Open in new window / Try shogun cloud
--- Log opened Fri Apr 19 00:00:28 2013
-!- utsavjain [~tiripS@] has left #shogun []01:27
-!- iglesiasg [] has quit [Quit: leaving]04:16
-!- vikram360 [~vikram360@] has joined #shogun04:41
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]05:59
-!- Shell_ [80c1080c@gateway/web/freenode/ip.] has joined #shogun06:07
-!- foulwall [] has joined #shogun07:26
foulwallsonney2k: sonne|work: which version of python do we use? I use a importlib in the PR, and need python version>=2.707:28
-!- Shell_ [80c1080c@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]07:41
-!- hoijui [] has joined #shogun08:12
foulwallsonney2k: I came up with an idea if I'm at leisure I can do some Chinese translation for shogun-web and make docs a Chinese version08:14
-!- travis-ci [] has joined #shogun08:42
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []08:42
-!- travis-ci [] has joined #shogun08:43
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []08:43
-!- foulwall [] has quit [Read error: Connection reset by peer]09:17
-!- gzhd79 [] has joined #shogun09:21
-!- vikram360 [~vikram360@] has quit [Ping timeout: 256 seconds]09:23
-!- foulwall [] has joined #shogun09:29
-!- foulwall [] has quit [Ping timeout: 272 seconds]09:36
-!- foulwall [] has joined #shogun09:38
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun09:41
-!- foulwall` [] has joined #shogun09:42
-!- foulwall [] has quit [Remote host closed the connection]09:43
-!- foulwall` [] has left #shogun []09:43
-!- foulwall [] has joined #shogun09:47
-!- foulwall [] has quit [Remote host closed the connection]09:48
-!- foulwall [] has joined #shogun09:51
foulwallsonney2k: Hey there?09:52
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun10:21
sonne|workfoulwall: hey!10:23
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:24
foulwallsonne|work: I've mentioned in the email. we need a module named importlib, just pip it on the server.10:29
sonne|workfoulwall: why do we need it?10:29
foulwallsonne|work: since we need to import the function at runtime, and the old __import__ is messy.10:31
foulwallsonne|work: importlib is just a wrapper of __import__10:31
sonne|workfoulwall: I don't understand why you need it though? Couldn't you just import everything statically?10:32
foulwallsonne|work: for the extensibility. when a new demo is written, there's no need to modify the, just put them in the demos' directory and they'll work.10:38
sonne|workfoulwall: ok I understand but for that you need to add checks that it doesn't import some random stuff10:39
sonne|workotherwise this introduces security issues10:39
foulwallit just import the modules with the prefix 'demos.'10:40
foulwallwhen import something that is not exist, it respond with a 40410:42
-!- iglesiasg [] has joined #shogun10:44
-!- mode/#shogun [+o iglesiasg] by ChanServ10:44
sonne|workok that should work10:44
foulwallWait I'll pull another request10:46
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:47
foulwallhide the funtion in the modules that is not permitted to invoke by user.10:47
sonne|workfoulwall: you basically have to make sure that no user input can harm the system10:49
sonne|workso try to avoid imports based on user import10:49
sonne|workfunction calls based on user input10:50
sonne|workalways only call hardcoded stuff when you see certain input10:50
sonne|workiglesiasg: please announce the workshop on twitter10:50
@iglesiasgsonne|work: ok10:52
* wiking is writing the unit tester for rocevaluation10:55
foulwallsonne|work: ok11:00
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun11:01
@iglesiasgsonne|work: done!11:01
@iglesiasgretweet it people!11:01
vgorbatihi guys, I am trying to write a unit test according to, but when I run 'make' in shogun/tests/unit it runs 40 tests, without my newly added unit test (some other tests are missing too). However, the obj file is generated. Can anyone help me?11:03
foulwalliglesiasg: retweeted and followed you11:03
wikingwtf: mathematics/ error: variable length array of non-POD element type 'SGSparseVector<float64_t>'11:04
wikingwho wrote this?!11:05
wikingand who merged it?!11:05
@iglesiasgfoulwall: nice, thank you!11:05
wikingok i know who wrote it....11:05
wikingheiko i will kill u11:07
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun11:08
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:09
wikingoh motherfucker11:12
wikingwhy is this not possible11:12
wikingto not to fucking commit broken code into the repo11:12
-!- gzhd79 [] has quit [Quit: gzhd79]11:12
@iglesiasgwiking: you ok man? :D11:12
wikingno because apparently whatever i try to make shogun work11:13
wikingit's fucking broken11:13
wikingok here's a very funny unit test that fails11:16
wikingoh even better11:17
wikingi have a segfault11:17
-!- shogun-notifier- [] has joined #shogun11:19
shogun-notifier-shogun: Viktor Gal :develop * ee8ab82 / tests/unit/evaluation/
shogun-notifier-shogun: Add unit test for ROCEvaluation11:19
shogun-notifier-shogun: as currently this little example fails11:19
wikingcan anybody tell me why this unittest fails?11:19
@iglesiasgwiking: what does it say when it crashes?11:20
wikingiglesiasg: [ERROR] assertion m_current_values.vector && idx<get_num_labels() failed in file labels/Labels.cpp line 5711:21
wikingunknown file: Failure11:21
wikingUnknown C++ exception thrown in the test body.11:21
wikingif i do the set_labels(vector)11:21
wikingit's even worse11:21
wikingit just segfaults11:21
@iglesiasgwiking: mm I don't see what is wrong either11:26
-!- abinash [75ef5e6b@gateway/web/freenode/ip.] has joined #shogun11:27
@iglesiasgwiking: I think it will work if you creat an SGVector<int32_t> with the labels and you give this to the CBinaryLabels constructor11:29
wikingiglesiasg: i've tried almost the same11:29
@iglesiasgbut what you are doing you work too idd11:29
wikinginstead of giving it to the constructor11:29
wikinguse label->set_labels(vector)11:29
wikingthat segfaults11:29
@iglesiasgyeah.. but try using it in the constructor11:29
@iglesiasgI bet that will work11:30
wikingbut what is the difference?!11:30
wikingi mean come on!11:30
@iglesiasgI am checking Labels code11:30
wikingthat *must* work11:30
@iglesiasgyeah, but I see something that might be weird11:30
wikingyes for sure11:30
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:30
@iglesiasgwiking: does it work fine using the SGVector in the constructor?11:31
sonne|workwiking: labels are copied with one11:31
sonne|work(in constructor)11:31
foulwallsonne|work: hi sonne.11:31
sonne|worknot with the set_labels()11:31
sonne|workfoulwall: hi!11:31
wikingsonne|work: should we remove set_labels?11:32
abinashHello I want to submit a code for the issue Probabilities for SVMs. What is the workflow for code submission11:32
sonne|workwiking: no11:32
sonne|workwiking: it should work11:32
foulwallsonne|work: Excuse me. and I've changed the demoHandler into a more security one.11:32
wikingctor version works11:32
@iglesiasgsonne|work, wiking : I think it might be something related to m_current_values and m_labels11:32
wikingsonne|work iglesiasg [ERROR] assertion m_current_values.vector && idx<get_num_labels() failed in file labels/Labels.cpp line 5711:33
wikingwith set_labels(vector)11:33
@iglesiasgset_labels and what wiking was using does not set m_current_values I think11:33
foulwallsonne|work: but it's a bit messy11:33
sonne|workfoulwall: btw please dont' do things like import * but just import what you need11:33
foulwallsonne|work: I've changed that.11:33
foulwallsonne|work: and i'll never do that.11:33
sonne|workiglesiasg: ahh I see11:33
@iglesiasgsonne|work: maybe it is a bug, I am not sure though11:34
wikingiglesiasg: well for sure it's a bug :)11:34
sonne|workwiking: no11:34
sonne|workwe have 2 functions11:34
sonne|workand set_values11:34
sonne|workthey don't set both11:34
wikingsonne|work: set_label(idx, value)11:34
sonne|workonly constructor does11:34
wikingsonne|work: using set_labels(vector) and set_label(idx, value)11:35
sonne|workthat's not nice though - but no idea (currenlty) how to do it better11:35
wikingwill not work with ROC11:35
wikingand this is a bug11:35
wikingas we let this happen11:35
sonne|workwiking: so do we ever want to be able to set these arrays *separately*11:36
sonne|workiglesiasg: could you please help abinash?11:36
wikingsonne|work: currently i have a usecase where yes i would like to11:36
sonne|workiglesiasg: thanks for twittering...11:36
@iglesiasgsonne|work: but what do m_labels and m_current_values represent? There must be a difference11:36
@iglesiasgsonne|work: no problem11:36
@iglesiasghey abinash11:36
@iglesiasgabinash: we have recently started using git flow in our development workflow11:37
sonne|workwiking, iglesiasg - yes sure there is a difference: consider e.g. binary labels. m_labels would then just be -1/+1 values11:37
wikingand what the hell is set_values anyways?11:37
-!- travis-ci [] has joined #shogun11:38
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []11:38
@iglesiasgabinash: see this mail on the mailing list for more info:
sonne|workwiking: the confidence values11:39
wikingsonne|work: and then set_labels ?11:39
@iglesiasgsonne|work: so m_labels would be the -1/+1 got it, but m_current_values?11:40
sonne|workiglesiasg: indeed no idea11:40
sonne|workI guess some subset stuff?11:40
@iglesiasgI thought so11:41
@iglesiasgsome of Heiko's magic for x-val11:41
wikingmmm we need to enable unit testing on travis11:41
@iglesiasgwiking, sonne|work: even if it is not a bug, I think it is a bit confusing11:42
sonne|workit is totally11:42
sonne|workI don't understand it11:42
@iglesiasgI wouldn't understand why the test didn't work without digging into Labels stuff11:42
wikingsonne|work: so now my answer for yesterday: shogun auROC: 0.900061 vs weka auroc: 0.95511:42
sonne|workwe have to ask hypergenious heiko to parse that I guess11:42
@iglesiasgmaybe we should rename so stuff11:42
@iglesiasgsonne|work: sure, let's see what he tells11:42
wikingsonne|work: so we are still off by a lot11:44
-!- abinash [75ef5e6b@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:45
sonne|workwiking: yeah so now tell us what you do11:46
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun11:46
wikingsonne|work: so now what's happening is that i use PolyKernel with 2.0 exponent11:46
wikingsonne|work: + normalize the input with p=2.011:47
wikingi.e. euclidean distance11:47
wikingand use libsvm with C=1.011:47
wikingbias enabled11:47
wikingepsilon set to 1E-1211:47
wikingsame settings in weka, using their smo11:48
wikingand of course run it on the same train/test set11:48
wikingi can right away try different kernels if u want me to, as i suspected that it should be better with other kernels11:50
wikingand using kernel normalization does not change the auroc11:50
sonne|workwiking: so with weka you use poly kernel11:52
wikingsonne|work: yes11:52
sonne|workin-homogeneous ?11:53
wikingsonne|work: the non-normalized polykernel11:53
sonne|workthen dont' normalize data in shogun either11:53
wikingyeah i've just tried both11:53
sonne|workby default the poly kernel does11:53
sonne|workset the identity kernel normalizer11:53
-!- HeikoS [] has joined #shogun11:55
wikingHeikoS: ARGH11:55
wikingHeikoS: get to work ! :)11:55
HeikoSwiking: ? :)11:55
wikingHeikoS: today you have a lot of task :)))11:55
wikingHeikoS: well first of all u merged code that breaks unit testing11:56
HeikoSjust arrived at werk :D11:56
HeikoSthats good, which ones?11:56
wikingHeikoS: second check the rocevaluator unittest11:56
HeikoSalready did11:56
wikingHeikoS: as i have nooooo clue why that fails11:56
HeikoSit fails?11:56
wikingHeikoS: yeps big time :)11:56
* sonne|work ~~~ooo~~~~~who~~let~~~the~~~dogs~~~out~~~ooo~~~~11:56
wikingsonne|work: with identitynormalizer: 0.91019411:57
HeikoSsonne|work: ha you are alright? :)11:57
HeikoSwiking: what dies it return?11:57
wikingHeikoS: well ASSERT error11:57
sonne|workHeikoS: wiking is biting you already :)11:57
wikingHeikoS: cd tests/unit && make11:57
wikingand you'll see11:57
sonne|workwiking: then don't normalize in shogun - like you do in weka11:57
wikingsonne|work: i mean i'm normalizing the input itself11:58
wikingsonne|work: i shouldn't?11:58
HeikoSwiking:  does the buildbot fail?11:58
wikingHeikoS: afaik buildbot doesn't run on develop branch11:58
HeikoSI see11:58
HeikoSoh thats not good11:58
HeikoSI am confused11:59
HeikoSthe buildbot runs on master?11:59
HeikoSand we commit to develop?11:59
HeikoSwhy is that?11:59
wikingHeikoS: yes11:59
wikingHeikoS: because travis should do tit :)11:59
wikingHeikoS: and it actually does most of it :P11:59
HeikoSI see11:59
sonne|workwiking: ohh man I guess you have to start from the beginning11:59
sonne|workwiking: you got data - you normalize than save that data11:59
HeikoSyes pls :)11:59
HeikoSwiking so when does code migrate from develop to master?12:00
sonne|workthen use exactly that data in weka and shogun?12:00
wikingsonne|work: L2Norm(input) -> PolyKernel(Exp=2.0) -> LIBSVM12:00
wikingHeikoS: release12:00
sonne|workwiking: nope12:00
wikingsonne|work: that's the current pipeline for the data12:00
wikingHeikoS: hahahah you grasped the most important part of the untittest :D12:01
HeikoSwiking: I see12:02
HeikoSwhat I dont really get is, why dont we run the buildbot on develop?12:02
wikingHeikoS: i mean with your comment... the problem is that currently that little snippet fails12:02
wikingHeikoS: hahahahahahahah12:02
wikingHeikoS: travis not enough?12:02
HeikoSwiking: so what does it return12:02
HeikoSwiking: not all architectures12:03
wikingHeikoS: [ERROR] assertion m_current_values.vector && idx<get_num_labels() failed in file labels/Labels.cpp line 5712:03
HeikoSso if we have errors that would not be detected there we will only notice when migrating to master12:03
wikingHeikoS: ok i guess then we have to discuss this with sonne|work12:03
HeikoSwiking: I see, so this seems to be a different problem12:03
HeikoSwiking: also this warning stuff, its so nice to see them in the buildbot12:03
HeikoSwiking:  wait12:04
HeikoSI actually saw commits by lambday compiling on the buildbot12:04
HeikoSwhy is that?12:04
wikingHeikoS: PR12:04
HeikoSdid he commit to master and I did not realise?12:04
wikingor no12:04
HeikoSexplain for dumb people please ;)12:04
wikingpr is not buildboted12:04
wikingi have no idea12:04
wikingHeikoS: explain me why that code fails12:04
wikingi'm really dumb as well in that case12:04
HeikoSwiking: ok will do12:04
HeikoSwiking: I notices some warnings in lambdays code in the build bot log12:05
HeikoSand then he fixed them and the warnings went away12:05
wikingbecause it was merged into master for some reason12:05
wikingwhich is NONO NOOOO good12:06
wikingbut i guess you did not realise12:06
wikingthat he sent the PR12:06
HeikoSno I did not12:06
wikingfor master branch12:06
wiking Merge pull request #961 from lambday/master ...12:06
wikingdba5cb777c Browse code12:06
wikingkarlnapf authored 15 days ago12:06
HeikoSI see12:06
wikingdont lie my friend, github catches u12:06
HeikoSso I have to make sure to merge into develop12:06
wikingwell just check the PR12:07
wikingwhether it's sent for master or develop branch12:07
HeikoSto know12:07
wikingif it's coming for master branch12:07
HeikoSand then travis catches things12:07
wikingjust drop it straight away12:07
wikingwe dont accept PR for master12:07
HeikoSwiking, and then when we release we merge develop into master?12:07
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun12:08
HeikoSwiking: okay! I knew this but forgot the check12:08
HeikoSgood that you tell me these things :)12:08
wikingHeikoS: indeed we do the merging12:08
wikingHeikoS: now it's going to be very interesting12:08
HeikoSwiking: so could you add something to travis that collects warnings?12:08
wikingHeikoS: becuase you got some stuff into master that is not in develop12:08
HeikoSalso for libshogun I would even vote to allow warnings as errors12:08
HeikoSas for python12:08
HeikoSbut you have to ask sonne|work and lisitsyn on this12:09
wikingHeikoS: so now we'll have to do some funky shit merging12:09
HeikoSwiking: shit, sorry for that12:09
HeikoSwell, we can rebase right?12:09
wikingHeikoS: heheh that's going to be a merge12:09
wikingand not a simple rebase12:09
wikingHeikoS: anyhow12:10
wikingHeikoS: why is that unit test fails12:10
wikingnow it's your turn12:10
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:10
HeikoSwiking: I will check12:10
HeikoSlisitsyn: could you have a thought on this out-of-synch master and develop? I fucked it up :(12:10
HeikoShow to restore that all in master is also in develop12:11
HeikoSwiking:  running tests12:11
-!- iglesiasg [] has quit [Ping timeout: 264 seconds]12:11
wikingHeikoS: cool12:11
HeikoSwiking:  so which tests did I make to fail?12:12
HeikoSsince only the ROC fail12:12
wikingHeikoS: no other fails12:12
wikingHeikoS: but that is already being taken care of ;)12:12
wikingHeikoS: dont worry about that12:12
wikingHeikoS: now it's ROC time12:12
wikingok? :)12:12
wikingsonne|work: another problem :)))12:13
-!- iglesiasg [] has joined #shogun12:13
wikingsonne|work: my input is DenseFeatures<float32_t>12:13
wikingsonne|work: and lets say i want to use Chi2Kernel12:13
wikingsonne|work: Chi2Kernel.cpp:6112:13
wikingfloat64_t* avec= ((CDenseFeatures<float64_t>*) lhs)->get_feature_vector(idx_a, alen, afree);12:13
wikingthis will lead to a segfault12:13
wikingi mean not this line12:13
wikingbut because of this12:14
wikinghow can we assure that no such case every can happen?12:14
wikingi.e. if the input is not CDenseFeatures<float64_t>12:14
wikingthen at least it fails with an assert12:14
wikingand not a segfault12:14
HeikoSwiking:  easy explanation12:14
HeikoSROC works on the scores12:15
wikingHeikoS: okok listening, what's the fix, is it already in?12:15
HeikoSnot the integer labels12:15
HeikoSI suggest the following:12:15
HeikoSadd a proper error message instead of the assert12:15
HeikoSthen, change the test a bit so that value vector is set12:15
HeikoSI will do that12:15
wikingHeikoS: ok just you know that this is confusing a lot12:15
HeikoSwiking:  the test is weird anyway12:16
wikingHeikoS: well this that we have set_label and set_value12:16
HeikoSROC on binary integers doesnt make any sense12:16
HeikoSset_value doenst work12:17
HeikoSif the value vector is not set12:17
wikingHeikoS: even if it doesn't make sense12:17
wikingwe should not die like this imo12:17
HeikoSwiking: true12:19
HeikoSwe should have some thoughts on the label vs values vs scores design12:19
HeikoSlisitsyn: ^12:19
wikingsonne|work: ok now i've beaten weka12:21
wikingsonne|work: as i suspected ;)12:21
wikingsonne|work: but i had the segfaults because of the prev error i was mentioning12:21
HeikoSwiking, sonne|work, lisitsyn we should have better error messages: I think it would be good to include the name of the method one is in class::name(): Error message12:22
HeikoSI never know where the asserts are coming from12:22
wikingHeikoS: lisitsyn sonne|work we should have a discussion about this: we have DenseFeatures<T> but then for example in some preprocessors and kernels have DenseFeatures<float64_t> hardcoded... this will cause segfaults... :(12:22
wikingsomehow it'd be great if we could do this part automatically12:22
wikinge.g. add a preprocessor to the DenseFeature and that will use the right 'version' (template paramter) of the preprocessor12:23
wikingany thoughts on this one?12:23
HeikoSwiking: the test might also fail if you have only labels from one class (might happen since random is used)12:26
HeikoSbetter to use odd/even for labels12:26
wikingHeikoS: heheh ok ok12:26
wikingHeikoS: i admit12:26
wikingnot the most sofisticated test12:26
wikingi just wanted to push something to show that something is wrong atm12:27
HeikoSwiking: I see, thats good12:28
-!- vgorbati__ [d4029f22@gateway/web/freenode/ip.] has joined #shogun12:29
@lisitsynvgorbati__: still having the same issue?12:30
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:30
@lisitsynwiking: yes that's why casts are let me cite you NOO ONOONON12:30
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun12:31
wikinglisitsyn: :D12:31
wikinglisitsyn: "you dont love me and i know NOW!" :D12:31
-!- abinash [75ef5e6b@gateway/web/freenode/ip.] has joined #shogun12:31
wikinglisitsyn: you mean that song? :P12:31
@lisitsynwiking: no, your "NONONO"12:32
-!- iglesiasg [] has quit [Read error: Connection reset by peer]12:32
wikinglisitsyn: 'fuckit!'12:32
-!- vgorbati__ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:33
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun12:34
abinashOk I got the workflow, but could someone explain how to submit code ?12:35
wikingabinash: PR12:36
wikingabinash: github pull request. make sure that it's against the develop branch12:36
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:37
-!- lambday [3ba2174d@gateway/web/freenode/ip.] has joined #shogun12:38
lambdayHeikoS: hi12:38
HeikoSlambday: hi!12:39
lambdayshould we always use SG_MALLOC while initializing SG-classes?12:39
HeikoSlambday: just a quick note: if you are doing PR's please always do them on the develop branch12:40
HeikoSlambday: the SGVector SGMatrix etc are treated differently12:40
HeikoSwe have automatic reference counting for them12:40
HeikoSyou *always* create them on the stack12:40
HeikoSrest is done internally12:40
HeikoSso always SGVector=SGVector(10)12:40
HeikoSnever SGVector*=new SGVector(10)12:40
@lisitsyndo not do SGVector a = SGVector(10);12:41
@lisitsyndo SGVector a(10);12:41
wikinghahah :)12:41
HeikoSlambday: you heard the man :D12:41
wikinglambday: now u know :)12:41
@lisitsynHeikoS: it creates an object two times12:41
HeikoSlisitsyn: yes you are in fact right12:42
HeikoSlisitsyn: could you help in rebasing the develop branch against master?12:42
@lisitsynHeikoS: alright let me check12:42
HeikoSwiking: how can I see the travis for my PR?12:42
HeikoSah found it!12:43
HeikoSlisitsyn: thanks12:44
-!- lambday [3ba2174d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:44
HeikoSlisitsyn: what do you think of warnings as errors for libshogun and python_modular on travis?12:44
-!- lambday_ [3ba2174d@gateway/web/freenode/ip.] has joined #shogun12:44
HeikoSwe have to remove those anyway12:44
lambday_sorry I got disconnected12:44
@lisitsynHeikoS: doc warnings?12:44
HeikoSin particular I like them for python since doxygen comments cannot be forgotten12:44
lambday_in one of my previous PR, I did something like SGSparseVector<float64_t> v[size-1];.... viktor and sonne|work pointed out that its not right12:44
@lisitsynI do not like that12:44
HeikoSlisitsyn: yes, and also libshogun gcc12:44
@lisitsynI'd rather get rid of all the dummy doc we have12:44
@lisitsynit is useless12:45
HeikoSlisitsyn: yes, do not merge dummy doc12:45
HeikoSbut the warning thing then helps you remembering to check12:45
HeikoSsince I tend to forget12:45
HeikoSand others do to12:45
@lisitsynHeikoS: yes but it is not an error12:45
HeikoSto get rid of dummy is responsibility of PR merger12:45
HeikoSlisitsyn: okay agreed12:45
HeikoSwhat about gcc warnings?12:45
HeikoSmy point is just: we tend to collect them12:46
@lisitsynHeikoS: I do not mind treating them as errors12:46
HeikoSsince its easy to oversee12:46
@lisitsynHeikoS: we should enable -Weffc++ :D :D12:46
HeikoSwhats that?12:46
@lisitsynwarnings from the effective C++ book by meyers12:46
HeikoSwhat do they tell you?12:47
@lisitsynI expect thousands in shogun12:47
@lisitsynvirtual destructor, operator=, blabla12:47
HeikoSI dont wont to be mean to people, just want to avoid piling up warnings12:47
HeikoSbut I guess one can also just check the warnings12:47
HeikoSwiking: could we collect them in travis and output at the end? or make it yellow as in the buildbot?12:48
wikingHeikoS: :))) google?12:48
@lisitsynHeikoS: so we want to rebase develop on top of last commit in master/12:48
HeikoSvgorbati: hi!12:48
-!- lambday_ [3ba2174d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:49
@lisitsynwiking: ^ ?12:49
HeikoSvgorbati: please use main channel for aksing questions, they might help others12:49
wikinglisitsyn: kindof12:49
HeikoSvgorbati: in order to make your unit test being executed, just use similar naming to the existing files12:49
HeikoSlisitsyn:  exactly12:49
@lisitsynI am curious if it would work without force push12:49
wikinglisitsyn: merge12:50
@lisitsynwiking: yeah merge is the only possible thing I see12:50
@lisitsynwait wait12:51
wikingi didnt touch anything12:51
@lisitsynwiking: but we have it already12:51
@lisitsynwiking: develop is on top of master d10e022a23a26b51e73a217c5014f3cf8d879ee612:52
@lisitsynthere is nothing to merge12:52
HeikoSsonne|work: could you explain me the problem with the sparse vector12:57
HeikoSwhy is SGSparseVector v[size-1]; wrong?12:57
@lisitsynHeikoS: []?12:57
HeikoSlisitsyn: ??12:57
@lisitsynSGSparseVector v[size-1];12:58
@lisitsynwhat does it mean?12:58
foulwallsonne|work: Hi there?12:58
HeikoSI see!12:58
HeikoSI guess he meant (size)12:58
HeikoSwith round brackets?12:58
HeikoSthats how I read it12:58
@lisitsynHeikoS: yes13:00
@lisitsynHeikoS: with [] it probably means create vector v and get its size-1 element13:01
@lisitsynI'd expect a segfault there13:01
HeikoSlisitsyn: yeah13:02
HeikoSI was confused13:02
HeikoSwhat am I saying13:03
HeikoSI am confused ;)13:03
-!- iglesiasg [] has joined #shogun13:04
-!- mode/#shogun [+o iglesiasg] by ChanServ13:04
HeikoSwiking: travis fails complaining about not being able to read tem pfiles13:04
wikingHeikoS: link13:04
HeikoScheck the tests on the bottom13:05
shogun-notifier-shogun: Heiko Strathmann :develop * 966dd6b / src/shogun/evaluation/ROCEvaluation.cpp:
shogun-notifier-shogun: replaced assert by proper error message13:05
shogun-notifier-shogun: Heiko Strathmann :develop * 65da3fd / src/shogun/labels/BinaryLabels.cpp:
shogun-notifier-shogun: added method name to error messages and re-formated13:05
shogun-notifier-shogun: Heiko Strathmann :develop * ce363f3 / src/shogun/labels/DenseLabels.cpp:
shogun-notifier-shogun: made set_cont to affect values/scores13:05
shogun-notifier-shogun: allocate value memory13:05
shogun-notifier-shogun: add values to parameters13:05
shogun-notifier-shogun: Heiko Strathmann :develop * a19166f / tests/unit/evaluation/
shogun-notifier-shogun: made unit-test work13:05
shogun-notifier-shogun: Heiko Strathmann :develop * 53378a5 / / (4 files):
shogun-notifier-shogun: Merge pull request #985 from karlnapf/develop13:05
shogun-notifier-shogun: Fixed ROC unit test13:05
HeikoSI still feel a bit uncomfortable that the buildbot doesn't run on develop ... like being in the dark. But should be fine with travis you are right13:06
@iglesiasgHeikoS, wiking, sonne|work : so it was bug in Labels at the end? :)13:07
wikingHeikoS: who creates those tmp files?13:07
HeikoSwiking: ehm, the tests?13:07
wikingHeikoS: no idea that's why i'm asking13:07
sonne|workwiking please take raw data for both13:08
sonne|workfoulwall: yes13:08
sonne|workHeikoS: SGSparseVector v[size-1] creates13:08
sonne|worksize-1 sparse vectors13:08
sonne|work(empty ones)13:08
sonne|workbut on the stack13:08
sonne|workvery dangerous thing - can make the system run out of stack13:09
foulwallsonne|work: What could I do next?13:09
sonne|workand with a dynamic parameter like size it is even a non-standard gnu extension13:09
@lisitsynsonne|work: bah I forgot about arrays, true13:10
sonne|workfoulwall: eye candy?13:11
sonne|workI mean some animation13:11
sonne|workgrid etc?13:11
foulwallsonne|work: ok13:12
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun13:12
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]13:14
-!- foulwall [] has left #shogun ["ERC Version 5.3 (IRC client for Emacs)"]13:18
-!- abinash [75ef5e6b@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]13:19
-!- travis-ci [] has joined #shogun13:24
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []13:24
-!- foulwall [] has joined #shogun13:26
-!- Marty28 [~marty@] has joined #shogun14:06
-!- foulwall_ [] has joined #shogun14:13
-!- foulwall [] has quit [Read error: Connection reset by peer]14:13
-!- Marty28 [~marty@] has quit [Quit: Colloquy for iPad -]14:15
HeikoSwiking: could you have a look at one of the travis ones? This is happeneing all the time14:17
shogun-notifier-shogun: lambday :develop * a28c730 / tests/unit/mathematics/
shogun-notifier-shogun: replaced variable-length array with SG_MALLOC in log_det_test_314:18
shogun-notifier-shogun: Heiko Strathmann :develop * a343111 / tests/unit/mathematics/
shogun-notifier-shogun: Merge pull request #986 from lambday/develop14:18
shogun-notifier-shogun: replaced variable-length array with SG_MALLOC in log_det_test_314:18
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]14:20
wikingHeikoS: the /tmp stuff?14:26
HeikoSwiking yes14:26
wikingHeikoS: no idea man14:27
wikingHeikoS: i'll try to google it but...14:27
wikingi'm not a travis expert ;)14:27
HeikoSwiking: ok14:28
HeikoSthen lets ask somebody else :)14:28
HeikoSwhy does it only fail on this particular build?14:28
shogun-notifier-shogun: Heiko Strathmann :develop * ff28dd0 / examples/undocumented/libshogun/labels_binary_fit_sigmoid.cpp:
shogun-notifier-shogun: removed assertions and print probabilities14:32
shogun-notifier-shogun: Heiko Strathmann :develop * 7f31c48 / tests/unit/labels/
shogun-notifier-shogun: added unit test for sigmoid fitting14:32
shogun-notifier-shogun: Heiko Strathmann :develop * d30f205 / / (2 files):
shogun-notifier-shogun: Merge pull request #988 from karlnapf/develop14:32
shogun-notifier-shogun: unit test for binary label sigmoid fitting14:32
-!- rishabh [~rishabh@] has joined #shogun14:35
-!- travis-ci [] has joined #shogun14:37
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:37
@lisitsynargh come on will they finally catch this asshole?14:49
@lisitsynHeikoS: watertown, usa14:52
@iglesiasglisitsyn: oh so the police killed the guy14:53
@lisitsyniglesiasg: yes one moron killed14:53
HeikoSI see14:54
@lisitsyniglesiasg: as I tweeted - these nations are troublesome for anyone - so usa should think of sheltering such animals14:54
@lisitsynthey usually do the same here14:54
@iglesiasglisitsyn: to what do you refer to with these nations?14:54
@iglesiasgwhere these guys that did it come from?14:54
@lisitsyniglesiasg: chechens, dagestanians14:55
@iglesiasgI see14:55
@lisitsyniglesiasg: they came from dagestan but they are chechens14:55
-!- travis-ci [] has joined #shogun14:55
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:55
@lisitsyniglesiasg: well it is adjacent regions14:55
@iglesiasgI see14:56
-!- foulwall_ [] has quit [Read error: Connection reset by peer]15:07
-!- foulwall [] has joined #shogun15:08
foulwallsonne|work: some eye candy15:09
@iglesiasgfoulwall: you are an artist! :)15:10
sonne|workfoulwall: nice. it could move a bit faster15:10
foulwalllisitsyn: sonne|work thankyou15:10
-foulwall:#shogun- I'll add some animation on 'clear'15:11
foulwallsonne|work: I'm wondering what kind of animation can be added to the cluster demo.15:12
sonne|workfoulwall: yeah would be cool15:12
sonne|workfoulwall: make the diameter of the circle from 0 -> r15:12
@iglesiasgfoulwall: if you are using k-means, you can do a nice one15:12
@iglesiasgfoulwall: since this method is iterative15:13
sonne|workiglesiasg: he only has the end result15:13
sonne|workit's the shogun output15:13
@iglesiasgfoulwall: mmm however that would require intermediate results15:13
@iglesiasgsonne|work: idd I noted that15:13
foulwallsonne|work: ok15:13
@iglesiasgsonne|work: I think it is worth trying15:13
sonne|workiglesiasg: the same would work for an svm15:13
sonne|workor svr15:13
sonne|workthink of it just training one iteration15:14
@iglesiasgaham true15:14
@iglesiasgI didn't come up with it for the svm and svr15:14
HeikoSfoulwall, could you connect the lines between the points in the regression function?15:15
HeikoSwhen you make the kernel width very small you see the individual points15:16
foulwallHeikoS: I note that, wait me and I'll fix that15:16
HeikoSfoulwall, this is really cool!15:17
foulwallHeikoS: thank you15:17
HeikoSfoulwall: you could do this for the GP classifier and regression15:17
HeikoSfoulwall, you know what would be even cooler?15:17
HeikoSif you could plot the underlying distribution15:17
HeikoSlike a heatmap15:17
HeikoSat least for classifier15:17
foulwallHeikoS: that's what I want to do next15:17
HeikoSI am amazed :D15:17
HeikoSkeep on the good work!15:18
sonne|workHeikoS, foulwall that is sth daniel has done already15:18
HeikoSsonne|work, ok, I didnt see that15:18
sonne|workit is also pretty cool15:18
foulwallsonne|work: I'll go for class. later i'll fix them.15:18
sonne|workfoulwall: thanks for your work15:19
HeikoSsonne|work seeing, wow that also very cool15:20
thoralfsonne|work: about text features again: i think it still makes sense to keep the preprocessing part outside shogun, but:15:25
-!- foulwall [] has quit [Remote host closed the connection]15:25
thoralfallowing files "feat1 feat2 ..." would be the way15:25
thoralfn-grams etc. could be created outside like feat1_feat2_... feat2_feat3_... etc.15:26
thoralfAll you have to to is to create TextFeature classes - and eventually something that maps them into a feature space.15:27
-!- sonne|work [] has quit [Quit: Leaving.]15:28
-!- Asix3 [452f9490@gateway/web/freenode/ip.] has joined #shogun15:32
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun15:37
lambdayHeikoS: hi15:37
-!- rishabh [~rishabh@] has quit [Ping timeout: 264 seconds]15:38
lambdayHeikoS: I added the sample_from_gaussian for sparse cov matrices and a unit test..15:38
lambdayHeikoS: shall I send the PR and you could take a look?15:38
lambdayHeikoS: also changed the array of pointers that in the log_det unit test to double pointer and used SG_MALLOC instead15:39
HeikoSlambday: I saw that, nice! already merged15:41
HeikoSsend the PR!15:42
HeikoSlooking forward to see it :)15:42
lambdayHeikoS: okay sending :)15:42
lambdayHeikoS: as of now, I used a constant diagonal matrix for sparse unit-test.. in the next unit test will add a random one15:44
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun15:46
-!- Asix3_ [452f9490@gateway/web/freenode/ip.] has joined #shogun15:52
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]15:53
-!- Asix3 [452f9490@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]15:53
-!- trtr3434 [] has joined #shogun16:03
-!- trtr3434 [] has left #shogun []16:03
-!- iglesiasg [] has quit [Quit: leaving]16:08
wikingHeikoS: mathematics/Statistics.cpp: In constructor 'shogun::CStatistics::log_det(shogun::SGSparseMatrix<double>)::EigenSimplicialLLT::EigenSimplicialLLT()':16:14
wikingmathematics/Statistics.cpp:2030:12: error: 'SimplicialCholeskyLLt' was not declared in this scope16:14
wikingmake[1]:  [mathematics/Statistics.cpp.o] Error 116:14
wikingmake[1]: Leaving directory `/home/niki/Dropbox/work/workspace/shogun/src/shogun'16:14
wikingfails on ubuntu16:14
wikingmake: *** [] Error 216:14
HeikoSoh, then we should add this PR recently16:14
HeikoSand a buildbot on develop :)16:14
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:14
HeikoSwiking: gsomix sent this PR recently to solve this problem16:15
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun16:15
lambdayHeikoS: I think I made a mistake.. checking again16:15
-!- Asix3_ [452f9490@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:16
wikingHeikoS: this is *master* branch16:18
HeikoSwiking: argh16:18
wikingso there should be definitely done something about this16:18
HeikoSmy fault again I guess :(16:18
-!- rishabh [~rishabh@] has joined #shogun16:18
HeikoSwiking, link16:18
HeikoSwiking: good chance to try the hotfix branch ;)16:18
wikingit is a good chance for that16:20
HeikoSlambday: no worries, not merged yet16:20
HeikoSwiking: have we already rebased cev against master?16:20
HeikoSsince I suggest to do this before16:21
HeikoSit might get messy otherwise16:21
-!- juanmanpr [c2fe7d88@gateway/web/freenode/ip.] has joined #shogun16:21
juanmanprhello guys, about gsoc, what kine of profiles are you searching on the students?16:22
juanmanprI am studying master in computer vision, and im really interested in machine learning16:23
juanmanpri have experience in C, C++, Matlab, and a little python16:23
HeikoSjuanmanpr: we are looking for motivated people than can code and have some background in the topic their project is about16:24
HeikoSjuanmanpr: the best way to get involved is to start sending patches and discuss project ideas16:24
HeikoSwith the mentors16:24
juanmanprBut, is there any special requirement that you are looking, for example to Interface and develop the general quadratic solver library libqp16:24
juanmanprok, where can I finf a list of menthors for shogun?16:25
HeikoSon our ideas page16:25
juanmanprthanks, i'll take a look16:25
HeikoSthere is no "hard" requirement, you just have to convince us of you :)16:25
HeikoSas a computer vision student, you should have the right background16:25
juanmanprYou also would like to see some code, right?16:26
HeikoSso just show us that you can code and are motivated by sending patches :) see our issues list on github for initial tasks you could do16:26
HeikoSyes, it is a requirement for your application to be considered16:26
HeikoSnot only show16:26
HeikoSbut actually merged to shogun codebase16:26
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:27
juanmanprI see, well, I better do something now!16:27
HeikoSjuanmanpr: looking forward to see a pull request!16:28
HeikoSI gotta go now, bye!16:28
-!- juanmanpr [c2fe7d88@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:32
-!- dvalcarce [] has joined #shogun16:37
-!- rishabh [~rishabh@] has quit [Ping timeout: 264 seconds]16:40
dvalcarcehi all16:40
wikinganybody seen this error before: /usr/local/include/shogun/lib/Signal.h:102:14: error: expected unqualified-id before 'protected'16:51
wikingok we have a problem16:59
wikingi mean it's not our problem16:59
wikingbut it's qt's fault17:00
wikingbasically qmake has a #define signals17:00
wiking-> we get that error with shogun if u use a code in qmake that uses shogun (headers)17:00
-!- shogun-notifier- [] has quit [Quit: transmission timeout]17:32
-!- foulwall [~foulwall@2001:da8:215:c252:edf7:a7fe:6f7f:ef7f] has joined #shogun17:40
foulwallHi, dvalcarce17:40
dvalcarcefoulwall: hi!17:41
foulwalldvalcarce: I am also applying for GSoC the interactive demos.17:41
foulwalldvalcarce: we can work together and make better demos.17:41
dvalcarceyeah, that's a good idea17:42
dvalcarcei was thinking the same thing17:42
foulwalldvalcarce: :-)17:42
dvalcarcei see your regression demo, it's cool17:43
dvalcarcemaybe we can split server and client sides work17:43
dvalcarcei don't know what sonney2k will think about it17:43
foulwalldvalcarce: good idea.17:44
foulwalldvalcarce: and dont' know either.17:45
dvalcarcelet's ask him then :P17:46
dvalcarcewhich part do you prefer?17:46
dvalcarcebecause right now i think we are doing duplicated work on django side17:47
dvalcarcei think we should define some basic interfaces17:48
foulwallboth are ok, and the front end needs more creativity.17:49
dvalcarcei agree, bootstrap is beautiful but it isn't enough17:50
foulwallI've considered building a interface, but for the security reasons...17:50
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun17:51
dvalcarcei understand17:52
dvalcarceit would be convenient to have a list of demos so we can plan the website, too17:52
thoralfBye.  See you monday.17:53
-!- thoralf [] has left #shogun ["Konversation terminated!"]17:53
foulwallsonney2k told me it's better to limit the access.17:53
vgorbatilisitsyn, hi are you there?17:55
dvalcarcewe need server communication, but we don't want a ddos17:56
foulwalldvalcarce: yes.17:57
dvalcarcesome kind of authentication?17:57
dvalcarceor maybe it is too overhead because of some demos?17:57
foulwalldvalcarce: auth is easy to implement but we need somekind of elastic way to make the demo code extensibility17:59
foulwalldvalcarce: code reuse and general interface17:59
foulwallbut a general interface may cause some function exposed to the bad user.18:02
dvalcarcewe can implement data size or computing time limitations18:04
-!- rishabh [~rishabh@] has joined #shogun18:11
-!- hoijui [] has quit [Ping timeout: 246 seconds]18:12
-!- se7en7 [~babu@2001:6b0:1:1430:21e:4fff:fec0:f4e2] has quit [Remote host closed the connection]18:13
-!- vikram360 [~vikram360@] has joined #shogun18:14
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun18:14
lambdayHeikoS: hi.. there?18:14
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]18:16
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun18:29
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit []18:33
-!- foulwall [~foulwall@2001:da8:215:c252:edf7:a7fe:6f7f:ef7f] has quit [Ping timeout: 245 seconds]18:41
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]18:47
-!- dvalcarce [] has quit [Quit: dvalcarce]18:54
-!- dvalcarce [] has joined #shogun18:54
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun18:57
-!- rishabh [~rishabh@] has quit [Ping timeout: 256 seconds]19:01
-!- foulwall [] has joined #shogun19:03
-!- foulwall [] has quit [Ping timeout: 258 seconds]19:16
-!- rishabh [~rishabh@] has joined #shogun19:19
-!- HeikoS [] has quit [Quit: Leaving.]19:25
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun19:26
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:30
-!- foulwall [] has joined #shogun19:37
foulwalllisitsyn: dvalcarce sonney2k: I've add another eyecandy for the clustering / svr demo.19:39
foulwallyou can see that on
-!- hoijui [] has joined #shogun19:40
foulwallit's late in beijing, i'll go sleeping. goodnight.19:40
dvalcarcenice :)19:41
dvalcarcegood night19:41
@lisitsynfoulwall: and nice19:42
foulwallHeikoS: I've connected the dots into line so svr demo can work correctly when the width is small.19:43
-!- foulwall [] has quit [Ping timeout: 258 seconds]19:52
-!- rishabh [~rishabh@] has quit [Ping timeout: 256 seconds]20:02
-!- rishabh [~rishabh@] has joined #shogun20:13
-!- dvalcarce [] has quit [Quit: dvalcarce]20:25
-!- sidbee [~sidbee@] has joined #shogun20:40
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun20:45
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:46
-!- hoijui [] has quit [Quit: Leaving]20:48
@lisitsynvgorbati: what you may do now is to replace this 100% check with something weaker20:54
-!- sidbee [~sidbee@] has quit []20:55
vgorbatilisitsyn: experimentally find some percentage, that will pass?20:55
@lisitsynvgorbati: kind of20:55
@lisitsynvgorbati: you may check the other thing though20:56
vgorbatilisitsyn: ?20:56
@lisitsynwhether weights are preserved20:56
@lisitsynlike for any vector x in initial space20:56
@lisitsynthe linear combination weights are the same20:57
@lisitsynfor initial space and embedded space20:57
vgorbatiwell, that will complicate things a little bit)20:58
vgorbatiit's like doing the big part of LLE itself20:59
vgorbatilisitsyn: I guess I can't receive the neighbors weights from CLocallyLinearEmbedding somehow, can I/21:01
-!- n4nd0 [] has joined #shogun21:14
-!- n4nd0 is now known as iglesiasg21:15
@lisitsynvgorbati: yes I don't like that way too21:17
-!- zxtx [] has quit [Ping timeout: 252 seconds]21:21
vgorbatilisitsyn: I think I will implement the "weakening" part on Monday - is it ok?21:23
@lisitsynvgorbati: totally21:26
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has joined #shogun21:30
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]21:32
-!- zxtx [] has joined #shogun21:36
-!- qpalzm [~ck@] has joined #shogun21:45
-!- lisitsyn [] has quit [Quit: Leaving.]21:51
-!- lisitsyn [] has joined #shogun21:52
-!- mode/#shogun [+o lisitsyn] by ChanServ21:52
-!- vgorbati_ [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]21:54
-!- vikram360 [~vikram360@] has quit [Ping timeout: 248 seconds]22:04
-!- gsomix [~Miranda@] has joined #shogun22:10
gsomixgood evening22:10
iglesiasghello gsomix22:10
iglesiasghow is gsoc preparation going?22:11
gsomixiglesiasg: bit by bit.22:15
@lisitsynthe dream city22:35
gsomixlisitsyn: I have one more
@lisitsynanyone watching for the 18th hour of that madness?22:44
-!- mdragan [6d66bb92@gateway/web/freenode/ip.] has joined #shogun22:48
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]22:52
-!- iglesiasg [] has quit [Ping timeout: 264 seconds]22:54
@lisitsynsuspect in custody23:17
-!- iglesiasg [] has joined #shogun23:40
-!- mode/#shogun [+o iglesiasg] by ChanServ23:40
@lisitsyniglesiasg: join me listening to scanner :D23:41
@lisitsyniglesiasg: I admit I am feeling bad already have been listening to it a hour or so23:41
@iglesiasglisitsyn: let me check23:43
@iglesiasgI have never listened to it23:43
@iglesiasglisitsyn: btw, have you seen?
@iglesiasgnice code I think23:44
@lisitsyniglesiasg: yes sure23:44
@iglesiasglisitsyn: what is this man?23:45
@iglesiasgis it like the police radio23:45
@lisitsyniglesiasg: yes23:45
@iglesiasgthat wouldn't make sense23:45
@lisitsynwhy not?23:45
@iglesiasghow is it possible we can listen to it?23:45
@iglesiasgshouldn't it be secret? hehe23:46
-!- kaushalb09 [] has joined #shogun23:46
@lisitsyniglesiasg: boston PD asked to not share addresses/etc but you can't control the internet you know23:46
@lisitsyniglesiasg: they are deadly tired23:47
@lisitsyn19th or 20th hour already23:47
@iglesiasgsomeone must be helping him IMHO23:47
@lisitsyniglesiasg: helping like?23:49
@lisitsyn2 guys23:50
@iglesiasglisitsyn: helping like keeping him at a place23:50
@lisitsynah they are going to fail this time23:50
@iglesiasgI bet he is at somenone's place23:50
@lisitsynyes may be23:50
@iglesiasgjust gut feeling23:50
@lisitsynthey have scanned basically all the city already I guess23:51
@lisitsyniglesiasg: they are going to have a trouble with sunsetting23:53
@iglesiasglisitsyn: why?23:54
@lisitsyniglesiasg: if they haven't found it with daylight..23:54
@lisitsyniglesiasg: 6 streets left23:56
@iglesiasglisitsyn: are they actually going house by house?23:58
@iglesiasgchecking inside, I guess not23:58
@lisitsyniglesiasg: yes23:58
@lisitsyniglesiasg: not the detailed inside I guess by yes - door-by-door23:58
@iglesiasglisitsyn: but this radio thing is fucked up23:59
@iglesiasgI mean I don't understand23:59
--- Log closed Sat Apr 20 00:00:28 2013