:44
Yanglittlesg('add_preproc', 'PRUNEVARSUBMEAN', divide_by_std); Invalid MEX-file '/home/liyang/MatlabWork/shogun-2.1.0/shogun-2.1.0/src/interfaces/matlab_static/sg.mexa64': /usr/local/MATLAB/R2012b/bin/glnxa64/../../sys/os/glnxa64/ version `GLIBCXX_3.4.15' not found (required by /usr/local/lib/  this means to update?12:45
YanglittleI'm not familiar with ubuntu.12:46
n4nd0Yanglittle: try sudo apt-get install libstdc++612:47
n4nd0or take a look to that post better12:48
sonne|workYanglittle: go to /usr/local/MATLAB/R2012b/bin/glnxa64/../../sys/os/glnxa6413:36
sonne|workmake a backup of the libs13:36
sonne|workand put the ones from usr system (in /lib and /usr/lib) there13:37
Yanglittlei'm trying13:42
n4nd0sonne|work, blackburn: another question about directors13:42
n4nd0void set(const SGVector<T> orig);13:42
n4nd0SGVector<T> create(const SGVector<T> orig);13:42
n4nd0for the first method it is fine when I give it a numpy array13:43
n4nd0but not for the second13:43
n4nd0mmm forget about it the second method should be static.. and I have not done it13:43
blackburnn4nd0: what is the reason to pass const value here?13:44
n4nd0blackburn: why not? It is not going to be changed inside13:45
n4nd0better to const if possible I think13:45
n4nd0and it works fine from python_modular for the method set13:45
Yanglittleit seems to fail, I should install the older version of GCC.13:49
n4nd0Yanglittle: mmmm I think it would be a bit weird if you actually need to do this13:50
blackburnn4nd0: why not reference?13:52
n4nd0blackburn: then it is a problem to do it from python for target interfaces13:53
blackburnahh thought so13:53
sonne|workn4nd0: const with references make sense but this not so much14:01
n4nd0sonne|work: why not?14:01
sonne|workn4nd0: it is like writing void foo(const int a);14:02
n4nd0I think that is fine14:02
blackburnn4nd0: makes no sense :)14:02
n4nd0blackburn: oh really?14:02
blackburnn4nd0: in case of foo(const int a)14:03
blackburnwhat is the differnce if you change it or not14:03
blackburnit is stored in the stack anyway14:03
n4nd0without const then you can make inside foo, a = 10 or whatever a equal to14:03
n4nd0oh yes14:03
blackburnyeah but this change won't change anything14:03
blackburnif it was reference you could change value outside14:03
n4nd0omg I am stupid14:04
blackburnbut this way you can't change anything anyway14:04
blackburnno you are not :)14:04
n4nd0a little bit at least I am :D14:04
blackburnit is ok to write const here but that doesn't affect anything14:04
sonne|workn4nd0: ok it protects you from changing 'a' inside of 'foo'14:08
n4nd0poor defensive programming14:09
sonne|workheiko: yeah sure - question is what do we do now?14:16
sonne|workmost of the code uses streaming stuff only to read in simple ascii files from shogun14:16
sonne|workso to get to stable we should somehow inactivate / blacklist the bad examples14:17
sonne|workmaybe implementing some ascii reader would be a good first workaround14:17
Yanglittle... I can't solve the problem,aaaa.15:10
heikosonne|work: Well either we invest the time (probably it *is* solvable eventually) or not. As we are very short on this, i dont know. We already removed all the non-working examples. I think the best way would be to start from zero and try to make things work in little steps starting from the most basic functionality15:16
n4nd0Yanglittle: library not found still?15:17
sonne|workheiko: I don't talk about removing examples15:17
sonne|workjust to disable them15:17
sonne|workheiko: and the streaming stuff is abused currently - no one actually really needs it so better have some simple reader for that file now15:18
sonne|workand then fix the streaming correctly(tm)15:18
heikogood plan15:18
heikobtw for me remove=disable15:19
heikoso we do not have a simple ascii file reader?15:19
heikoI didnt even know that15:19
Yanglittleso who can save me... - -15:20
n4nd0Yanglittle: can you describe a little bit more what's going on please15:20
sonne|workheiko: we have that actually!15:21
Yanglittlesg('add_preproc', 'PRUNEVARSUBMEAN', divide_by_std); Invalid MEX-file '/home/liyang/MatlabWork/shogun-2.1.0/shogun-2.1.0/src/interfaces/matlab_static/sg.mexa64': /usr/local/MATLAB/R2012b/bin/glnxa64/../../sys/os/glnxa64/ version `GLIBCXX_3.4.15' not found (required by /usr/local/lib/   I tried the methods on google, but failed.15:22
Yanglittlei trid to make a soft link to different libstdc++, failed.15:23
sonne|workYanglittle: but you removed the libstdc++ stuff from that dir before doing so right?15:24
n4nd0Yanglittle: please do: ldd usr/local/MATLAB/R2012b/bin/glnxa64/../../sys/os/glnxa64/
sonne|workYanglittle: then please move it to some backup dir15:27
n4nd0I am off for a rest now, see you later15:27
-!- n4nd0 [] has quit [Quit: leaving]15:29
-!- Yanglittle [deb20afd@gateway/web/freenode/ip.] has quit [Quit: Page closed]15:30
-!- shogun-notifier- [] has joined #shogun15:38
shogun-notifier-shogun: iglesias :master * 13f88b1 / src/shogun/lib/SGVector.cpp,src/shogun/lib/SGVector.h:
shogun-notifier-shogun: set method in SGVector wrapper of copy constructor for SWIG interfaces15:38
shogun-notifier-shogun: Soeren Sonnenburg :master * 74a5c82 / src/shogun/lib/SGVector.cpp,src/shogun/lib/SGVector.h:
shogun-notifier-shogun: Merge pull request #947 from iglesias/structure15:38
shogun-notifier-shogun: set method in SGVector wrapper of copy constructor for SWIG interfaces15:38
-!- FSCV [~FSCV@] has joined #shogun15:44
-!- travis-ci [] has joined #shogun15:45
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:
-!- travis-ci [] has left #shogun []15:45
-!- FSCV_ [] has joined #shogun15:47
-!- FSCV [~FSCV@] has quit [Ping timeout: 264 seconds]15:49
shogun-buildbot_build #832 of deb2 - static_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: iglesias <>16:00
shogun-buildbot_build #827 of bsd1 - libshogun is complete: Success [build successful]  Build details are at
-!- blackburn [] has quit [Quit: Leaving.]16:05
shogun-buildbot_build #376 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: iglesias <>16:10
shogun-buildbot_build #959 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: iglesias <>16:28
shogun-buildbot_build #377 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>16:35
shogun-buildbot_build #833 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at
shogun-buildbot_build #960 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Soeren Sonnenburg <>16:38
-!- FSCV_ [] has quit [Ping timeout: 252 seconds]16:55
-!- FSCV_ [~FSCV@] has joined #shogun16:56
-!- FSCV_ [~FSCV@] has quit [Ping timeout: 245 seconds]17:13
-!- FSCV [~FSCV@] has joined #shogun17:19
-!- FSCV [~FSCV@] has quit [Ping timeout: 272 seconds]17:26
-!- FSCV [~FSCV@] has joined #shogun17:32
-!- zxtx [] has quit [Ping timeout: 255 seconds]17:35
-!- zxtx [] has joined #shogun18:05
-!- zxtx [] has quit [Ping timeout: 256 seconds]18:10
-!- shogun-notifier- [] has quit [Quit: transmission timeout]18:38
-!- n4nd0 [] has joined #shogun18:44
sonne|work :-)18:46
n4nd0sonne|work: we got a nice answer from Patrick, did you see?18:47
-!- blackburn [~blackburn@] has joined #shogun19:40
-!- heiko [] has quit [Quit: Leaving.]20:18
blackburnsonney2k: I'll add lcov support soon20:21
-!- FSCV [~FSCV@] has quit [Ping timeout: 256 seconds]20:36
-!- FSCV [~FSCV@] has joined #shogun20:37
@sonney2kblackburn, still awake?23:08
blackburnsonney2k: yes23:08
blackburnsonney2k: what's up?23:09
n4nd0party time! :)23:09
@sonney2kblackburn, what was your gsoc handle again23:09
blackburnsonney2k: sergey_lisitsyn23:10
@sonney2kblackburn, OK I've submitted some dummy23:12
blackburnsonney2k: is that better to early submit?23:12
blackburnatgh what's with my word order23:12
@sonney2kblackburn, the sooner the better.23:13
@sonney2kwe can incrementally improve all the time23:13
-!- hoijui [] has joined #shogun23:13
blackburnsonney2k: yes but what is the reason to keep improves on melange?23:14
@sonney2kblackburn, what is worrying me a lot is that we might not get down to 0 issues on the buildbot23:14
blackburnsonney2k: due?23:14
@sonney2kblackburn, and we should somehow mention all the other gsoc events and have all the old ideas / stuff on the website23:14
blackburnthe wave?23:14
@sonney2kdue before the wave in 2 weeks if we get accepted yes23:15
blackburnI see, lets try to get rid of these issues next weeks then23:15
@sonney2kbut recall that students started to look already now (the years before and this year too)23:15
@sonney2kman I wish we had inf time23:17
@sonney2kthen we could do some kaggle stuff and grow shogun with kaggle :)23:18
blackburnsonney2k: yeah that's what you should do instead of your job :)23:19
blackburnbut we didn't yet discovered how to get money automagically23:19
@sonney2kblackburn, yeah sb has to beat the no free lunch theorem23:21
-!- hoijui [] has quit [Ping timeout: 245 seconds]23:30
n4nd0people, can you tell me if you can see the document without problem?23:33
n4nd0blackburn: ^23:33
@sonney2kn4nd0, could you please have a look at the website?23:35
n4nd0sonney2k: sure, what do you want me to look at?23:36
@sonney2kn4nd0, ahh my bad did sth wrong23:37
n4nd0sonney2k: aham I see23:37
n4nd0we want a GSoC level23:37
@sonney2kblackburn, I've added and
n4nd0and GSoC 201x below23:37
@sonney2kn4nd0, but then we cannot have multiple pages for a gsoc23:38
@sonney2kI think we should have a)23:38
@sonney2kideas list23:38
@sonney2kb) announcement that we particiapte23:38
n4nd0yeah then we wouldn't be able23:38
@sonney2kc) summary23:38
n4nd0I understand23:38
n4nd0what about an old level23:38
@sonney2kI need to copy that stuff from the old website23:38
n4nd0and we put GSoC 2012 and GSoC 2011 inside23:38
n4nd0or would you like to have inner levels in those GSoCs too?23:39
@sonney2kn4nd0, is it so bad to have some content on the website?23:39
n4nd0then it is fine as it is23:39
@sonney2kI want to copy this stuff from my blog23:39
@sonney2kblackburn, we can also use that style23:39
@sonney2kI need to copy a bit of .css for that to look nice though23:40
@sonney2kn4nd0, I would also copy the europython2010 / whatever conference we talked about shogun stuff there23:41
n4nd0it makes sense yeah23:42
@sonney2kn4nd0, yeah we still haven't moved all content from the old website to the new layout23:47
@sonney2kit is about time to do that23:47
@sonney2kn4nd0, do you recall this nice 'feature matrix' we had23:48
@sonney2kwhere we compared lots of ml toolboxes?23:48
n4nd0sonney2k: yes23:48
@sonney2kthat should come back...23:48
n4nd0I guess it is easy to draw a table in django23:49
@sonney2kI had this python script that generates it from the google doc23:49
@sonney2kn4nd0, could you please move this gsoc information over to gsoc201223:54
-!- zxtx [] has joined #shogun23:54
@sonney2kboth the 2013 and the 2012 should be moved23:54
n4nd0sonney2k: sure23:54
n4nd0sonney2k: I will edit it now23:55
n4nd0we have to improve the style of some of the articles... they look a bit ugly :S23:55
@sonney2kyes do what you can :)23:57
--- Log closed Fri Mar 22 00:00:15 2013