Open in new window / Try shogun cloud
--- Log opened Wed Jun 27 00:00:17 2012
n4nd0good night!00:18
-!- n4nd0 [] has quit [Quit: leaving]00:19
CIA-18shogun: Soeren Sonnenburg master * rc9ae118 / src/interfaces/modular/Transfer.i : fix compile error on java modular -
-!- av3ngr [] has joined #shogun01:15
-!- av3ngr [] has quit [Client Quit]01:15
-!- ipondering [~iponderin@] has joined #shogun04:35
-!- ipondering [~iponderin@] has left #shogun []04:35
-!- ipondering [~iponderin@] has joined #shogun04:36
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has joined #shogun05:18
-!- cheng [80fa500f@gateway/web/freenode/ip.] has joined #shogun06:14
-!- Marina [55b3252a@gateway/web/freenode/ip.] has joined #shogun07:24
Marinahello everyone, i have the following question:07:25
Marinaafter training a svm with a weighted degree kernel07:25
Marinais it possible to get the w vector ?07:26
Marinai know that i can get the weight of the support vecotrs - the alphas - or is it possible to reonstruct the w by the alphas?07:27
-!- cheng [80fa500f@gateway/web/freenode/ip.] has quit [Quit: Page closed]07:59
@sonney2kMarina, the w is 10^20 dimensional or so but sparse - so it is stored in a tree - so no not really08:04
Marina@sonney2k thanks!!08:44
-!- Marina [55b3252a@gateway/web/freenode/ip.] has quit [Quit: Page closed]08:46
-!- uricamic [~uricamic@2001:718:2:1634:a5ac:d325:440c:2bf2] has joined #shogun08:51
CIA-18shogun: Soeren Sonnenburg master * r60e54ed / (51 files in 7 dirs): Merge pull request #603 from puffin444/master (+11 more commits...) -
-!- Netsplit *.net <-> *.split quits: ipondering09:06
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]09:07
-!- Netsplit over, joins: ipondering09:08
-!- n4nd0 [] has joined #shogun09:42
-!- ipondering_ [~iponderin@] has joined #shogun09:44
CIA-18shogun: Soeren Sonnenburg master * rbd0fd92 / (2 files): remove german umlaut from code -> ascii only -
-!- ipondering [~iponderin@] has quit [Ping timeout: 264 seconds]09:48
-!- ipondering_ is now known as ipondering09:48
CIA-18shogun: Soeren Sonnenburg master * r4a337b8 / src/shogun/evaluation/MachineEvaluation.h : fix typo -
CIA-18shogun: Soeren Sonnenburg master * rd74ef69 / (3 files): fix warnings in modular interfaces -
-!- gsomix [~gsomix@] has joined #shogun10:21
gsomixgood morning10:21
-!- ipondering [~iponderin@] has quit [Quit: ipondering]10:24
n4nd0good morning gsomix10:30
-!- heiko [] has joined #shogun11:25
-!- cheng [] has joined #shogun11:42
-!- blackburn [~blackburn@] has joined #shogun11:46
-!- blackburn [~blackburn@] has quit [Read error: Operation timed out]12:34
-!- pluskid [~pluskid@] has joined #shogun12:43
-!- cheng [] has quit [Quit: Leaving.]12:53
-!- cheng [] has joined #shogun13:30
-!- n4nd0 [] has quit [Ping timeout: 246 seconds]13:31
-!- n4nd0 [] has joined #shogun14:01
-!- cheng [] has quit [Quit: Leaving.]14:42
-!- heiko [] has quit [Ping timeout: 264 seconds]14:53
-!- pluskid [~pluskid@] has quit [Ping timeout: 246 seconds]15:37
-!- pluskid [~pluskid@] has joined #shogun15:37
-!- alexlovesdata_ [82955843@gateway/web/freenode/ip.] has joined #shogun16:13
-!- heiko [] has joined #shogun16:16
-!- n4nd0 [] has quit [Quit: leaving]16:24
-!- blackburn [5f43cee3@gateway/web/freenode/ip.] has joined #shogun16:35
-!- pluskid [~pluskid@] has quit [Quit: Leaving]16:37
-!- zxtx [] has quit [Ping timeout: 248 seconds]16:52
-!- uricamic [~uricamic@2001:718:2:1634:a5ac:d325:440c:2bf2] has quit [Quit: Leaving.]16:53
-!- blackburn [5f43cee3@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:58
-!- alexlovesdata_ [82955843@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:15
-!- blackburn [5f43cee3@gateway/web/freenode/ip.] has joined #shogun17:18
@sonney2khmmhh whenever I need n4nd0 he is not there :D17:57
@sonney2kmust be that we live in different timezones ;)17:57
blackburnsonney2k: I just got first publication ever :D18:02
@sonney2kblackburn, congrats18:02
@sonney2know a 100 more and you will be professor :D18:02
blackburnhahah lol18:03
@sonney2kand don't forget to do shogun coding practicals then :D18:03
heikoblackburn, congrats :)18:03
blackburnsonney2k: n4nd0 is usually here nights18:03
blackburnI hope JMLR paper could be accepted too - we will do an attempt18:03
@sonney2kthe dimred stuff18:03
@sonney2kblackburn, can you please fix the libshogun example?18:07
blackburnsonney2k: didn't I?18:07
@sonney2kI'd like to get things back to normal18:07
@sonney2kdid you?18:07
* sonney2k checks18:07
blackburnsonney2k: ok I'll fix18:08
@sonney2khmmhh puffin is no longer here - we also have a problem with nlopt18:08
@sonney2kblackburn, it is not yet in debian/ubuntu right?18:08
blackburnI have to get home (will be here in 35 mins)18:08
blackburnsonney2k: what is not yet in debian?18:09
@sonney2kor is it?18:09
blackburnno idea let me check18:09
blackburnit is very easy to install18:09
@sonney2kblackburn, how did you add support for nlopt then?18:09
blackburnsonney2k: installed from sources18:10
blackburnit is as easy to install as shogun18:10
blackburnwell if one have no nlopt he would not have some great stuff from puffin18:11
blackburnsounds fair, isn't it?18:11
@sonney2kblackburn, well examples crash currently with nlopt not beinga vailable it seems18:11
blackburndifferent issue then18:11
blackburnok give me 30-35 mins for a bike ride18:16
blackburnbrb :D18:16
-!- blackburn [5f43cee3@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:16
heikosonney2k I read a bit about how German Vereine work18:16
heikosome open-source project are registered as one18:16
heikoI think that has some nice benefits18:16
heikohowever, quite some effort is needed to start this off18:17
heikoBut these money problems wouldnt be there since the Verein has the right to collect/gain money, as long as it is used for purposes of its manifest18:17
heikoIll check if there is like an internet-version of it18:18
CIA-18shogun: Soeren Sonnenburg master * r9d3c1a1 / (5 files): fix compile errors in java and csharp modular -
-!- gsomix [~gsomix@] has quit [Ping timeout: 246 seconds]18:32
@sonney2kheiko, any idea why modelselection_parameter_tree.cpp fails?18:32
heikoIll check18:33
@sonney2ksonney2k, how many people do we need to sign the manifest?18:33
heikowhich is a lot for us :)18:33
@sonney2kheiko, all Germans?18:33
heikoI think so, yes, but let me checlk18:33
@sonney2kwell, Gunnar, Chris, Heiko, Nico, Alex, Soeren18:34
@sonney2kone missing18:34
heikomembers can be non-german18:36
@sonney2kand founders?18:37
heikobut I cannot find anything for the founders18:37
heikoseems to be allowed18:39
heikoIll ask my parents later, my father knows about that stuff18:39
@sonney2kheiko, cool18:40
@sonney2kI think this really is the best option for us18:40
-!- zxtx [~zv@] has joined #shogun18:40
heikoThere is another alternative without notar, but I think then we would have to pay taxes, so e.V. would be best18:41
-!- gsomix [~gsomix@] has joined #shogun18:42
@sonney2kheiko, so we would have to create sth along these lines
@sonney2kthen pay ~100 EUR for notar/ 'Gericht;18:44
@sonney2kand done18:44
heikowhich is fine for shogun18:44
CIA-18shogun: Soeren Sonnenburg master * rcd63a59 / examples/undocumented/libshogun/regression_gaussian_process.cpp : disable example when NLOPT is not available -
-!- zxtx [~zv@] has quit [Ping timeout: 240 seconds]18:48
CIA-18shogun: puffin444 master * rf4b4c8f / examples/undocumented/libshogun/regression_gaussian_process.cpp : Allowed the Gaussian Process Regression example to behave gracefully -
CIA-18shogun: Soeren Sonnenburg master * r6e9cc41 / examples/undocumented/libshogun/regression_gaussian_process.cpp : Merge pull request #606 from puffin444/master -
CIA-18shogun: Chiyuan Zhang master * r48af5bd / (2 files): added subset features to modular interface. -
CIA-18shogun: Chiyuan Zhang master * rf5ef596 / (3 files in 2 dirs): Added ShareBoost python modular interface and example -
CIA-18shogun: Chiyuan Zhang master * r21fbe11 / (3 files): slight improvements to some other MC py-modular examples -
CIA-18shogun: Soeren Sonnenburg master * reade5bd / (7 files in 2 dirs): Merge pull request #605 from pluskid/multiclass -
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has joined #shogun18:53
puffin444hey heiko18:54
heikohey puffin44418:54
puffin444looks like I broke the build.18:54
heikoyeah :) I was just about checking ...18:54
puffin444I just sent in a fix.18:54
heikoor do you already know the problem?18:54
heikosoeren already merged it18:55
puffin444Yes. In the example I forgot to handle the case when NLOPT is not available.18:55
puffin444Oh good.18:55
@sonney2kpuffin444, I will revert my fix18:55
heikothere also was an error in another test18:55
@sonney2kbut there is one more problem in the other example18:55
heikosonney2k, is this parameter tree example still failing?18:55
puffin444Which one?18:55
@sonney2kheiko, yes18:55
heikok let me check it now18:55
@sonney2ksparse subset stuff is what blackburn will fix18:56
-!- blackburn [~blackburn@] has joined #shogun18:57
@sonney2k...who just logged in :D18:57
blackburnwhat's up?18:57
heikopuffin444 modelselection_parameter_tree.cpp18:57
puffin444Oh I see what's wrong. The parameter is called "Likelihood Model," not "model"18:58
CIA-18shogun: Soeren Sonnenburg master * r7bac16a / examples/undocumented/libshogun/regression_gaussian_process.cpp : Revert "disable example when NLOPT is not available" -
heikosonney2k, I just checked modelselection_parameter_tree.cpp. it doesnt fail here18:59
heikoah wrong file sorry18:59
blackburnsonney2k: why did you revert it?19:00
blackburnwhat is wrong?19:01
heikopuffin444, sonney2k, I spotted the problem19:01
heikopuffin444, you renamed he parameters19:01
heiko"model" -> "Likelihood Model"19:01
heikoplease dont do that, always use lower case single words, or if needed underscores19:02
blackburnpuffin444: I can share some experience on SSD19:02
-!- zxtx [~zv@] has joined #shogun19:02
puffin444blackburn, how is it?19:02
blackburnpuffin444: ok I can say word 'awesome' or say win7 fully reboots in 20s19:03
@sonney2kread speed is really impressive19:04
blackburnI am going to install 4 gb ram more19:04
@sonney2kthough write speed at least for mine is not19:04
blackburnand start to switch19:04
@sonney2kat least not when compiling19:04
blackburnto new one19:04
blackburnsonney2k: hmm can we set up compilation to separate dir?19:04
blackburnsonney2k: these files are pretty small19:05
blackburnwell okay actually I'll test compilation on ssd tomorrow19:05
blackburnmaybe modern ssds are better with it19:05
heikoblackburn whats the problem with ssds?19:07
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:07
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has joined #shogun19:08
@sonney2kyeah I think newer ssd's don't have these stallation problems19:09
blackburnheiko: is there any problem?19:09
heikoblackburn, what problem is there with ssd's?19:09
blackburnheiko: I don't know any :)19:09
heikoI got one and everything compiles ultra fast19:09
blackburnbut Soeren said compilation is a little slow19:09
blackburnokay nice19:09
heikoI got one of the ibm ones, think most recent one19:10
heikowithout any problems19:10
blackburnI have samsung (just like notebook vendor)19:10
blackburnbut I didn't manage to install good os here yet19:10
@sonney2kAPPLE SSD TS256A here19:11
@sonney2kwhatever that means :D19:11
@sonney2kheiko, so do you fix the bug?19:12
@sonney2kand blackburn are you working on it?19:12
heikopuffin444 did you change any other names?19:12
@sonney2kwhich reminds me - wiking - how is it going?19:12
heikoif not I will change these back to nold values19:12
blackburnkick ass time? sonney2k19:13
@sonney2kwiking, you can send small patches already now19:13
@sonney2kno need to wait until tomorrow19:13
@sonney2kblackburn, well wiking promised the PR with everything until tomorrow...19:13
@sonney2kso better ask :D19:13
@sonney2kblackburn, have you seen the movie 'kick-ass' ?19:14
blackburnhmm let me google it19:14
@sonney2kblackburn, you should :D19:14
blackburnprobably no19:14
blackburnwill do19:15
@sonney2k...but fix the bug before19:15
puffin444heiko, a few. I am fixing this now.19:15
puffin444If you don't mind I'd like to use the name "likelihood_model"19:17
heikopuffin444 sure, could you change that in the example that failed?19:18
puffin444Yes, I will19:18
heikosonney2k, do you remember the problem of appending features?19:23
-!- n4nd0 [] has joined #shogun19:23
heikowe discussed that and agreed on adding a convienience method that appends two given features19:24
n4nd0sonney2k: I see in the logs you were looking for me?19:24
blackburnOkay sparse transposed is a little buggy19:24
heikois there anyway of appending features from the CFeatures interface?19:24
heikocopying is ok19:24
puffin444heiko, I have a pull request that should fix it.19:26
CIA-18shogun: puffin444 master * rc34d4bd / (7 files in 5 dirs): Fixed some parameter names and examples. -
CIA-18shogun: Heiko Strathmann master * rdb17b24 / (7 files in 5 dirs): Merge pull request #607 from puffin444/master -
heikopuffin444, thanks :)19:27
puffin444Your welcome. I am leaving, but I should be back later today.19:27
heikosame here19:28
puffin444To get some of those kernel derivatives up and running.19:28
-!- puffin444 [62e3926e@gateway/web/freenode/ip.] has quit [Quit: Page closed]19:28
-!- heiko [] has quit [Ping timeout: 264 seconds]19:37
n4nd0sonney2k: tell me what was it if so and I will check it later :)19:46
@sonney2kn4nd0, fixed it already19:50
@sonney2kn4nd0, some compile error in modular java interfaces19:51
@sonney2kheiko, I do remember - did any problems appear19:52
@sonney2kn4nd0, so you want spain to win tonight?20:12
blackburnsonney2k: you were pretty inactive about football20:12
n4nd0sonney2k: of course20:19
n4nd0sonney2k: btw, coming back to the discussion of the features20:22
n4nd0what I told you about yesterday, the MatrixFeatures20:23
blackburnmatrix features?20:23
n4nd0for the HMSVM20:23
n4nd0each example is represented with a "matrix"20:23
n4nd0so each of the features of an example is a sequence20:24
n4nd0BUT it is not compulsory that all the features have the same length20:24
n4nd0blackburn: do you see what I mean?20:25
blackburnso variable size matrix?20:25
n4nd0yeah, maybe20:25
n4nd0blackburn: what type of shogun features do you think I should use?20:26
n4nd0I think I need to create something new20:26
blackburnhmm well some new one I think20:26
n4nd0let me show you20:27
n4nd0I sketched them a bit20:28
n4nd0although now this needs to be changed20:28
blackburnI actually though about it20:28
n4nd0since Nico has told me that each of the features can have different length20:28
blackburnfor images20:28
blackburnso I second this idea20:28
n4nd0what do you think I should use to handle this?20:29
n4nd0that the length of the features may change20:29
n4nd0I have not thought about it yet (I just got the mail)20:29
n4nd0brain storm :D20:29
n4nd0I think I can use directly the dynamic object array20:30
n4nd0mmmm although I don't want to store objects but something simple as integers20:31
blackburnwell some array of matrices20:31
n4nd0not exactly matrices though20:31
n4nd0since one of the dimensions is not fixed20:32
n4nd0I think it should be something like an array of StringFeatures20:36
blackburnn4nd0: I am sorry but I want portugal to win :D20:37
@sonney2kn4nd0, yeah portugal is much easier to beat ;)20:38
n4nd0blackburn: haha it's fine20:38
@sonney2kn4nd0, well have an array of SGMatrices then20:38
blackburnportugal is easy to beat? lets see heh20:38
@sonney2kblackburn, well Germans beat them ...20:39
n4nd0sonney2k: but the thing is that each SGMatrices are supposed to be of fixed num_rows and num_cols20:39
@sonney2kbut spain is tough20:39
blackburninduction does not work here20:39
n4nd0what I need doesn't have one of the dimensions fixed20:39
@sonney2kn4nd0, ???20:39
@sonney2kI thought you have several matrices of different shape?20:40
@sonney2khow can they not be fixed?20:41
n4nd0sonney2k: sorry, I don't know what I wrote before20:48
n4nd0they were speaking to me and I lost the focus20:48
n4nd0sonney2k: so it is not several matrices of different shape20:48
n4nd0so each matrix represent a training example20:48
n4nd0all the training examples (all the matrices then) have the same number of features20:49
n4nd0=> the number of columns for example is the same for all the matrices20:49
n4nd0BUT each of the features can be of different length20:49
n4nd0both within the same example and among different examples20:49
n4nd0=> the number of rows (even for a matrix) is not fixed20:50
-!- n4nd0 [] has quit [Ping timeout: 252 seconds]20:56
@sonney2kn4nd0 - but then it is ok - just use some array of matrices20:59
@sonney2kthe same way like string/sparse features do it20:59
@sonney2kyou cannot set them all at once though then21:00
blackburnsonney2k: uh21:12
CIA-18shogun: Sergey Lisitsyn master * rc9b592d / src/shogun/features/SparseFeatures.cpp : Fixed get transposed in sparse features -
-!- gsomix [~gsomix@] has quit [Ping timeout: 250 seconds]21:19
@sonney2kblackburn, hurray21:29
@sonney2kblackburn, looks like you are right21:33
@sonney2kportugal looks pretty good21:33
-!- zxtx [~zv@] has quit [Ping timeout: 264 seconds]21:33
blackburnvery active and fast21:35
@sonney2kI bet 0:021:35
blackburnand christianny21:35
blackburnno penalty kicks please :)21:36
@sonney2kI am too tired already now...21:36
shogun-buildbotbuild #972 of python_static is complete: Success [build successful]  Build details are at
blackburnronaldo has broken aim22:32
@sonney2kbias term needs a fix22:32
blackburnmay be he uses (w,x)+b and ball prefer (w,x)-b22:33
@sonney2kfrom what I have seen spain is mostly harmless today22:34
blackburnyeah it should be kind of luck22:34
@sonney2ktoo bad that there was no early goal22:34
@sonney2kthese 0:0 games are :((22:35
blackburnway too much yellow22:35
blackburn8-9 already?22:35
@sonney2kI wonder when the first red card will be given22:35
blackburnprobability is high already22:36
@sonney2kportugal can change 2 more times right?22:36
blackburnyes I think so22:37
@sonney2kso maybe they have some plan...22:37
blackburnto fix bias22:37
@sonney2kI recommend valgrind ;-)22:39
blackburnlets have some goal22:40
blackburnI always laugh hearing ?????? yells22:40
blackburnrussian fans bought tickets in hope russia will go through :D22:41
@sonney2kblackburn, some early goal please22:43
@sonney2kotherwise I will fall asleep22:43
@sonney2kthe referee is pretty good though!22:45
blackburnhe spends so much time22:45
@sonney2kohh well penalties are pretty likely...22:48
blackburnsonney2k: russian commentator said 'no sleep! penalties are near!' :D22:53
@sonney2kblackburn, are you the commentator?22:53
blackburnI am not, why?22:54
@sonney2kbecause you said  'no sleep! penalties are near!' :D22:54
blackburnthat was dangerous22:54
@sonney2kyes first dangerous thing22:54
@sonney2kbut portugal can still change 2 times...22:55
blackburncustodio is going to appear22:55
@sonney2kI think he will change 2 at once22:56
blackburngood auROC22:57
blackburnbad bias22:57
@sonney2kalright chocolate and stuff will keep me awake22:57
@sonney2ksome sweet beer :D23:01
blackburnbeer with chocolate?23:02
-!- zxtx [~zv@] has joined #shogun23:02
@sonney2kseems like both teams aim for penalty kicks23:03
blackburn12? 13?23:06
blackburnall yellow23:06
blackburnI thought more23:07
@sonney2khow many via ronaldo?23:07
blackburnno idea23:07
blackburnpedro pedro23:07
@sonney2kdo they have rose shoes?23:08
@sonney2kthe spanish I mean23:08
blackburnI didn't notice23:08
blackburnronaldo has23:08
@sonney2krose/blue or?23:09
blackburnisn't it very individual?23:09
@sonney2kno idea23:09
blackburnspain is tough now23:10
@sonney2kcome on FFWD23:10
@sonney2kyeah spains standard tactics works now23:11
@sonney2kportugal is out of energie23:11
@sonney2kenergy regularizer does it's job23:11
blackburnhah kerzhakoved23:11
blackburn(new word came from russian player - means bad miss) :D23:12
@sonney2k1 more minute...23:13
blackburnyeah almost done23:13
@sonney2kawful game23:14
@sonney2khope tomorrow there will be one goal within the first 15 minutes23:14
@sonney2know it is time for penalty roulette23:14
blackburnI almost got tree stuff done during this match23:14
@sonney2kin that case there will be a 0:0 tomorrow too :23:16
@sonney2kmy bet is that ronaldo will shoot above the goal23:18
blackburnlets see23:19
blackburnhe could23:19
@sonney2klets see if they can turn it around23:20
@sonney2k0:0 after penalties :D23:21
blackburnreally near to save23:22
@sonney2ksome goals after all23:23
blackburnpepe is kind of bull23:23
blackburnI thought he would break net23:23
blackburnsymmetric still23:23
blackburnidentical goals identical saves23:23
@sonney2ktoo bad23:25
@sonney2kthat's it then23:26
blackburnspain probably23:26
-!- zxtx [~zv@] has quit [Ping timeout: 245 seconds]23:26
@sonney2kcould have been the other way round aswell23:27
blackburnbad bar miss23:27
blackburnokay lets see germany-spain on sunday :D23:28
@sonney2knahh I don't want to see that ;)23:29
blackburnwould you like to see italy in finals?23:33
-!- n4nd0 [] has joined #shogun23:38
n4nd0blackburn: :)23:39
@sonney2kn4nd0, lucky you :D23:40
@sonney2kblackburn, surely *not*23:40
n4nd0sonney2k: c'mon ... we deserved it in the extra time!23:40
@sonney2kn4nd0, nope23:41
-!- zxtx [~zv@] has joined #shogun23:41
n4nd0haha I see in the logs you have been talking about the match23:41
@sonney2kspain was not superior this time23:41
n4nd0not until the extra time23:42
@sonney2kbut I don't expect germany to win against italy (or spain for that matter)23:42
@sonney2kn4nd0, yes23:42
n4nd0I think that Germany can do it23:42
@sonney2kbut still just 1 chance23:42
@sonney2kitaly played pretty rough last time23:43
@sonney2kwasn't it in semi-finals too?23:43
n4nd0wow and what about Ramos' penalty?23:43
n4nd0I couldn't believe that23:43
n4nd0I am Real Madrid supporter and suffered the one against Bayern ...23:43
@sonney2kwiking promised to send a PR thursday night23:44
@sonney2kso someone around/after the game :D23:44
@sonney2ktime to sleep23:45
n4nd0sonney2k: good night!23:45
-!- zxtx [~zv@] has quit [Ping timeout: 244 seconds]23:48
--- Log closed Thu Jun 28 00:00:17 2012