Open in new window / Try shogun cloud
--- Log opened Mon Oct 22 00:00:17 2012
wikingsonney2k: gottit00:01
wikingthnx heaps!00:01
@sonney2kI think now we should focus on polishing website / infrastructure00:02
@sonney2kwiking, I added some minimal dev page00:03
@sonney2kplease improve it00:03
wikingwill try00:03
wikingjust trying to finish up a paper here00:03
@sonney2kgood luck00:03
@sonney2kn4nd0, fern4ndo really did a great job... he wrote a quite flexible system00:03
blackburnsonney2k: bern4rd :D00:03
@sonney2kalright have to change notebooks00:04
@sonney2klack of sleep....00:04
blackburnbernardo :D00:04
blackburnsonney2k: how can I send email from that email?00:15
wikingblackburn: the from field in a mail you can put anything ;)00:15
blackburnwell yeah with some hacks00:16
blackburnbut normally?00:16
wikingno hacks00:16
wikingany email program will do that for u00:16
blackburnY one could ever use email programs in 201200:17
-!- sonne_ [~sonne@] has joined #shogun00:36
sonne_hmmhh I think we should use d3js / mathjax for our html documentation...01:14
sonne_err website01:14
sonne_not the real doc01:14
n4nd0sonney2k: yeah, I got the mail01:22
n4nd0good night people!01:24
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.] has quit [Quit: Page closed]01:24
-!- audy [] has quit [Changing host]01:25
-!- audy [~audy@unaffiliated/audy] has joined #shogun01:25
sonne_ptizoom, still around?01:35
-!- sonne_ [~sonne@] has quit [Ping timeout: 260 seconds]01:57
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]02:03
shogun-buildbotbuild #123 of nightly_none is complete: Failure [failed compile]  Build details are at
shogun-buildbotbuild #140 of nightly_default is complete: Failure [failed test]  Build details are at
-!- n4nd0 [] has joined #shogun07:19
-!- n4nd0 [] has quit [Ping timeout: 246 seconds]08:30
ptizoomto all the sonne*08:33
ptizoomptizoom is still around and somewhat try to discover the art of machine learning...08:34
ptizoomI try to match the best LR parser (with shift, reduce,goto) to some counts of occurrence strings_b "is just after" string_a...08:38
ptizoomand construct, thereof, with such bi-gram likely set of terminal tokens...08:41
ptizoomI am sure shogun will come into play ... somewhere?08:41
-!- n4nd0 [] has joined #shogun09:02
-!- n4nd0_ [] has joined #shogun09:02
-!- n4nd0 [] has quit [Client Quit]09:02
-!- n4nd0_ is now known as n4nd009:02
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun09:17
blackburnwiking: I've got a clearcase work again!09:18
blackburnn4nd0: I have one idea10:25
blackburnnot sure whether it is not crazy, probably it is10:26
blackburncan we develop a structured model that fits skeleton of a human body?10:26
blackburnfrom say multiple cameras, I have such data around10:26
wikingu could do that10:27
n4nd0blackburn: it sounds fancy10:42
n4nd0are you thinking of something like pose tracking?10:43
blackburnn4nd0: yes11:22
blackburnwell one pic should be enough11:22
blackburnbut I do not like to cheat so lets try to develop it11:23
n4nd0blackburn: I think pose tracking is a bit challenging so as to be ready for Friday :O11:24
blackburnn4nd0: we have 'true' data11:24
blackburnI think it is possible to fit it to train data11:24
blackburnn4nd0: sonney2k: I've discussed with guys who have been to USA, they say it is impossible to get a visa in my 'situation'11:54
n4nd0blackburn: :(, your situation = the time span you have?11:54
blackburnmore specifically no foreign passport yet which takes a few weeks11:55
n4nd0that's bad12:01
n4nd0do you know if you can get something temporary?12:01
blackburnno idea12:13
blackburnprobably not12:13
wikingmmmm either the dot product in eigen has something going on13:26
wikingor i dont know13:26
blackburneigen's dot product is ok I think :)13:27
blackburnwe use it in all linear models IIRC13:28
-!- n4nd0 [] has quit [Ping timeout: 260 seconds]13:51
-!- n4nd0 [] has joined #shogun14:03
shogun-buildbotbuild #571 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Heiko Strathmann <>14:21
shogun-buildbotbuild #570 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Heiko Strathmann <>14:26
wikinghow is this possible15:01
wikingsame input15:01
wikingdifferent output15:02
wikingwith mosek15:02
wikingit optimizes differently15:02
wikingbut the whole thing is exactly the same :D15:02
-!- n4nd0 [] has quit [Ping timeout: 256 seconds]15:08
blackburnwiking: A MIRACLE - I found out how to find a commit I wanted to find in clearcase :D15:08
blackburnholy day15:09
@sonney2kblackburn, try it anyway15:19
blackburnsonney2k: about visa?15:19
@sonney2kblackburn, yup15:19
@sonney2kI asked a bit further15:20
blackburnI ordered a foreign passport, once I get it we will see if there is enough time15:20
@sonney2kusually up to 30 organizations apply15:20
@sonney2kabout 5 are accepted15:20
@sonney2k(similar ratio as gsoc)15:20
blackburnso we have no chance15:20
blackburnwell we have but not the big chance15:21
@sonney2kand notifications will be sent out within 1 week15:21
blackburn1 week?!15:21
@sonney2kso it is looking quite good15:21
blackburnahh within15:21
blackburnI thought 1 week *before*15:21
blackburnsonney2k: so do we have 4 shoguneers?15:22
@sonney2kblackburn, wiking, n4nd0, heiko15:23
blackburnheiko is out with NIPS15:23
@sonney2kblackburn, is that decided already?15:23
blackburnI am very probably out with no visa15:23
blackburnsonney2k: about heiko? well he will be visiting NIPS15:24
@sonney2kblackburn, when did he say that?15:24
blackburnhmm am I wrong?15:24
@sonney2k<heiko1> well, then us three at least15:25
@sonney2kthat is the last thing I heard from him15:25
@sonney2k(that was before n4nd0)15:26
blackburnhhmm okay15:26
blackburnanyone else?15:26
blackburnsonney2k: does chris want?15:26
@sonney2kseems not15:26
@sonney2kmaybe me but not too likely15:26
blackburnI am unable to say anything before middle of november15:27
@sonney2kdoesn't matter15:29
blackburnsonney2k: why?15:31
@sonney2kwell we apply assuming that everything works out - if not well bad luck. we cannot do anything about it15:33
blackburnah okay15:33
-!- n4nd0 [] has joined #shogun16:22
-!- yoo__ [2eda6d52@gateway/web/freenode/ip.] has joined #shogun16:46
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:59
-!- n4nd0 [] has quit [Quit: leaving]17:00
-!- heiko1 [] has joined #shogun17:23
-!- heiko1 [] has quit [Remote host closed the connection]17:24
@sonney2kptizoom, do you understand why v_array doesn't work with perl?17:36
@sonney2k(I don't get it)17:36
@sonney2kptizoom, ahh I understand. it clashes with parse_label ( a perl function)17:38
-!- yoo__ [2eda6d52@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:51
@sonney2kptizoom, I managed to get perl / pdl include path etc autodetection for configure working17:55
@sonney2kso what is left to do now is typemaps typemaps typemaps17:55
shogun-buildbotbuild #463 of deb2 - static_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Soeren Sonnenburg <>18:35
ptizoomsonney2k: donot mind you give your patch for the effect your git dev branch ... !18:41
ptizoomalso you have 2 ways of getting typemaps... make it pure perl, or use it through pdl.18:43
ptizoomand remind kindly , I have only done (still buggy!) dense string/vector/matrix conversions.18:44
ptizoomremains... sparse stuff, and general dims arrays. IO FILE* and CODE callback have not been thought through.18:46
ptizoomalso the little bits for "warn" and "print" callbacks through threads... I cannot understand Perl_IO yet!18:48
-!- naywhayare [] has quit [Quit: leaving]18:49
ptizoomand I thought reusing PDL is more attractive than pure is like adding salt and sugar together.18:49
shogun-buildbotbuild #572 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at  blamelist: Soeren Sonnenburg <>19:03
shogun-buildbotbuild #573 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Soeren Sonnenburg <>19:07
shogun-buildbotbuild #464 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.] has joined #shogun19:24
-!- in3xes [~in3xes@] has joined #shogun19:29
-!- ptizoom [] has quit [Quit: Ex-Chat]19:33
-!- ptizoom [] has joined #shogun19:36
-!- ptizoom [] has quit [Client Quit]19:39
-!- blackburn [~blackburn@] has joined #shogun19:54
wikingthis is considered a big error or?19:58
blackburnwiking: ehm error of?19:59
wikingwell this is the difference20:00
-!- in3xes [~in3xes@] has quit [Read error: Connection reset by peer]20:00
wikingbetween my code and the reference code20:00
wikingsupposedly the same dot product20:00
wikingof the a sparse vector20:00
blackburndepends on scale of vectors..20:01
blackburnare they L2=1?20:01
blackburnthey are not20:01
blackburnwhat is abs(dot(v,v)-dot(v',v'))/dot(v,v)?20:01
blackburnpretty small relative error then20:03
wikingwell all in all this accumulates20:03
wikingto a quite bit of an error on the learning part20:03
wikingbut then i guess there's not much i can do about it at this point ...20:04
-!- naywhayare [] has joined #shogun20:23
wikingfloat -> double error :D21:16
blackburnI thought about that21:22
blackburnfloats sucks21:22
wikingthis really fucked me up a whole fucking day21:23
wikingor even a 1.5 :(21:23
wikingok i should test this new sosvm solver21:28
wikingwith other data :P21:29
blackburnSLEP guys get upset with me because of lost citation :D21:50
blackburnI hope they will accept my sorry21:50
blackburnsonney2k: seen mail?22:00
wikingO-SVM: 89.8022:15
wikingBMRM:   88.6022:15
wikingCC:   90.3022:15
wikingMC:     40.5022:15
blackburnwiking: what is the data and what is going on?22:33
shogun-buildbotbuild #574 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at  blamelist: Sergey Lisitsyn <>22:41
wikingblackburn: i've just added that new so classifier to so_multiclass.cpp22:56
wikingto see if it at all works22:58
wikingnow i'm trying to recover svmlight22:58
wikingso that this new solver can work with svmlight as well22:58
wikingnot only with msoek22:58
wikingi need an optimizer that solves this23:18
wikingmin 0.5 w*w + C sum_i C_i \xi_i s.t. x_i * w > rhs_i - \xi_i23:18
-!- heiko [] has joined #shogun23:49
--- Log closed Tue Oct 23 00:00:17 2012