Open in new window / Try shogun cloud
--- Log opened Sun Sep 02 00:00:17 2012
-!- emrecelikten [] has quit [Ping timeout: 272 seconds]01:44
-!- romi_ [~mizobe@] has joined #shogun06:32
-!- gsomix_ [~gsomix@] has joined #shogun06:47
-!- gsomix_ [~gsomix@] has quit [Ping timeout: 264 seconds]06:55
-!- gsomix_ [~gsomix@] has joined #shogun07:01
-!- gsomix_ [~gsomix@] has quit [Read error: Connection reset by peer]07:09
-!- naywhayare [] has quit [Ping timeout: 272 seconds]08:54
-!- naywhayare [] has joined #shogun08:54
-!- blackburn [~blackburn@] has joined #shogun09:01
CIA-27shogun: Sergey Lisitsyn master * r6579577 / (10 files in 2 dirs): Merge pull request #766 from puffin444/master (+8 more commits...) -
shogun-buildbotbuild #491 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: puffin444 <>09:16
shogun-buildbotbuild #492 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>, puffin444 <>09:16
CIA-27shogun: Sergey Lisitsyn master * r47ed83f / src/shogun/regression/gp/StudentsTLikelihood.cpp : Fixed position of eigen3 guard in student t likelihood class -
shogun-buildbotbuild #493 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>09:20
CIA-27shogun: Sergey Lisitsyn master * rf89d7b8 / (2 files): Fixed more guards in GPs -
shogun-buildbotbuild #494 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>09:25
CIA-27shogun: Sergey Lisitsyn master * ra97a0ed / src/shogun/regression/gp/LaplacianInferenceMethod.h : More guard fixes for #766 -
shogun-buildbotbuild #495 of deb1 - libshogun is complete: Success [build successful]  Build details are at
CIA-27shogun: Sergey Lisitsyn master * r442ceac / (2 files in 2 dirs): Fixed sparse copy subset method and its example -
blackburnsonney2k: okay some leaks left and we would be ready09:38
-!- blackburn1 [~blackburn@] has joined #shogun09:58
-!- blackburn [~blackburn@] has quit [Ping timeout: 272 seconds]10:00
-!- n4nd0 [] has joined #shogun10:00
CIA-27shogun: Sergey Lisitsyn master * rad4ac3c / src/shogun/lib/Map.h : Fixed CMap destroy method to avoid memory leaks -
blackburn1n4nd0: 6 leaky examples left!10:56
n4nd0blackburn1: good job!10:56
n4nd0which ones?10:57
blackburn1n4nd0: base_*, two from GPs and so_multiclass_BMRM10:57
n4nd0I thought that the commit sonney2k did some days ago solved so_multiclass_BMRM10:58
blackburn1n4nd0: btw so_multiclass_BMRM now produces ~20% accuracy10:58
blackburn1am I wrong it was 100%?10:58
n4nd0I think it was yes10:58
n4nd0I think that solved so_multiclass_BMRM10:58
n4nd0but I didn't check it10:59
blackburn1it leaks on apply10:59
blackburn1and ~20%11:00
blackburn111 sometimes11:00
n4nd0mmm that's weird11:01
n4nd0we should mail michal and ask him11:01
n4nd0I would say it was more than 20%, 100% actually11:02
n4nd0I am not sure though11:02
blackburn1n4nd0: I think it is related to features loading11:02
n4nd0aham, int read_data?11:03
n4nd0has something been changed there?11:04
blackburn1yeah in commit you pointed to11:04
blackburn1it appears to be rando11:05
blackburn1n4nd0: no it is all ok with features11:12
blackburn1I've inserted liblinear there11:12
blackburn1n4nd0: do you have time to track what caused the problem?11:13
n4nd0in so_multiclass_BMRM?11:13
n4nd0blackburn1: ^11:13
n4nd0sure, I can do that later today11:14
blackburn1okay thanks11:14
n4nd0I am with another thing right now, but will look into it later11:14
blackburn1I'm now going to get some food in :D and then will fix regression examples11:14
@sonney2kn4nd0, any chance bern4rd has time this evening?11:30
n4nd0no idea...11:30
n4nd0I can give you his mail if you want so you can tell him11:31
@sonney2kn4nd0, kthx11:33
n4nd0blackburn1: around?11:59
blackburn1n4nd0: yes12:02
n4nd0there is something weird with my data repo I don't get12:02
n4nd0if I am in shogun/data12:02
n4nd0git st12:02
n4nd0# On branch master12:02
n4nd0nothing to commit (working directory clean)12:02
n4nd0but when I am in shogun/12:02
n4nd0git st12:02
n4nd0# On branch master12:02
n4nd0# Changes not staged for commit:12:02
n4nd0#   (use "git add <file>..." to update what will be committed)12:02
n4nd0#   (use "git checkout -- <file>..." to discard changes in working directory)12:02
n4nd0#modified:   data (new commits)12:02
blackburn1what if you go to data and git pull origin master?12:02
n4nd0I think that then git st in shogun/data says that there are new commits12:02
n4nd0nah, it says the same12:03
n4nd0everything the same after that12:03
n4nd0git data I just have one remote called origin that points to shogun-toolbox/shogun-data-git12:05
n4nd0.git at the end ^12:05
n4nd0lunch time, see you later12:05
-!- n4nd0 [] has quit [Quit: leaving]12:06
blackburn1strange, n4nd012:14
blackburn1sonney2k: genosse! sind wir ready for release more or less?12:15
blackburn1what is on your list?12:15
-!- emrecelikten [] has joined #shogun12:16
-!- bern4rd [533886e2@gateway/web/freenode/ip.] has joined #shogun16:16
bern4rdhi shogun-toolbox16:17
bern4rdhi sonney2k16:17
-!- blackburn1 [~blackburn@] has quit [Ping timeout: 264 seconds]17:31
-!- n4nd0 [] has joined #shogun17:40
-!- emrecelikten [] has left #shogun []18:30
-!- in3xes [~in3xes@] has joined #shogun18:41
-!- n4nd0 [] has quit [Quit: leaving]18:57
-!- blackburn999 [d5578a74@gateway/web/freenode/ip.] has joined #shogun19:01
@sonney2kbern4rd, around?19:16
@sonney2kn4nd0, you need to do git submodule update19:17
CIA-27shogun: iglesias master * r2482cb1 / examples/undocumented/libshogun/so_multiclass_BMRM.cpp : * fix memory leaks in so_multiclass_BMRM example -
CIA-27shogun: Soeren Sonnenburg master * r7e780c2 / examples/undocumented/libshogun/so_multiclass_BMRM.cpp : Merge pull request #767 from iglesias/master -
CIA-27shogun: Sergey Lisitsyn master * r684cd3a / (7 files in 4 dirs): GP regression memory issues fixes -
blackburn999sonney2k: only base_* left19:39
blackburn9993 examples19:39
@sonney2kbern4rd, alright - no idea if you come back but let me explain a few issues I need help with: the 'download now' knob needs to link to the latest http:// src code download url or the latest news like
@sonney2kbern4rd, the rss feed at the bottom should do the feed for news19:49
@sonney2kbern4rd, I have no idea how to modify the sqlite database to use 'sort_order' columns instead of 'order' ( I got conflicts when attempting to convert the sqlite database)19:50
@sonney2kn4nd0, please run python multiclass_qda.py20:08
@sonney2k(it crashes here...)20:08
bern4rdI'm here,20:12
@sonney2kbern4rd, hey20:13
@sonney2kbern4rd, biggest issue to me is really that I don't know how to rename the 'order' fields to 'sort_oder' in the sqlite db20:13
bern4rdThe rss is not done, but if you want I think I can do it20:13
@sonney2kbern4rd, ok lets do it later (I did this a couple of times so not highest prio)20:14
bern4rdabout order, I have seen you changed the name in models.py20:14
bern4rdbut is necessary to update the db20:15
-!- n4nd0 [] has joined #shogun20:15
@sonney2kn4nd0,  please run python multiclass_qda.py20:15
@sonney2k<sonney2k> (it crashes here...)20:15
@sonney2kblackburn999, does python regression_gaussian_process_demo.py20:15
@sonney2kTraceback (most recent call last):20:15
@sonney2k work for you?20:15
-!- blackburn999 [d5578a74@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:16
@sonney2kbern4rd, I had to - I didn't manage to convert the sqlite db to mysql w/o that20:16
@sonney2kbern4rd, the django way with loaddata (fixtures) didn't work20:16
n4nd0sonney2k: it crashes here too ... I am on it20:17
@sonney2kn4nd0, does python also crash?20:17
@sonney2kblackburn ^ ?20:18
n4nd0sonney2k: yeah, all the way20:18
bern4rdso are you using sqlite3?20:18
n4nd0sonney2k: any clue what change may have caused it?20:18
@sonney2kbern4rd, on ?20:19
@sonney2kn4nd0, no idea20:19
@sonney2kn4nd0, btw python works and gives 100% accuracy - any ideas what is different to the C++ example?20:19
n4nd0sonney2k: I guess that the data is not the same basically20:20
n4nd0in any case, so_multiclass_BMRM in C++ prints out the error not the accuracy20:20
n4nd0I don't know if you noticed about that20:20
n4nd0IIRC so_multiclass_BMRM was really toy data20:20
@sonney2kn4nd0, we need to unify examples.. otherwise it is getting messy: e.g. all MC examples should use the same MC data set20:21
-!- gsomix [~gsomix@] has joined #shogun20:29
-!- gsomix [~gsomix@] has quit [Client Quit]20:29
-!- in3xes [~in3xes@] has quit [Read error: Connection reset by peer]20:30
-!- in3xes [~in3xes@] has joined #shogun20:32
@sonney2kn4nd0, and while we are at it - I think we should also modify graphical examples to be able to be run as batch job creating the figure on disc not only interactively...20:33
-!- bern4rd [533886e2@gateway/web/freenode/ip.] has quit [Quit: Page closed]20:33
-!- bern4rd [533886e2@gateway/web/freenode/ip.] has joined #shogun20:34
CIA-27shogun: Soeren Sonnenburg master * r7bb1224 / (3 files): fix svr and krr examples -
@sonney2k<sonney2k> bern4rd, on ?20:36
@sonney2kbern4rd, ^ did you see this?20:36
bern4rdI am reading about20:37
n4nd0sonney2k: I don't understand that, sorry20:38
@sonney2kn4nd0, basically currently graphical examples are not run on make check-examples20:39
@sonney2kn4nd0, but we could have another target 'create-figures' that would run the graphical examples and each example generates a .png on disc20:39
@sonney2kso we can directly put these online (website...)20:40
bern4rdyou changed from order to sort_order, but is not working. I guess that is because the database is not updated so the sqlite code has "order". I am trying to update the sqlite code.20:40
n4nd0aham, I see now20:40
n4nd0yes, that's a good idea to implement20:40
@sonney2keverything we have to do manually is bug prone...20:40
@sonney2kblackburn - more bugs: run in ocr application ./predict20:49
@sonney2kblackburn ohh it seems all applications are b0rken20:50
CIA-27shogun: Soeren Sonnenburg master * re455486 / applications/edrt/ : fix Euclidian -> Euclidean -
CIA-27shogun: Sergey Lisitsyn master * r308c1b3 / src/shogun/regression/GaussianProcessRegression.cpp : Update src/shogun/regression/GaussianProcessRegression.cpp -
-!- blackburn [d5578a74@gateway/web/freenode/ip.] has joined #shogun21:00
-!- gsomix [~gsomix@] has joined #shogun21:00
CIA-27shogun: Sergey Lisitsyn master * rfaf87e8 / examples/undocumented/libshogun/regression_gaussian_process_fitc.cpp : Update examples/undocumented/libshogun/regression_gaussian_process_fitc.cpp -
@sonney2kbern4rd, any ideas? I mean it is really hard now to develop the site (since one cannot locally run things)21:04
blackburnsonney2k: okay lets check apps21:05
-!- in3xes_ [~in3xes@] has joined #shogun21:05
n4nd0sonney2k: qda_multiclass breaks in ~SGVector21:06
@sonney2kblackburn, but before fix python also crash :)21:06
bern4rdThe problem is that syncdb only update new tables, but not changes in existing tables. I have read about South and Django Evolution that is a tool to edit django database easily. I have installed an i am learning about it21:06
blackburnsonney2k: arts is working21:06
blackburnsonney2k: 2-3 weeks before marina was asking me how to use arts21:06
blackburnit seems I helped more or less21:06
@sonney2kbern4rd, that is too much I think - all we need to do is alter the tables and rename the columns from 'order' to 'sort_order'21:06
blackburnhowever I still do not know what does it do21:06
@sonney2kblackburn, it predicts the start of genes on your dna :D21:07
-!- in3xes [~in3xes@] has quit [Ping timeout: 276 seconds]21:07
bern4rdah ok, for that with the sql command line should be enough21:07
blackburnsonney2k: yeah I said it predicts something she could be interested about21:07
@sonney2kbern4rd, exactly21:07
blackburnsonney2k: I am not interested in things that do not lead to obtain some vodka21:09
@sonney2kblackburn, ohh you could find the gene to tweak that allows you to get high w/o vodka :D21:09
blackburnI am in21:09
blackburnplease guide me where to start21:10
@sonney2kblackburn, release first - drugs later :D21:11
blackburnsonney2k: are base_* leaks blocking you?21:12
blackburnI am unsure it is a correct test now, may be we can live with it21:12
@sonney2kblackburn, errors in applications, graphical examples, base leaks, website...21:13
@sonney2kbern4rd, how are the screenshots done?21:13
@sonney2kI mean the where you have /bigpicture/ links?21:13
bern4rdin i have the method showPicture, this method loads the template bigpicture.html and give the name of the picture selected21:17
bern4rdI got the pictures from the other shogun web-page21:17
@sonney2kbern4rd, ahh so it is taking the picture from static/figures21:18
@sonney2kand the small preview?21:18
@sonney2kdid you resize manually?21:18
bern4rdi resized them manually21:19
n4nd0sonney2k: how is it possible that get_mean causes a call to ~SGVector? I included the method get_column_vector too21:19
bern4rdthe big picture is in static/figuares and the small pictures in static/design21:21
bern4rdi thought in resize them with html and/or css but although you resize them to a smaller one you are loading the big one and needs more time to load the page21:22
@sonney2kbern4rd, ok understood. I will add a few more...21:23
@sonney2kbern4rd, admin page is still under /admin right?21:23
@sonney2kfor some reason I have problems logging in21:23
bern4rdyes, using /admin/ url21:24
CIA-27shogun: Sergey Lisitsyn master * r10beb7a / applications/edrt/tutorial_examples/ : Made isomap tutorial example use normal distance instead of idiot cosine based one -
CIA-27shogun: Sergey Lisitsyn master * r091ebdc / applications/edrt/tutorial_examples/ : Made mds tutorial example use normal distance instead of idiot cosine based one -
blackburnI'm ok don't ask :D21:27
@sonney2kbern4rd, hmmhh for some reason I cannot log in (and I've reset the password...)21:27
bern4rddo you mean in
CIA-27shogun: Sergey Lisitsyn master * r2cbba38 / applications/ocr/ : Updated OCR application due to multiclass labels -
blackburnhelp! I need somebody21:30
bern4rdfor me is working with login:root pass:toor21:30
n4nd0blackburn: what is it?21:30
blackburnI need any of you :D21:30
blackburnn4nd0: okay21:30
blackburnn4nd0: please go to applications/ocr and run train21:30
blackburnwith latest git21:30
blackburnit will update saved SVM model21:31
blackburnso please commit it too21:31
blackburnI am unable to do that because I am using browser and I am under win7 now21:31
blackburneverything is against me lol21:31
n4nd0mmm I cannot really update right now man ...21:32
n4nd0I am trying to find out why qda breaks in ~SGVector21:32
n4nd0is it urgent?21:32
n4nd0I mean like more urgent than fixing qda example21:32
blackburnit depends when sonney2k want to release :D21:32
n4nd0it is another example that breks indeed21:32
blackburnah no, fix it21:32
n4nd0blackburn: can you please take a look to the pastebin I put ^21:33
n4nd0I don't know why can it be that the code there calls ~SGVector21:33
blackburnyeah I've seen - what is the exact problem?21:33
blackburnso it calls destructor and you've got no vector anymore?21:34
n4nd0but why does it call the destructor?21:34
n4nd0it makes no sense for me21:34
@sonney2kblackburn, no release today at least21:35
@sonney2kbern4rd, ok works thx21:35
blackburnn4nd0: how did you detect that?21:35
blackburnsonney2k: yeah I think so too21:35
n4nd0blackburn: that the constructor is called?21:35
n4nd0valgrind complains about it21:35
blackburnn4nd0: destructor?21:35
n4nd0and I am printing it out too, just to confirm21:35
n4nd0yeah, destructor sorry21:35
blackburnwhere is it called? right in that method?21:35
n4nd0after SGVector's constructor21:36
n4nd0it provokes the call to it21:36
@sonney2kblackburn, I forgot to put the vodka on ice - so no way we can release today21:36
blackburnsonney2k: hot vodka tastes better!21:36
blackburnand smells21:36
n4nd0blackburn: somehow, the constructor is calling the destructor this causes to call unref, free_data, and boom!!21:39
blackburnehhm that's strange for sure21:40
CIA-27shogun: Sergey Lisitsyn master * r60f20df / applications/vw/vw.cpp : Updated VW application -
n4nd0I have no idea how it comes that the destructor is called21:44
@sonney2kbern4rd, ok I've put more screenshots under
blackburnsonney2k: can we drop sailfish?21:45
blackburnit is commented whole way21:45
@sonney2kblackburn, n4nd0 regarding the website - shouldn't we have applications instead of screenshots?21:46
@sonney2kerr preview21:46
blackburnyes that's better21:46
blackburnn4nd0: me neither actually21:49
@sonney2kblackburn, drop it21:49
blackburnokay will drop tomorrow21:49
blackburnI have to get some sleep now21:49
@sonney2kblackburn, n4nd0 - what do we do with the follow us on twitter?21:50
n4nd0sonney2k: any clue about the destructor thing?21:50
@sonney2kfollow us on github?21:50
blackburnsonney2k: create an twitter account?21:50
@sonney2kn4nd0, no - buried in the website code21:50
n4nd0put a link to blackburn's twitter21:50
n4nd0sonney2k: ok21:50
blackburnn4nd0: hah that would be strange shogun representative21:51
@sonney2kblackburn, can github twitter when we push things etc?21:51
@sonney2kbuildbot too?21:51
@sonney2kthen twitter makes sense21:51
blackburnno, I think twitter should be for different things21:51
blackburnShogun now have SUPER-X algorithm21:51
blackburnhas I mean :D21:52
n4nd0that is a good idea actually21:52
blackburnsonney2k: nobody is interested about our commits21:52
n4nd0I like it21:52
n4nd0bern4rd and I are twitter's experts :D21:52
@sonney2kn4nd0, so then create some shogun-toolbox twitter account and we can link it21:53
@sonney2kblackburn, github has twitter notifications...21:53
n4nd0we are not good at making accounts, we are good at page-ranking twitter ;)21:53
blackburnwhatever - it makes totally no sense to expose our commits to twitter21:53
@sonney2kblackburn, n4nd0 could you maintain the twitter stream?21:54
blackburnwe should have an account tracking new features and algorithms21:54
blackburnI think I can handle that21:54
@sonney2klike at least weekly news?21:54
@sonney2kok then create a twitter shogun-toolbox account21:54
blackburnokay leaving - lets chat tomorrow more :)21:54
blackburnsee you later21:54
n4nd0blackburn: good night21:54
@sonney2kblackburn, n821:55
-!- blackburn [d5578a74@gateway/web/freenode/ip.] has quit [Quit: Page closed]21:55
@sonney2kn4nd0, it is here
@sonney2kbern4rd, is there any chance to align the arrows in new news box?22:00
gsomixgood evening22:00
bern4rdsonney2k: I have done it for one table, i go to dine and i will do the others22:01
bern4rdprobably is possible, but is html/css stuff22:01
@sonney2kn4nd0, any idea what we should put at the bottom? I mean this 'check out also these great products' things?22:01
@sonney2kbern4rd, it looks weird currently22:02
@sonney2kbern4rd, btw how is the NEWS updated?22:02
bern4rdeach time you call /home url the parser is called, here I check if there is any file in /news directory that is more recent than the last date updatet22:04
bern4rdin other apps i have done i directly create a method loadNews (for example) that remove all elements and reload them again22:05
bern4rdthe first option is not working good?22:05
@sonney2kbern4rd, how is the file getting into the /news directory?22:06 when the object is created22:07
bern4rdin def __init__() the variable newsfolder22:07
@sonney2kbern4rd, I don't get it - do I have to run in some cron job or ?22:08
n4nd0good night guys22:14
-!- n4nd0 [] has quit [Quit: leaving]22:14
@sonney2kbern4rd, ?22:19
-!- gsomix [~gsomix@] has quit [Ping timeout: 252 seconds]22:20
-!- romi_ [~mizobe@] has quit [Ping timeout: 246 seconds]23:07
-!- romi_ [~mizobe@] has joined #shogun23:19
--- Log closed Mon Sep 03 00:00:17 2012