Open in new window / Try shogun cloud
--- Log opened Wed Apr 11 00:00:00 2012
--- Day changed Wed Apr 11 2012
n4nd0so tomorrow I will make sure that this cover tree is actually giving the right answers00:00
n4nd0and if it does, I will start to port it :)00:00
blackburnn4nd0: okay00:00
n4nd0good night :)00:00
blackburngood night00:00
-!- n4nd0 [] has quit [Quit: leaving]00:01
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking]00:03
-!- wiking [] has joined #shogun00:08
-!- wiking [] has quit [Changing host]00:08
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun00:08
harshit_good night everyone :)00:58
-!- harshit_ [~harshit@] has quit [Quit: Leaving]00:59
PhilTilletblackburn, not sleeping?01:07
-!- av3ngr [av3ngr@nat/redhat/x-fslqwfzaqdfhtuml] has joined #shogun01:12
-!- PhilTillet [] has quit [Remote host closed the connection]01:15
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has quit [Quit: Page closed]01:23
@sonney2kwiking, btw the qnorm 'bug' should not have been critical because for mkl we only considered positive kernel weights02:30
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking]04:16
-!- pluskid [~chatzilla@] has joined #shogun04:55
-!- pluskid [~chatzilla@] has quit [Client Quit]04:56
-!- pluskid [~chatzilla@] has joined #shogun05:41
-!- gsomix [~gsomix@] has quit [Ping timeout: 245 seconds]05:51
-!- pluskid [~chatzilla@] has quit [Quit: ChatZilla [Firefox 11.0/20120314124128]]06:31
-!- n4nd0 [] has joined #shogun07:29
-!- wiking [] has joined #shogun08:33
-!- wiking [] has quit [Changing host]08:33
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun08:33
n4nd0sonne|work: I finished cheking all the results, this cover tree gives the right output :)08:33
n4nd0sonne|work: at least the same as our current KNN with qsort08:34
n4nd0and the interesting part, the query time is quite shorter!08:35
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection]08:35
-!- wiking [] has joined #shogun08:36
-!- wiking [] has quit [Changing host]08:36
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun08:36
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun08:51
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking]08:59
n4nd0blackburn: hey, good morning!09:01
n4nd0blackburn: I just sent a small pull request for KNN09:02
blackburnn4nd0: morgen :)09:02
blackburndamn I thought I pushed my changes09:02
n4nd0I didn't see them in the repo09:03
n4nd0no problem :)09:03
n4nd0blackburn: a detail for the cover tree09:03
n4nd0blackburn: should I substitute directly the current one we have or use another name (CoverTree2 for example)09:04
@sonney2kn4nd0, so JL's covertree works right?09:04
@sonney2kmorning blackburn09:04
n4nd0sonney2k: it is not JL's implementation09:04
n4nd0sonney2k: it is this one
@sonney2khmmhh - admittedly I am a big fan of JL's stuff.09:05
@sonney2kwhat he does is reall fast(tm)09:05
blackburnsonney2k: morgen. today is the day? ;)09:05
n4nd0sonney2k: yeah but I don't really understand that implementation09:05
@sonney2kthat is the downside of his code - but just ask him then09:06
@sonney2kblackburn, not my day09:06
n4nd0sonney2k: mailing list?09:06
@sonney2ktomorrow I would say :D09:06
@sonney2kn4nd0, personal email09:06
n4nd0sonney2k: ok, in any case, do you think it pays off to start porting this implementation?09:07
@sonney2kI would rather try to get JL's implementation to work09:07
n4nd0ok ...09:07
n4nd0this one presents a good improvement in time though09:07
n4nd0Shogun's time: 00:00:00.32880409:07
n4nd0CoverTree time 00:00:00.05365009:07
blackburnsonney2k: btw I managed to beat all my competitors with 'brand new libedrt' :D09:08
-!- uricamic [9320543b@gateway/web/freenode/ip.] has joined #shogun09:08
n4nd0that is just 100 queries using 10000 training vectors09:08
n4nd0sonney2k: another thing I was thinking about our KNN09:08
n4nd0sonney2k: right now we have N?log(N) for every query, because we sort the distances from the query vector to all the training vectors09:09
@sonney2kn4nd0, well then benchmark JLs impl against this one09:09
@sonney2kuricamic, hi...09:09
n4nd0sonney2k: yeah, I can do that ... but we won't know if the results are correct09:10
@sonney2kare you back from vacations? did you think of some small patch?09:10
n4nd0sonney2k: so it won't fell safe in any case09:10
@sonney2kn4nd0, but you can compare the output to the qsort/other covertree09:10
n4nd0sonney2k: yes, to the cover tree I am using now, the comparison is done09:10
n4nd0sonney2k: I am quite certain it is correct09:11
uricamicsonney2k: yes, I plan to work on it this week, and I have one question09:11
@sonney2kn4nd0, yeah - I meant you can do the same with JL09:11
n4nd0sonney2k: I have tried a very long test (10000 training vector, 10000 queries with 5 neighbors) and it works09:11
n4nd0sonney2k: no, I cannot do that with JL09:11
@sonney2kand the winner is the one we integrate09:11
uricamicsonney2k: is it possible to write very simple interface for the libbmrm as a patch, or hsould I choose something different?09:11
@sonney2kn4nd0, why not?09:12
@sonney2kuricamic, I would prefer something completely different09:12
@sonney2kblackburn, do you have some *small* idea for uricamic?09:12
uricamicsonney2k: ok09:12
n4nd0sonney2k: when I query 10 points with k = 10 with JL, I get 10 neighbors for the first query09:12
n4nd0sonney2k: 1 for each of the others ...09:12
blackburnsonney2k: I proposed parzen stuff before09:12
@sonney2kuricamic, ^09:12
@sonney2kso feel free09:13
blackburnsonney2k: do you think it worths?09:13
@sonney2kuricamic, other relatively simple algos are also ok09:13
uricamicok, I will start with the Parzen windows09:13
blackburnah or some nadaraya-watson or etc09:13
n4nd0sonney2k: but I can try to check again if I am doing correctly this test, but I think so09:13
@sonney2kfor example some iterative method for ridge regression (not computing pinv)09:13
@sonney2kn4nd0, well ask JL directly in CC me in the email why this is the case and give him your code snippet09:14
n4nd0sonney2k: good idea :) thank you09:14
@sonney2kuricamic, that was for you: you could for example also do some iterative method for ridge regression (not computing pinv)09:15
@sonney2kif that is more in your field of expertise09:15
@sonney2kas optimization guru :)09:15
uricamicsonney2k: ok, I will look at it and then decide09:16
@sonney2kuricamic, just look in the shogun/regression/ directory... these all are pinv based things09:16
uricamicsonney2k: ok, thanks09:17
n4nd0sonney2k: about the other strategy for KNN, the qsort one, right now we are in N?log(N) but that can be done in N easy09:18
n4nd0sonney2k: instead of sorting all the distances and taking the K shortest09:18
n4nd0sonney2k: loop through all the distances and keep a record of the K shortest one found09:18
blackburnn4nd0: really, how? ;)09:20
@sonney2kn4nd0, yes that would be faster for small k09:20
blackburnit won't be N still I mean09:21
@sonney2kn4nd0, I guess no one is using k>1109:21
@sonney2kblackburn, it is N*k09:21
n4nd0sonney2k: I think it is something like N?k?log(k)09:21
blackburndepends on structure you will use to store neighbors09:21
blackburnbut with lame way it is even Nk^209:22
n4nd0using just a vector of k elements09:22
n4nd0that you order in every iteration09:22
n4nd0is just N?k?log(k)09:22
n4nd0and I know nothing lamer than that ...09:22
blackburnbut you would need to compare with elements09:22
blackburnin worst case with each element09:22
@sonney2kn4nd0, I thought you meant sth like going through the N elements and keeping a history of the k neighreast09:22
n4nd0sonney2k: yeah, that is the idea09:23
blackburnn4nd0: I used fibonacci heap for that before :D09:23
n4nd0blackburn: what complexity gives with fibonacci heap?09:23
n4nd0I normally solve this problem using std::set09:23
blackburnminimum is log N amortized09:24
n4nd0you insert in log(k)09:24
blackburnso k*log N09:24
blackburninsert is O(1) in the fib heap09:24
n4nd0sonney2k: bye09:24
n4nd0blackburn: so you insert in O(1) and later you sort it?09:24
blackburnyes something like that09:25
blackburnactually you could measure it too09:25
n4nd0but when you said09:25
n4nd009:22 < blackburn> but you would need to compare with elements09:25
n4nd009:22 < blackburn> in worst case with each element09:25
n4nd0you are playing there with worst-time complexity of qsort??09:25
n4nd0if not, I don't see what you mean09:25
blackburnokay let us have x1,x2,x3,x4,...09:25
blackburnwhen you collect neighbors09:26
blackburnlet we have some collection already09:26
blackburnand if we need to know if new vector x is a neighbor09:26
blackburnwe would need to compare all these distances right?09:27
n4nd0you just compare with the one that gives the largest distance09:27
n4nd0among these neighbors09:27
n4nd0that's why you need to have it sorted09:27
n4nd0so you have this collection of k tentative neighbors09:27
blackburnokay and if you have distance 1.509:28
blackburnwhile largest is 509:28
blackburnand smallest is 1.2509:28
n4nd0then 5 pops out09:28
blackburnyou would need to compare to place it?09:28
n4nd01.5 enters09:28
n4nd0the largest goes out09:28
n4nd0the new one goes in, and you sort again09:28
blackburnokay got it09:29
n4nd0N?k?log(k) :D09:29
n4nd0"assuming" that qsort is n?log(n) for n elements09:29
n4nd0so we will normally have complexity N because N >> k09:30
n4nd0and that would make our current KNN to work in O(N) instead of O(N?log(N))09:30
n4nd0blackburn: did I convince you?09:30
blackburnn4nd0: just measure :)09:30
n4nd0blackburn: I think that is my homework for today09:31
n4nd0I am going to change it, measure and mail JL09:31
blackburnyeah a lot of things to do09:31
-!- nickon [] has joined #shogun09:31
n4nd0blackburn: I was thinking of that because I remember you said that scikits people wrote that our KNN is slow09:34
n4nd0blackburn: and AFAIK they don't have cover tree either09:34
blackburnn4nd0: kdtree09:37
blackburnor balltree09:37
n4nd0I think kd-trees work fine in low dimensionality problems09:38
n4nd0blackburn: I have just realized that KNN comes with SG_PROGRESS, how is that used?09:43
blackburnI would avoid kdtree09:43
n4nd0blackburn: ok, thank you09:44
n4nd0blackburn: why would you avoid kd-trees?09:44
blackburnn4nd0: that thing should be everywhere actually09:44
blackburnn4nd0: because it would work only for simplefeatures09:44
sonne|workn4nd0: mail JL first then start the play stuff :)09:46
sonne|workn4nd0: btw JL is located in NY so you cannot quickly get an answer09:46
n4nd0sonne|work: :( haha ok, I was starting to play first, such a kid :P09:46
n4nd0sonne|work: the sooner I mail him, the better then09:47
blackburnsonne|work: is there anything except covertree and vw he has we can integrate?09:47
sonne|workblackburn: his vw stuff changed completely09:47
sonne|workthey have some massiviely parallel thing in there now09:48
blackburnsonne|work: I have seen that - bfgs-l is here now09:48
n4nd0particle filters! :P09:48
-!- av3ngr [av3ngr@nat/redhat/x-fslqwfzaqdfhtuml] has quit [Ping timeout: 246 seconds]10:06
n4nd0sonne|work: there is sth weird, JL's non-templated implementation gives the correct number of neighbors while the templated (the one I was using for my test) not10:07
sonne|workthen sb introduced a bug :)10:07
n4nd0sonne|work: I am going to try to find out why is this, the mail may wait10:08
n4nd0yeah, either I did it in the test or it was in this templated version10:08
sonne|workn4nd0: maybe you can then even compare the 2 cover trees first10:08
n4nd0I think the first alternative is more likely10:08
n4nd0sonne|work: which two ones?10:08
sonne|workthe one you dig up and JLs10:09
blackburnn4nd0: could you please remind me how does non-templated stuff works?10:15
n4nd0blackburn: the point is fixed to: typedef float* point10:16
blackburnn4nd0: how distances are computed then?10:17
sonne|workfloats are really crazy ...
n4nd0blackburn: normal Euclidean distance10:18
n4nd0blackburn: I think I didn't get the point of the question sorry10:18
blackburnn4nd0: hardcoded? can you change it?10:18
n4nd0blackburn: you mean like how many elements do they use?10:18
blackburnn4nd0: no how hard it could be to adapt non-templated?10:18
blackburnor in other words is it any possible ?10:19
n4nd0blackburn: so take a look to the distance function for example10:19
-!- wiking [] has joined #shogun10:19
-!- wiking [] has quit [Changing host]10:19
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun10:19
blackburnsorry overwhelmed with stuff at job but was curious :)10:19
blackburnn4nd0: I mean can you provide you own distance there?10:19
wikinganybody knows how can i force an ext3 fs to be clean?10:19
blackburnno idea10:20
n4nd0blackburn: I think that by changing that function that can be done10:20
n4nd0wiking: no idea either10:20
wikinghehehe trying to hack with tune2fs and dumpfs :>10:21
sonne|workwiking: hi ... what errors where you getting with clang? because here it compiles fine w/ clang...10:21
n4nd0blackburn: there are things I don't get anyway10:21
wikingsonne|work: will give you the list10:21
n4nd0blackburn: if I set k = 9, I get 10 neighbors10:21
wikingsonne|work: just now i'm trying to save my work while the FS is collapsing :)10:21
n4nd0blackburn: the k used is the k given +110:22
blackburnn4nd0: Iguess you get object itself?10:22
n4nd0blackburn: aham, let me check10:23
sonne|workwiking: why do you want the fs to become clean10:23
sonne|workwiking: just mount it10:23
wikingsonne|work: i cannot boot10:23
wikingjust in single user mode10:23
n4nd0blackburn: I don't really know if it is because of that10:24
sonne|workwiking: then boot in single user mode copy stuff over and run fsck10:24
wikinghehehe i'm actually on it10:24
wikingthe problem here is that it's an nfs10:24
wikingand i don't have space to save that much data only when i have network10:25
wikingbut i think10:25
wikingiiii think i've made it now :)10:25
wikingdamn raid controllers...10:39
-!- sonne|work [~sonnenbu@] has left #shogun []10:46
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]10:50
-!- sonne|work [~sonnenbu@] has joined #shogun11:10
CIA-64shogun: puffin444 master * r6f128a5 / (4 files in 3 dirs):11:16
CIA-64shogun: Mean predictions in GaussianProcessRegression now solved using Cholesky11:16
CIA-64shogun: decomposition. Predicted Covariance Matrix is also accessible. -
CIA-64shogun: Soeren Sonnenburg master * ra27e1ae / (8 files in 5 dirs):11:16
CIA-64shogun: Merge pull request #406 from puffin444/master11:16
CIA-64shogun: Gaussian Process Regression -
CIA-64shogun: Soeren Sonnenburg master * r2d37d40 / src/shogun/classifier/KNN.cpp :11:17
CIA-64shogun: Merge pull request #440 from iglesias/knn-covertree11:17
CIA-64shogun: Minor bugfixes in KNN -
n4nd0it was my bad, there were some bugs in my implementation of the point I was using for JL11:22
n4nd0I am sorry for the confusion11:22
sonne|workn4nd0: so now benchmark please :D11:23
sonne|workthen we can decide11:24
n4nd0sonne|work: there is one issue though11:24
n4nd0I believe that JL doesn't give the points in order (closest to farthe)11:25
n4nd0the neighbors I mean11:25
n4nd0so a test to ensure that the results are correct would take me a bit more time11:26
n4nd0we can benchmark it though and assume that the results are correct (they must be)11:26
sonne|workwell try if they are exactly the same...11:30
sonne|workthen you know for sure that JL's covertree gives (un)ordered results11:31
sonne|workhowever, it is very cheap if you would just compute the distance to these k points and sort accordingly (compared to all the n log(n) stuff before)11:32
shogun-buildbotbuild #662 of cmdline_static is complete: Failure [failed test_1]  Build details are at  blamelist: root@localhost.(none),, jacobw@localhost.(none)11:50
shogun-buildbotbuild #640 of r_static is complete: Failure [failed test_1]  Build details are at  blamelist: root@localhost.(none),, jacobw@localhost.(none)11:54
shogun-buildbotbuild #641 of octave_static is complete: Failure [failed test_1]  Build details are at  blamelist: root@localhost.(none),, jacobw@localhost.(none)11:58
-!- gsomix [~gsomix@] has joined #shogun12:03
gsomixhello all12:04
shogun-buildbotbuild #628 of python_static is complete: Failure [failed test_1]  Build details are at  blamelist: root@localhost.(none),, jacobw@localhost.(none)12:10
-!- Johannes__ [55e5c1c3@gateway/web/freenode/ip.] has joined #shogun12:12
Johannes__Hello, where can i find the shogun-java-modular package the installation guide talks about on ubuntu?12:13
Johannes__I can find some of the others through the usual channels but not that one.12:14
n4nd0Johannes__: hey! it doesn't appear directly in my apt-get either, is it said around there to include any special repository for apt-get??12:16
n4nd0sonne|work: I got results of the benchmark :)12:16
Johannes__n4nd0: Doesen't mention any special repository to include as far as I can see.12:35
n4nd0Johannes__: aham, then I don't really know12:35
n4nd0Johannes__: have you considered installing from source?12:36
n4nd0sonne|work: JL's is considerably faster12:36
n4nd0sonne|work: it is reasonable to some extent since in his version the algorihtms for creation of the tree and query are batch algorihtms12:37
Johannes__Considered, yes but will probably look around for alternatives first, tried the macport install on my laptop without success aswell ;p12:37
n4nd0Johannes__: but you are on ubuntu or in mac?12:37
Johannes__both, the shogun-java-modular question was in regards to ubuntu12:38
n4nd0Johannes__: ahm12:39
n4nd0Johannes__: anyway, it should not be hard to install in ubuntu from source, just say it if you want some help with that12:40
Johannes__Yea perhaps that's the way to go, we'll see. Thanks for the offer and answers regardless :)12:42
n4nd0no problem :)12:44
sonne|workhi gsomix13:08
sonne|workn4nd0: with batch you mean - one has to specify *all* train data at once and *all* test data at once?13:08
n4nd0sonne|work: yes, both for the training data and the test data13:22
n4nd0sonne|work: actually the way it works is doing a cover tree for the test data too13:22
sonne|workso you end up with k*T indices?13:23
sonne|workuhh and what are the memory requirements for that?13:23
n4nd0I have not measured it myself13:23
n4nd0but in the paper they say that a cover tree needs O(n) in memory13:24
sonne|worksounds great...13:25
n4nd0the build time of the other's cover tree really sucks13:25
sonne|workbuild time?13:26
sonne|workwhy? what?13:26
sonne|workyou mean compile time13:26
sonne|workor construction of the tree?13:26
n4nd0to create the cover tree13:26
sonne|workwell I think it is pretty clear that we want JL's tree :)13:27
n4nd0yeah ... bad new for my health status :P13:27
n4nd0sonne|work: so how would you like that we use it? use the same code and provide wrappers?13:28
n4nd0sonne|work: is it better to re-write?13:28
n4nd0btw, I didn't check the correctnes of the results in JL yet!13:28
n4nd0I just assumed that they are ok13:28
sonne|workn4nd0: will I said so upfront :D13:36
sonne|workn4nd0: it depends - a shallow wrapper is easier to maintain in case there are fixes to JL's covertree13:36
-!- nickon [] has quit [Quit: ( :: NoNameScript 4.22 :: )]13:45
-!- n4nd0 [] has quit [Quit: leaving]14:08
-!- Johannes__ [55e5c1c3@gateway/web/freenode/ip.] has quit [Quit: Page closed]14:15
-!- PhilTillet_ [~Philippe@] has joined #shogun14:20
-!- PhilTillet_ [~Philippe@] has quit [Ping timeout: 240 seconds]14:33
gsomixsonne|work, moin. I'm working on SGVector, but slow. I think I can turn on "berserk mode" in Friday or Saturday.15:10
sonne|workgsomix: before doing that - did you fix the  SG_PROGRESS stuff?15:11
sonne|workyes ok15:11
sonne|workno problem15:11
-!- pluskid [~chatzilla@] has joined #shogun15:46
sonne|workbuildbot failing16:02
gsomixsonne|work, about SG_PROGRESS. should I do overall progress on all four cycle?16:03
sonne|workone overall progress (approximate is sufficient)16:04
gsomixsonne|work, ok.16:04
gsomixsonne|work, and about references. I figured out, that references is true way.16:09
sonne|workgsomix: ahh how come?16:19
sonne|workgsomix: what changed your mind16:19
sonne|workpluskid: could you please set a proper username?16:20
sonne|workpluskid, uricamic these are the instructions16:20
pluskidsonne|work: you mean set username to real name?16:21
gsomixsonney2k, wiking wonderful analyzed this situation, hehe. :)16:21
gsomix* sonne|work,16:22
-!- harshit_ [~harshit@] has joined #shogun16:22
sonne|workpluskid: yes16:22
pluskidok, changed16:22
sonne|workit helps you to get credit!16:22
pluskidbut we cannot update already committed code maybe?16:23
sonne|workgsomix: I see :)(16:23
uricamicsonne|work: I have already set it, sorry that I haven't done it for the first commit16:23
sonne|workpluskid: no that is not possible16:23
sonne|workbut for new pull requests at least16:23
pluskidok, see it16:23
sonne|workpluskid: we have a couple of weird contributors in there16:25
sonne|workinvalid-email-address (You didn't set your git config correctly)16:25
sonne|workis the best one (blackburns favourite :)16:25
pluskidit consists of a lot of people :D16:26
sonne|workuricamic: no problem16:29
sonne|workwe all are not yet git experts16:29
-!- vikram360 [~vikram360@] has joined #shogun17:02
pluskid /quit17:11
pluskidfailed to quit...17:12
-!- pluskid [~chatzilla@] has quit [Quit: ChatZilla [Firefox 11.0/20120314124128]]17:12
-!- pythonroar [d2198538@gateway/web/freenode/ip.] has joined #shogun17:16
-!- pythonroar [d2198538@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:26
-!- vikram360 [~vikram360@] has quit [Ping timeout: 265 seconds]17:31
-!- vikram360 [~vikram360@] has joined #shogun17:32
gsomixI'm going to a dance classes.17:39
-!- uricamic [9320543b@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:40
-!- puffin444 [230bf329@gateway/web/freenode/ip.] has joined #shogun17:47
-!- harshit_ [~harshit@] has quit [Ping timeout: 244 seconds]18:06
-!- harshit_ [~harshit@] has joined #shogun18:24
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has joined #shogun18:32
-!- PhilTillet [] has joined #shogun18:33
puffin444Do you know when/if Sergey or Soeren will be on IRC?18:36
PhilTilletI have no idea :p18:36
PhilTilletI think they are eagerly waiting for the number of slots to be announced =D18:36
blackburnpuffin444: I'm partially available18:37
blackburnI am first one ;)18:37
puffin444So blackburn it looks like I broke the build :(18:38
blackburnI've seen the mail18:38
blackburnfirst of S heh18:38
blackburnlet me check logs18:39
blackburnpuffin444: does it work on your machine?18:39
puffin444That's what I don't understand18:40
blackburnany errors detected by valgrind?18:42
-!- PhilTillet [] has quit [Ping timeout: 245 seconds]18:45
puffin444There are no memory leaks. However, it does complain about freeing memory. This happens to all the examples though.18:46
puffin444In Valgrind18:46
blackburnpuffin444: do you copy the covariance matrix when you return it?18:47
puffin444Hmm... maybe not.18:48
blackburnI mean is it a member?18:48
blackburnah no18:48
puffin444But the problem is that I can't reproduce this error on my system, so I don't really know how to fix it.18:49
blackburnlet me try to compile and test that18:49
blackburnpuffin444: actually the problem is some matrix is free'd two times18:51
puffin444Aah I see18:51
puffin444blackburn, I am extremely sorry to do this, but I need to leave the IRC immediately18:51
blackburnpuffin444: no problem18:51
puffin444I'll check and find this double-freed matrix.18:51
-!- puffin444 [230bf329@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:51
blackburnyeah and I'll try to reproduce18:52
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]18:56
-!- flxb [] has joined #shogun18:57
flxbare kernels in shogun centered?18:58
-!- PhilTillet [~Philippe@] has joined #shogun19:00
-!- n4nd0 [] has joined #shogun19:04
flxbor is it normalized in any way if I use custom kernels?19:07
-!- harshit_ [~harshit@] has quit [Ping timeout: 260 seconds]19:11
-!- harshit_ [~harshit@] has joined #shogun19:17
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has joined #shogun19:26
blackburnflxb: we have kernel normalizers for that19:27
-!- flxb [] has quit [Quit: flxb]19:28
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:40
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has joined #shogun19:51
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun19:52
-!- harshit_ [~harshit@] has quit [Ping timeout: 245 seconds]19:54
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds]19:58
-!- PhilTillet [~Philippe@] has quit [Ping timeout: 245 seconds]20:00
-!- harshit_ [~harshit@] has joined #shogun20:08
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun20:11
wikingblackburn: here?20:15
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:18
-!- harshit_ [~harshit@] has quit [Ping timeout: 252 seconds]20:40
-!- harshit_ [~harshit@] has joined #shogun20:49
-!- PhilTillet [~Philippe@] has joined #shogun21:09
@sonney2khey therre21:42
@sonney2kohh the man with the fast covertree :)21:43
n4nd0haha not yet :)21:43
n4nd0I won't say until we can use it within shogun21:43
@sonney2khmmh does anyone know when the business day of pacific time is over21:45
@sonney2klooks like in 5-6 hours but I guess at google they work a bit longer so lets say 8 hrs ;-)21:45
n4nd0I am around the channel in case I get something, I have seen you there too21:47
@sonney2kno I will go to bed need to recover a bit21:47
n4nd0we'll know tomorrow then21:49
CIA-64shogun: Soeren Sonnenburg master * rc80508e / src/shogun/mathematics/Math.cpp : Fix compile error in twonorm() when lapack is not available -
* gsomix have returned from dance floor.21:51
PhilTillethey sonney2k21:51
wikingoops sonney2k sorry about that bug :(21:52
@sonney2kyeah luckily we have that buildbot (last year we didn't and that was really causing us quite some headache21:55
wikingjavac] symbol  : constructor RealFeatures(org.jblas.DoubleMatrix)21:58
wiking    [javac] location: class org.shogun.RealFeatures21:58
wikinghow come?21:58
@sonney2kgot to leave train21:59
wikingwas working till now21:59
wikingdoh these days...21:59
gsomixgood night22:14
@sonney2knite gsomix22:16
@sonney2kwiking, what is the exact error?22:16
@sonney2kseems a little unspecific...22:16
n4nd0good night gsomix22:16
wikingsonney2k: i think i had some trouble with detecting jblas22:16
wikingnow i'm recompiling and i guess it'll work now22:17
@sonney2knot in CLASSPATH ?22:17
wikingit was actually22:17
@sonney2kwiking, are you using java interface of shogun?22:17
wikingi mean the problem was with22:17
wikingthat javac was defined to be gcj22:18
wikingand that had some troubles detecting the jblas (it's at default location)22:18
@sonney2khmmhh maybe22:18
wikingbut now trying with sun's javac22:18
wikingok now it's fixed22:18
wikingonly thing now is22:18
wikingno that is my error22:19
wikingok now compiled fine22:19
wikingyeah i'm using sometimes java interface22:19
@sonney2kwiking, so are you using the java interace?22:19
@sonney2kahh ok22:19
wikingyep yep22:19
wikingneed to22:19
wikingi have some features coming from hadoop22:19
@sonney2kI haven't used it except for crafting some examples22:20
@sonney2kso all feedback welcome22:20
wikingand i don't want to do constantly a transforming the features into some other formats22:20
@sonney2kwiking, btw you should be even able to serialize a shogun object from java and load it later from python :)22:20
wikingso in order to be able to simply use hadoop created sequence file i just use java interface22:20
wikinglet's see multi class linear machine22:21
wikingwith homo kernel mapping :))22:21
wikingmm i wonder what's happening22:26
wikingstill calculating :)))22:26
wikingi thought it's a fast linear machine22:27
wikingcan it be that this linear machine is slower than a kernel machine with linear kernel?22:31
@sonney2kwiking, liblinear?22:36
@sonney2kno should be very fast22:37
-!- PhilTillet [~Philippe@] has quit [Ping timeout: 252 seconds]22:37
wikingException in thread "main" java.lang.NullPointerException22:43
wikingat bme.iclef.shogun.Classify.classify(
wikingException in thread "main" java.lang.NullPointerException22:43
wikingat bme.iclef.shogun.Classify.classify(
wikingmmm not htis22:43
wikingok again my bug22:44
wikingok i'm gonna time this one now :)22:50
n4nd0time to sleep today, good night guys!22:53
-!- n4nd0 [] has quit [Quit: leaving]22:53
-!- puffin444 [230bf329@gateway/web/freenode/ip.] has joined #shogun23:36
-!- puffin444 [230bf329@gateway/web/freenode/ip.] has quit [Client Quit]23:36
-!- muddo [~muddo@gateway/tor-sasl/muddo] has left #shogun ["Leaving"]23:51
-!- blackburn [~qdrgsm@] has joined #shogun23:58
--- Log closed Thu Apr 12 00:00:19 2012