Open in new window / Try shogun cloud
--- Log opened Mon Apr 15 00:00:28 2013
-!- shogun-notifier- [] has quit [Quit: transmission timeout]00:12
-!- hoijui [] has quit [Ping timeout: 248 seconds]00:16
-!- iglesiasg [] has quit [Quit: leaving]01:37
-!- akshaydixi [~akshay@] has joined #shogun02:13
shogun-buildbotbuild #360 of nightly_default is complete: Failure [failed test]  Build details are at
-!- lisitsyn [] has quit [Quit: Leaving.]06:34
-!- gsomix [~gsomix@] has quit [Quit: Leaving]07:20
-!- Calvin-O [73f82292@gateway/web/freenode/ip.] has joined #shogun07:25
-!- hoijui [] has joined #shogun07:39
-!- zuhao [] has joined #shogun07:42
-!- akshaydixi [~akshay@] has quit [Ping timeout: 260 seconds]08:37
-!- akshaydixi [~akshay@] has joined #shogun08:47
-!- sonne|work [] has joined #shogun08:49
-!- Calvin-O [73f82292@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]08:52
-!- gsomix [] has joined #shogun08:58
-!- iglesiasg [] has joined #shogun09:27
-!- mode/#shogun [+o iglesiasg] by ChanServ09:27
gsomixiglesiasg: hey, how are you? what is your project at this year?09:29
@iglesiasggsomix: I am aimint at the structured learning project09:32
@iglesiasggsomix: what about you?09:32
-!- abinash [~abinash@] has joined #shogun09:34
abinashsonney2k: I want to work on the issue Fix and enable blacklisted integration tests.09:36
abinashCould you please be more elaborate it09:37
gsomixiglesiasg: improving of usability again or improving of dimensionality reduction module09:37
@iglesiasgaham the second one with sensei lisitsyn :)09:38
gsomixiglesiasg: yep. it's cool project: ml+modern cpp+data structures...09:40
-!- lambday [3ba21749@gateway/web/freenode/ip.] has joined #shogun09:43
-!- foulwall [~foulwall@2001:da8:215:6819:7dfe:cce2:8883:9ea2] has joined #shogun09:52
foulwallhello guys09:52
foulwalliglesiasg: why the turns out different figure when I click the 'cluster' button ?09:55
sonne|workfoulwall: hey there!09:55
foulwallsonne|work: I've write a clustering demo09:55
gsomixfoulwall: hi. nice support vector regression demo, btw09:55
sonne|workI put it live09:55
sonne|workand I realized we still have some minor issues09:55
foulwallyeah, the title of that page there are incorrect and a lot of improve we can make09:56
sonne|workone was a bit critical - we needed to call .get_labels() to get the  numpy array - that is necessary due to an older numpy version09:56
sonne|workfoulwall: but what is important is that you validate input09:57
sonne|workfoulwall: e.g. for degree of kernel I could enter 'adfadf'09:57
sonne|workand then that causes an exception09:57
sonne|workso you should validate this in the server side (in python) and also in .js09:58
sonne|workfoulwall: back to your question - yes clustering can return different results - it is initialized with some  random number09:58
foulwallok, I'll do that, and pull request09:58
@iglesiasgfoulwall: regarding the clustering example, it is using an algorithm kmeans that I think is randomly initialized09:59
sonne|workfoulwall: one more thing - please be cautious when taking code from other sources - the license has to be compatible with gplv3.09:59
foulwalloh , I'll take care.10:00
gsomixabinash: try asking sonne|work instead sonney2k now :)10:04
abinashsonne|work I want to work  on the issue Fix and enable blacklisted integration tests10:04
abinashcould you please elaborate it10:05
sonne|workhi abinash10:06
abinashHello sonne|work10:06
sonne|worklook at this - inside is a variable blacklist10:07
sonne|workthese are the failing test / examples10:07
sonne|workso just pick one10:07
sonne|workand analyze it to death :)10:08
sonne|workif you know why it fails tell us :)10:08
abinashok thanks, I would try it10:08
sonne|worksome for sure might be tough10:09
sonne|workand others very easy10:09
abinashOk, I am thinking to pickup kernel_distance_modular.py10:10
abinashLet me see whether it is tough or not10:11
-!- abinash [~abinash@] has quit [Quit: Leaving]10:12
foulwallsonne|work: iglesiasg a demo for clustering,left click for positive dots and right click for negative dots10:15
@iglesiasgfoulwall: nice!10:16
foulwalliglesiasg: thanks10:17
gsomixfoulwall: right cliks installs both dots: negative and positive, hm10:17
foulwallgsomix: sorry for that10:17
foulwallI'll fix it10:17
sonne|workfoulwall: cool10:17
@iglesiasggsomix: really? didn't note it here10:17
-!- lisitsyn [] has joined #shogun10:17
sonne|workfoulwall: how did you point the circle?10:17
sonne|workfoulwall: it seems like separated dots?10:18
foulwalluse the polar system10:18
sonne|workfoulwall: no I mean doesn't d3 allow you to draw a circle?10:18
@iglesiasgfoulwall: kmeans should work also fine for more than two classes so I guess the demo could also support that somehow10:19
lisitsynsonne|work: btw our kmeans is b0rken for say manhattan distance10:19
foulwallok i'll make d3 draw the circle10:19
gsomixiglesiasg: just press right button, hold, and move aside10:19
gsomixfoulwall: cool demo :)10:20
@iglesiasggsomix: mmm I only get clue dots here doing that10:20
lisitsynsonne|work: IIRC manhattan distance kmeans objective is minimized with median not mean10:20
@iglesiasgclue -> blue10:20
sonne|workgsomix: doesn't happen here10:20
sonne|workgsomix: which browser?10:20
gsomixhm, strange. Chrome 26.0.1410.64 m, win6410:21
sonne|worklisitsyn: why?10:21
lisitsynsonne|work: recall the objective of kmeans - it is not minimized with mean for arbitrary distance10:22
sonne|worklisitsyn: that's arbitrary no?10:22
lisitsynsonne|work: but we put centers to the mean right?10:22
lisitsynsonne|work: does mean minimize inner cluster distance in case of manhattan distance?10:22
lisitsynmean is a minimizer of least squares (euclidean)10:23
lisitsynnot the sum of absolute differences10:23
lisitsynafaik kmeans is not generic in that10:24
lisitsynit could work with spectral clustering when data is embedded to some euclidean space first10:24
sonne|worklisitsyn: I see what you mean but it is still OK as is if we document it10:24
foulwallthe demo is only a preview, I've found lotsof bugs in it, wait me and I'll show the perfect one.10:24
lisitsynsonne|work: yes sure - I am just totally unsure if results are any reasonable with other metrics10:25
gsomixfoulwall: sonne|work: iglesiasg: same in Firefox 20. probably I have problems with mouse...10:28
-!- lambday [3ba21749@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:29
@iglesiasghehe yeah it makes sense10:29
@iglesiasgfoulwall: sure, no need to rush :)10:29
foulwallgsomix: It's the code's problem10:29
foulwallgsomix: I've found the bug10:30
gsomixfoulwall: ^__^10:34
-!- mode/#shogun [+o lisitsyn] by ChanServ10:35
-!- iglesiasg [] has quit [Ping timeout: 260 seconds]10:36
sonne|workfoulwall: I think it would make more sense instead of right mouse button clicks to have some color buttons (say 10 or so) from which one can choose from10:37
sonne|workand depending on which one is active you get a different label/color circle10:37
sonne|workthis probably is easier too10:37
sonne|workand would work with unlimited number of classes10:37
sonne|workfoulwall: anyway now that you have this with left / right mouse keep it - we can at least use it for binary classification examples10:38
sonne|work(or the first two colors)10:38
-!- eric___ [2eda6d58@gateway/web/freenode/ip.] has joined #shogun10:40
eric___mmmh.. my irc is still down? anyone here ?10:41
@lisitsyneric___: we hear you!10:41
sonne|workeric___: yes you keep asking that question no?10:41
-!- eric___ [2eda6d58@gateway/web/freenode/ip.] has quit [Client Quit]10:42
@lisitsynsonne|work: he must be using a telegraph wire for internet connection10:43
sonne|workit is always the same10:43
sonne|workI think I tried 3 times to answer his question10:43
@lisitsynsonne|work: btw I have a strong reason to don't like JL's v_array10:59
@lisitsynI still didn't manage to get it working under M$ VS10:59
@lisitsynso it is just to win a few milliseconds may be11:00
@lisitsynbut spend all your life to debug it heh11:00
@lisitsynsonne|work: we finally got andreas11:02
@lisitsyn13 mentors11:02
sonne|worklisitsyn: yeah pushing people did help :-)11:04
-!- foulwall [~foulwall@2001:da8:215:6819:7dfe:cce2:8883:9ea2] has quit [Ping timeout: 245 seconds]11:16
-!- iglesiasg [] has joined #shogun11:17
-!- mode/#shogun [+o iglesiasg] by ChanServ11:17
sonne|worklisitsyn: how many did I say we should be?11:30
sonne|workok so who is missing?!11:31
@iglesiasgsonne|work: you are talking about signing up in melange?11:33
@iglesiasgsonne|work: mm I think I never mentioned that to Patrick, maybe he is the one missing?11:33
@lisitsynsonne|work: stegle?11:33
@lisitsyniglesiasg: he did that11:33
@iglesiasgok then11:34
@lisitsynsonne|work: no stegle is here11:34
@lisitsynsonne|work: daniel simpson of heiko11:35
@lisitsynis missed11:35
sonne|workthats it I think11:36
sonne|workI will tell heiko11:36
-!- akshaydixi [~akshay@] has quit [Ping timeout: 246 seconds]11:59
-!- k_matsuda [] has joined #shogun12:04
-!- k_matsuda [] has quit [Quit: Leaving...]12:14
-!- iglesiasg [] has quit [Ping timeout: 256 seconds]12:15
-!- k_matsuda [] has joined #shogun12:15
-!- gsomix [] has quit [Read error: Connection reset by peer]12:42
-!- erlenda [] has joined #shogun13:00
-!- gsomix [] has joined #shogun13:05
-!- d [ca4eaca2@gateway/web/freenode/ip.] has joined #shogun13:23
-!- d is now known as Guest326013:24
-!- zuhao [] has quit [Quit: zuhao]13:32
-!- Guest3260 [ca4eaca2@gateway/web/freenode/ip.] has quit [Quit: Page closed]13:33
-!- zuhao [] has joined #shogun13:43
-!- k_matsuda [] has quit [Remote host closed the connection]13:46
-!- foulwall [~foulwall@2001:da8:215:6819:adc7:ed8d:b359:185f] has joined #shogun13:47
foulwallhey guys14:00
foulwalljust finished bug fix14:01
foulwallno big change14:02
gsomixfoulwall: works proper, cool :)14:03
foulwallsonney2k: gsomix: I'll update a pull request for this and the bug fix for the previous svr demo14:04
sonne|workfoulwall: nice14:05
foulwallI added some check on the input so illegal input can be corrected14:05
sonne|workfoulwall: very good. btw I just tried out svr again. it would make sense to draw lines between the blue points14:07
sonne|workfoulwall: if someone draws some crazy function it looks a bit weird with all those gaps14:07
sonne|workfoulwall: ohh and I noted that one can put points outside of the coordinate system14:08
sonne|workmaybe better limit that14:08
sonne|workfoulwall: sorry one more wish... please make the range 0...1 on the x & y axis... this is better for the polynomial kernel14:09
foulwallsonne|work: you means the user can put points in the negative area?14:09
sonne|workleft of the y axis currently14:10
sonne|workjust next to the axis14:10
sonne|workand below they x axis too14:10
sonne|workfoulwall: sorry for all these (minor) comments...14:10
foulwallsonne|work: that's easy14:10
foulwallsonne|work: no problem, i'll fix those.14:11
foulwallsonne|work: but the point left and bottom are legal and reasonable ones, and won't cause bugs14:12
sonne|workbut they are negative?14:13
foulwallsonne|work: yep14:13
foulwallsonne|work: anyway i'll fix it.14:15
foulwallsonne|work: how to set parameter if I set the axis to range 0...1, the result turn out to be ugly14:40
foulwallsonne|work: nothing need to be changed. my mistake14:41
sonne|workfoulwall: sorry I didn't understand14:42
foulwallsonne|work: I've made a nonsense, ignore it14:43
-!- foulwall [~foulwall@2001:da8:215:6819:adc7:ed8d:b359:185f] has quit [Ping timeout: 245 seconds]14:50
-!- gsomix [] has quit [Quit: Miranda NG! Smaller, Faster, Easier.]15:13
-!- iglesiasg [] has joined #shogun15:28
-!- mode/#shogun [+o iglesiasg] by ChanServ15:28
-!- erlenda [] has quit [Remote host closed the connection]15:54
-!- hoijui [] has quit [Ping timeout: 252 seconds]16:02
-!- k_matsuda [] has joined #shogun16:11
-!- lisitsyn [] has quit [Quit: Leaving.]16:52
-!- hoijui [] has joined #shogun16:54
-!- iglesiasg [] has quit [Ping timeout: 260 seconds]17:05
-!- foulwall [~foulwall@2001:da8:215:c252:341b:429b:b736:ca2b] has joined #shogun17:12
foulwall@sonney2k I adjusted the svr demo's x range to 0...1, but the result figure  is not as good as the previous one, I guess the parameter like cost, tubeeps need changing, no idea how I change them.  I don't know how svr works, help me17:17
foulwallresult curve is not so fitted17:18
foulwallsorry for last nonsense.17:37
-!- foulwall [~foulwall@2001:da8:215:c252:341b:429b:b736:ca2b] has quit [Remote host closed the connection]17:38
-!- shogun-notifier- [] has joined #shogun17:39
shogun-notifier-shogun: lambday :develop * b8612b5 / / (3 files):
shogun-notifier-shogun: added sample_from_gaussian for dense matrices17:39
shogun-notifier-shogun: lambday :develop * 98ba848 / / (3 files):
shogun-notifier-shogun: Merge branch 'test' into feature/sample_gaussian_dense17:39
shogun-notifier-shogun: lambday :develop * 5a7a730 / / (3 files):
shogun-notifier-shogun: Merge branch 'feature/sample_gaussian_dense' into develop17:39
shogun-notifier-shogun: lambday :develop * 6c96678 / / (3 files):
shogun-notifier-shogun: removed new, produces memory leak17:39
shogun-notifier-shogun: lambday :develop * 982ee70 / tests/unit/mathematics/
shogun-notifier-shogun: removed new, produces memory leak17:39
shogun-notifier-shogun: lambday :develop * 9e66058 / / (3 files):
shogun-notifier-shogun: fixed memory leak in sample_from_gaussian17:39
-!- heiko [] has joined #shogun17:40
-!- LeStealth [8e018170@gateway/web/freenode/ip.] has joined #shogun17:52
-!- LeStealth [8e018170@gateway/web/freenode/ip.] has quit [Client Quit]17:52
-!- blackburn [~blackburn@] has joined #shogun17:52
-!- BenYu [8e018170@gateway/web/freenode/ip.] has joined #shogun17:53
-!- iglesiasg [] has joined #shogun18:09
-!- mode/#shogun [+o iglesiasg] by ChanServ18:09
-!- blackburn is now known as lisitsyn18:12
-!- lisitsyn [~blackburn@] has quit [Quit: Leaving.]18:13
-!- lisitsyn [~blackburn@] has joined #shogun18:13
-!- mode/#shogun [+o lisitsyn] by ChanServ18:14
-!- hoijui [] has quit [Ping timeout: 245 seconds]18:19
shogun-buildbotbuild #361 of nightly_default is complete: Failure [failed test]  Build details are at
-!- heiko [] has quit [Quit: Leaving.]18:21
-!- Heiko [] has joined #shogun18:22
-!- Heiko [] has quit [Client Quit]18:22
-!- HeikoS [] has joined #shogun18:24
HeikoSsonney2k, registered my nickname18:24
-!- iglesiasg [] has quit [Ping timeout: 248 seconds]18:36
-!- BenYu [8e018170@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:06
-!- mode/#shogun [+o HeikoS] by ChanServ19:14
@sonney2kHeikoS, I will have to disable the website buildbot frontend19:16
@sonney2ktoo many spam bots...19:16
-!- shogun-buildbot [] has quit [Quit: buildmaster reconfigured: bot disconnecting]19:18
-!- shogun-buildbot [] has joined #shogun19:18
@sonney2kshogun-buildbot, force build nightly_default19:18
shogun-buildbotbuild forced [ETA 35m38s]19:18
shogun-buildbotI'll give a shout when the build finishes19:18
@sonney2kshogun-buildbot, force build 'rpm1 - libshogun'19:19
shogun-buildbotbuild forced [ETA 7m25s]19:19
shogun-buildbotI'll give a shout when the build finishes19:19
shogun-buildbotbuild #1010 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Soeren Sonnenburg <>19:20
@sonney2kshogun-buildbot, force build 'deb3 - modular_interfaces'19:22
shogun-buildbotbuild forced [ETA 44m36s]19:22
shogun-buildbotI'll give a shout when the build finishes19:22
shogun-buildbotbuild #424 of rpm1 - libshogun is complete: Success [build successful]  Build details are at
-!- medeeiip [~medeeiip@] has joined #shogun19:43
shogun-buildbotbuild #1011 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at
@HeikoSsonney2k: the above fail could have been prevented if python tests would be performed before merge ...20:02
-!- HeikoS [] has left #shogun []20:03
shogun-buildbotbuild #362 of nightly_default is complete: Failure [failed test]  Build details are at
@sonney2kHeikoS well it is your responsibility - you've merged this :P20:13
-!- hoijui [] has joined #shogun20:19
-!- medeeiip [~medeeiip@] has quit [Ping timeout: 252 seconds]20:29
-!- lambday [0e8b6149@gateway/web/freenode/ip.] has joined #shogun20:30
lambdaylisitsyn: hi...20:31
@lisitsynlambday: hey20:31
lambdaytwo questions - one.. say i am already in a feature branch... made some commits, but then I want to work on another feature branch.. and finish and push that one first..20:32
lambdayso,  if i just do git checkout develop... and then start the new feature branch... it will work?20:33
@lisitsynyou may change your branches20:33
@lisitsynno problem20:33
lambdaylisitsyn: alright.. thanks... another question20:33
lambdayi think we'll be needing the SGSparseMatrix to eigen's SparseMatrix quite often.. we talked about it and you told that it will slow down the swig stuffs... is there any way, we could define this, some util may be, so that it doesn't harm much?20:35
@lisitsynlambday: yes, just implement it in other class20:35
@lisitsynthat is not exposed to swig20:35
lambdayummm.. okay let me check20:37
-!- shogun-notifier- [] has quit [Quit: transmission timeout]20:39
-!- gsomix [~gsomix@] has joined #shogun20:44
gsomixgood evening20:44
-!- lambday [0e8b6149@gateway/web/freenode/ip.] has quit []21:04
-!- medeeiip [~medeeiip@] has joined #shogun21:05
-!- iglesiasg [] has joined #shogun22:01
-!- mode/#shogun [+o iglesiasg] by ChanServ22:01
@sonney2klisitsyn, errm is there any way to place an abstract when registering for the workshop?22:07
@lisitsynsonney2k: abstract of?22:08
@lisitsynsonney2k: there is a field for that already22:09
@sonney2klisitsyn, ok then22:21
@sonney2kiglesiasg, lisitsyn please polish the CfP for the workshop...22:21
@sonney2ktook me ages I am pretty rusy22:22
@lisitsynsonney2k: oh 2 days at TU?22:23
@sonney2klisitsyn, not?22:23
@sonney2konly I am not around22:23
@lisitsynsonney2k: I don't mind - just asking if this is already signed22:23
@sonney2kno it is not22:23
@sonney2klisitsyn, I am sending this to gunnar and chris etc etc too22:27
@sonney2kalright cu22:27
@iglesiasgsonney2k: I think it is pretty well, nothing I can come up with to make it better ATM22:27
@lisitsynsonney2k: yes it is pretty well-written22:27
@sonney2kI don't like it that much22:27
@iglesiasgsonney2k: oh BTW, I guess we don't need to apply for the registration but maybe I am wrong22:29
-!- medeeiip__ [~medeeiip@] has joined #shogun22:41
@iglesiasgsonney2k, lisitsyn: one thing, since we are now merging to develop with the PRs, we lost this nice message from travis22:46
@iglesiasgI mean the one that appears here:
@iglesiasgbut not here:
@iglesiasgI think it would be nice if we can still keep that22:47
@lisitsyniglesiasg: could you please check the travis doc to find out if it is possible to set up a default branch?22:48
@iglesiasglisitsyn: ok, let me google a bit22:48
-!- medeeiip_ [~medeeiip@] has joined #shogun23:12
medeeiip_in shogun io module in some function like CAsciiFile::get_string_list() uint8_t* is used instade of char*. What is the reason behind doing so? afterall the file will contain char value.....23:14
-!- medeeiip__ [~medeeiip@] has quit [Quit: Leaving]23:14
-!- medeeiip_ [~medeeiip@] has quit [Client Quit]23:15
-!- gsomix [~gsomix@] has quit [Ping timeout: 256 seconds]23:15
-!- zuhao [] has quit [Quit: zuhao]23:22
-!- hoijui [] has quit [Ping timeout: 260 seconds]23:37
--- Log closed Tue Apr 16 00:00:28 2013