Open in new window / Try shogun cloud
--- Log opened Sun Mar 29 00:00:19 2015
-!- curiousguy13 [~curiousgu@] has quit [Ping timeout: 255 seconds]04:36
-!- curiousguy13 [~curiousgu@] has joined #shogun04:49
-!- pickle27 [] has joined #shogun04:58
-!- pickle27 [] has quit [Remote host closed the connection]07:16
-!- x4i [~A2o@unaffiliated/x4i] has joined #shogun10:15
-!- besser82 [] has joined #shogun10:22
-!- besser82 [] has quit [Changing host]10:22
-!- besser82 [~besser82@fedora/besser82] has joined #shogun10:22
-!- mode/#shogun [+o besser82] by ChanServ10:22
-!- curiousguy13 [~curiousgu@] has quit [Ping timeout: 248 seconds]10:36
-!- x4i [~A2o@unaffiliated/x4i] has quit [Quit: irc]10:44
-!- curiousguy13 [~curiousgu@] has joined #shogun10:49
-!- nszceta [] has joined #shogun12:52
-!- nszceta [] has quit [Quit: Textual IRC Client:]13:24
-!- HeikoS [] has joined #shogun13:29
-!- mode/#shogun [+o HeikoS] by ChanServ13:29
-!- vortex_ape [~vortex_ap@] has joined #shogun13:51
-!- nileshK [uid75191@gateway/web/] has quit [Quit: Connection closed for inactivity]14:09
@HeikoSlisitsyn: heya14:25
-!- shogun-notifier- [] has joined #shogun14:26
shogun-notifier-shogun: Wu Lin :develop * 7c758d4 / src/shogun/ (10 files):
shogun-notifier-shogun: minor fix14:26
shogun-notifier-shogun: Heiko Strathmann :develop * 768671c / src/shogun/ (10 files):
shogun-notifier-shogun: Merge pull request #2785 from yorkerlin/develop14:26
shogun-notifier-shogun: minor fix14:26
-!- travis-ci [] has joined #shogun14:28
travis-ciit'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 []14:28
shogun-buildbotbuild #259 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at  blamelist: Heiko Strathmann <>, Wu Lin <>14:31
-!- nileshK [uid75191@gateway/web/] has joined #shogun14:35
nileshKHeikoS: Hi14:35
@HeikoSnileshK: hi14:35
@HeikoSnileshK: how are things with the lpboost?14:35
nileshKYeah. I trying to get something working on lpboost14:36
nileshKSince now I have got it compiling, I am trying to figure out how to create a toy example14:36
nileshKwhich will just act a proof of concept and add it to the examples set14:37
@HeikoSnileshK: the error you got seems to be of a different source14:37
nileshK"different source"? I dont think I understand this14:38
nileshKAlso, I needed your help on one more thing14:40
nileshKLook at this: ui/GUIClassifier.cpp. This thing tries to use the lpboost14:40
shogun-notifier-shogun: Sergey Lisitsyn :develop * 4e5f194 / / (23 files):
shogun-notifier-shogun: Moved hashed features into separate directory14:41
shogun-notifier-shogun: Sergey Lisitsyn :develop * ecc6f4d / src/shogun/ (4 files):
shogun-notifier-shogun: Fixed some broken include paths14:41
shogun-notifier-shogun: Sergey Lisitsyn :develop * aaaaaf3 / tests/unit/ (5 files):
shogun-notifier-shogun: Adjusted paths to hashed features in unit tests14:41
shogun-notifier-shogun: Sergey Lisitsyn :develop * 060bb0f / / (32 files):
shogun-notifier-shogun: Merge pull request #2739 from lisitsyn/refactor/hashed_to_separate_dir14:42
shogun-notifier-shogun: Moved hashed features into separate directory14:42
shogun-buildbotbuild #3375 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>14:43
shogun-buildbotbuild #260 of trusty - libshogun - viennacl is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>14:43
-!- vortex_ape [~vortex_ap@] has quit [Ping timeout: 252 seconds]14:46
-!- vortex_ape [~vortex_ap@] has joined #shogun14:46
shogun-buildbotbuild #261 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>14:47
shogun-notifier-shogun: Khaled Nasr :develop * 88269cd / examples/undocumented/libshogun/ (4 files):
shogun-notifier-shogun: fixed uncought exception problem in neuralnets c++ examples14:50
shogun-notifier-shogun: Sergey Lisitsyn :develop * 529ab43 / examples/undocumented/libshogun/ (4 files):
shogun-notifier-shogun: Merge pull request #2535 from khalednasr/bugfix14:50
shogun-notifier-shogun: fixed uncought exception problem in neuralnets c++ examples14:50
shogun-buildbotbuild #3376 of deb1 - libshogun is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #262 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at  blamelist: Khaled Nasr <>, Sergey Lisitsyn <>14:53
-!- travis-ci [] has joined #shogun14:55
travis-ciit's Sergey Lisitsyn'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
-!- HeikoS [] has quit [Quit: Leaving.]15:02
-!- HeikoS [] has joined #shogun15:08
-!- mode/#shogun [+o HeikoS] by ChanServ15:08
-!- HeikoS [] has quit [Quit: Leaving.]15:14
-!- travis-ci [] has joined #shogun15:17
travis-ciit's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []15:17
-!- curiousguy13 [~curiousgu@] has quit [Ping timeout: 255 seconds]16:37
-!- curiousguy13 [~curiousgu@] has joined #shogun16:49
nileshKHello anybody there16:59
@lisitsynnileshK: yes17:00
nileshKI was a bit doubtful regarding the implementation of LpBoost that Shogun has17:00
nileshKDoes it really work?17:00
nileshKAnd has it been tested?17:00
@lisitsynnileshK: ohh its pretty old17:01
@lisitsynI think it was tested somehow17:01
@lisitsynnileshK: sonney2k_ is the man to ask actually17:02
nileshKI need some help in getting an example of it working.17:02
nileshKCan you help me right now?17:02
nileshKHe doesnt seem to be be online now.17:02
@lisitsynI can try17:03
nileshKSo, see LpBoost is inherited from a class CLinearMachine17:03
nileshKBut it doesnot have a constructed implemented which can accept the data17:04
@lisitsynwhat data?17:04
nileshKas in for that matter even SVMOcas is also inherited from CLinearMachine17:04
@lisitsynyes they are all linear17:04
nileshKdata => training set17:05
nileshKCSVMOcas* lin_svm = new CSVMOcas(C, rf_feats, labels);17:05
@lisitsynthat's ok17:06
@lisitsynjust do set_features17:06
nileshKAlso regarding doing the set_features thing17:06
nileshKis there a specfic format in which the training data has to be presented17:07
nileshKJust to clarify my point what kind of data files can I use with LpBoost17:07
nileshKfrom the shogun-data that I downloaded17:08
@lisitsynas far as I can see LpBoost requires sparse features17:08
nileshKYeah. so how do I proceed17:08
@lisitsynwhat language are you using shogun in?17:09
nileshKI am using C++17:09
@lisitsynok I see17:10
@lisitsynyou need to construct CSparseFeatures
@lisitsynyou can create sgsparsematrix or sgmatrix17:11
@lisitsynor load it from file17:11
nileshKok. Assuming I created the CSparseFeatures17:12
nileshKThe next step would be to the train function?17:12
@lisitsynyeah you just create CLpBoost thing17:12
@lisitsyncall set_features(<your features>)17:12
@lisitsyncall set_labels(<your labels>)17:12
@lisitsynand train17:13
nileshKI will try this and get back to you17:14
nileshKTill what time will you be online?17:14
@lisitsynnileshK: whole day but can disappear for some moments :)17:14
nileshKOk. I might require your help for small issues I might stuck into. :)17:15
nileshKThanks a lot!17:15
nileshKI will ping you after trying this17:15
@lisitsynnileshK: sure17:17
-!- shogun-notifier- [] has quit [Quit: transmission timeout]17:50
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 265 seconds]18:37
nileshKlisitsyn: Hi19:23
@lisitsynnileshK: hey19:24
nileshKI was checking this file out:
nileshKIt had an example of reading sparefeatures19:24
@lisitsynah yeah19:24
@lisitsynit uses libsvm19:25
nileshKbut can you point me to a dataset that could be read via that function19:25
nileshKPlease check the load_data function19:25
@lisitsynjust a second19:27
@lisitsynthese should work19:30
nileshKOk. I will try and get back. Thanks :)19:31
nileshKLisitsyn: The dataset worked correctly.20:24
nileshKI was trying to modify the code to generate an example for lpBoost on lines with this code20:24
nileshKThis code I have adapted from teh random_fourier_classfication20:25
nileshKIt is compiling well for me20:25
nileshKg++ lpBoost_classfication.cpp -lshogun -L/home/nilesh/shogun-install/lib -I/home/nilesh/shogun-install/include20:25
nileshKI am using this data set20:26
nileshKBut on executing with this command :20:26
nileshK./a.out --dataset ../../data/a1a20:26
@lisitsynexport LD_LIBRARY_PATH=/home/nilesh/shogun-install/lib20:27
@lisitsynthen try that again20:27
@lisitsynnileshK: cool :) hope lpboost works as well20:31
nileshKYeah I am now adding the testing part to it.20:31
nileshKLet me see how that goes through20:31
-!- pickle27 [] has joined #shogun20:43
-!- vortex_ape [~vortex_ap@] has quit [Ping timeout: 256 seconds]20:46
-!- vortex_ape [~vortex_ap@] has joined #shogun20:47
-!- HeikoS [] has joined #shogun21:04
-!- mode/#shogun [+o HeikoS] by ChanServ21:04
@HeikoSlisitsyn: around?21:07
@lisitsynHeikoS: yes21:08
@HeikoSlisitsyn: hi how are things?21:08
@HeikoSfeeling better?21:08
@lisitsynHeikoS: yeah much better :)21:08
@lisitsynwhat about you?21:08
@HeikoSyeah same21:09
@HeikoSparents are visiting in London21:09
@HeikoSthats fun21:09
@HeikoShey I wanted to ask aobut the manual stuff21:09
@HeikoSwant to keep that going21:09
@HeikoSwhats needed?21:09
@lisitsynwell you know I didn't approach that since then21:09
@HeikoSI didnt know21:10
@lisitsynhmm well21:10
@HeikoSshall we try to tackle the html thingi now?21:10
@lisitsynI am not sure I can do that now but I can try tomorrow21:10
@lisitsynthe thing is that we would need our own 'visit' thing21:11
@lisitsynthat prints node21:11
@lisitsynbecause we mess with bootstrap and classes are not enough21:11
@HeikoSshouldn not be too hard should it?21:11
@lisitsynshould be easy21:11
@lisitsynHeikoS: if we go with better naming of functions it would look really amazing21:12
@lisitsynpeople will love it21:12
@lisitsynget_something() is non-sense in java21:12
@HeikoSlisitsyn: yeah I like tha tone21:13
@HeikoSlisitsyn: I think we should combine it with the manual thing21:13
@HeikoSintegrate it straight away to get these things out together21:13
@HeikoSone *big* downside though is that the API docs dont work anymore21:13
@HeikoScurrently, syntax is the same21:14
@HeikoSand that is good21:14
@HeikoSso not sure about details there21:14
@lisitsynshould be solvable21:14
@lisitsynI need to force myself to work on doc/distribution of shogun21:14
@lisitsyninstead of neural nets :D21:14
@lisitsynsomething made me to refactor some code there21:15
@HeikoSlisitsyn: I mean as far as I know we cannot generate API for different languages21:15
@HeikoSbut if we had the examples, then this is fine as it shows how to use the API21:15
@HeikoSlisitsyn: do you ahve an idea how to define this visit thing?21:15
@lisitsynHeikoS: I have seen some example21:15
@lisitsynactually in they define visit_* thing21:16
@lisitsynbut I've seen some other example21:16
@lisitsynhow to add custom stuff there21:16
@lisitsynHeikoS: that's it21:19
@HeikoSok let me check21:21
@HeikoSlisitsyn: so its just self.body.append21:21
@lisitsynlooks like21:22
@lisitsynlets try21:22
@HeikoSIll just pase his code in ours and see what happens21:22
@lisitsynHeikoS: I just tried to self.body.append21:25
@HeikoSlisitsyn: doesnt seem like its doing something21:25
@HeikoSlisitsyn: but he does someting as self.body.append and then self.context.append21:26
@lisitsynhah stop I was doing some useless stuff21:27
@HeikoSlisitsyn: ?21:28
@HeikoSlisitsyn: the code is actually not executed21:28
@HeikoSI mean the visit code21:28
@lisitsynyeah I tried to modfiy visit_sgexample21:28
@lisitsynwhich is not used anyway21:29
@HeikoSlisitsyn: mmh21:29
@HeikoSwhy is that?21:29
@HeikoSwas the orginal example different?21:30
@lisitsynHeikoS: we use ShogunExample21:30
@lisitsynbut not sgexample21:30
@HeikoSlisitsyn: its this phase 4 they talk about21:30
@HeikoSdoesnt seem to change anything?21:31
@HeikoScheck description of phase 421:32
@lisitsynI managed to add a tag21:32
@HeikoSThis phase converts the resolved doctrees to the desired output format, such as HTML or LaTeX. This happens via a so-called docutils writer that visits the individual nodes of each doctree and produces some output in the process.21:32
@lisitsynwe are close21:32
@lisitsynHeikoS: ok I can commit now21:35
@HeikoSok go21:35
@lisitsynbut it doesn't work21:35
@HeikoSwhat did you do then?21:35
@lisitsynHeikoS: I added divs with proper role21:36
@lisitsynaround the code21:36
@HeikoSwhy doesnt it work?21:36
@lisitsynsome bootstrap stuff21:36
@lisitsynit breaks due to some strange layout I guess21:36
@lisitsynor maybe no ids21:36
@lisitsynlet me check21:37
@lisitsynHeikoS: we have multiple tab panel21:38
@HeikoSill wait for the push21:38
@lisitsynthis is tricky21:38
@HeikoSshould not be21:38
@HeikoSbecause we just iterate?21:38
@HeikoSand its all tree structure?21:38
@lisitsynHeikoS: yeah our template makes it very tricky21:39
@lisitsynHeikoS: I mean21:40
@lisitsynwe have tabpanel that serves python/java/..21:40
@lisitsynand it has tab-content that's basically whole page21:40
@lisitsynand inside of that we have another tab-content21:40
@lisitsynthis gets really tricky21:41
@lisitsynokay anyway that's html problem now21:41
@lisitsynI'll commit now and continue later21:42
@HeikoScool Ill check21:46
@HeikoSlisitsyn: wait sergey21:47
@HeikoSits different21:47
@HeikoSits not a whole page21:47
@HeikoSwe changed that21:47
@HeikoSthe template still has the thing globally21:47
@HeikoSbut only the code blocks are duplicated ,the page itself is not21:47
@lisitsynHeikoS: yes and it won't work globally21:47
@lisitsynI don't think it can work with nested tab-content21:48
@HeikoSbut do we need that even?21:48
@HeikoSah I see21:48
@lisitsynthere is only one section for tabs21:48
@lisitsynbut multiple tab sets21:49
@lisitsynlike for every snippet21:49
@lisitsynbut it's other problem21:49
@lisitsynI think I can fix that21:49
@lisitsynneeds some javascript21:49
@HeikoSdont know anything about this21:50
@HeikoShow doe sthat work?21:50
@lisitsynHeikoS: I think we need to add callbacks for these 'global' tabs21:53
@lisitsynand invisible 'local' tabs21:54
@lisitsyninvisible tab headers I mean21:54
@lisitsynso when one selects some global tab21:54
@lisitsynall corresponding local tabs are selected21:54
@lisitsyndo you follow the idea?21:55
nileshKLisitsyn: I am still having one more issue.21:57
@lisitsynnileshK: what's it?21:57
nileshKThis is the code we have for LpBoost example21:57
nileshKI am not able to collect the output after training21:58
nileshKthe object with the data.21:58
@lisitsynI don't get it21:58
@lisitsynwhat output?21:58
nileshKLike in this example if you check line no:21:59
nileshKYou see they are getting the weights w after training21:59
@lisitsynyou want the learned weight vector?22:00
nileshKI dont see how to collect the training output22:00
nileshKBut then if you move to line : 15922:00
nileshKwill just giving that w be good enough22:00
@lisitsynthat's random features specific22:01
nileshKI am trying to find an analogy b/w this code and lpboost example and create a similar one22:01
@lisitsynyou don't need to do anything with weights22:01
@lisitsynonce learned you can apply lpboost22:01
@lisitsynit will use learned weights22:01
nileshKthen how do you check the output for the test samples22:02
@lisitsynnileshK: line 176 of your example22:02
@lisitsynit returns binary labels with predicted classes22:02
nileshKthat code is not workgin22:03
@lisitsynwhat error?22:03
@HeikoSlisitsyn: actually, not sure22:04
@HeikoSbut I dont know much javascript22:04
@HeikoScan you gist an example?22:04
nileshKThis is the error which comes while training it22:06
nileshKThis is the gdb output22:10
-!- pickle27 [] has quit [Remote host closed the connection]22:30
@lisitsyn HeikoS: yeah a bit later22:33
@lisitsynnileshK: uh then it looks broken22:33
@lisitsynmay be some cplex version incompatibility..22:34
nileshKHow strong are you about that?22:34
@lisitsynno clue :)22:35
@HeikoSlisitsyn:  cool thanks!22:36
@HeikoSlisitsyn:  its a pitty we have no tests for thouse cplex things22:36
@HeikoSlisitsyn: sonney2k_ suggested to drop them as not open source22:37
@HeikoSany other way to implement such linear programing things?22:37
@lisitsynwell it is written before jesus christ22:37
-!- curiousguy13 [~curiousgu@] has quit [Ping timeout: 256 seconds]22:37
@lisitsynin 1953 or so22:37
@lisitsyndon't know22:37
@lisitsynI guess there is no unified cool open source solution for LP22:38
nileshKHeikoS: what do you mean by other way?22:38
@lisitsynw/o cplex22:38
nileshKCplex is good right?22:38
nileshKThere are other libraries lpsolve22:38
@lisitsyncplex is good but not open source yeah22:38
nileshKLpsolve is good option in that case22:39
@HeikoSnileshK: do you know other lp solvers?22:39
@HeikoSnileshK, lisitsyn could be a cool GSoC project22:39
nileshKI think yes. Because the amount of work that is currently required to get these things working22:40
nileshKis quite enough22:40
nileshKHeikoS just confirm one thing, do know whether lpBoost is tested?22:41
@HeikoSnileshK: no idea22:41
@HeikoSthis is all very old stuff before all of us were around22:42
@HeikoSso you will have to try, sorry22:42
@HeikoSbut any findings are absolutely welcome22:42
@HeikoSso document them on github issues22:42
@HeikoSthis increases probability of getting them fixed and working again22:42
nileshKYeah. As soon as i get everything working right and I am sure. I will write some examples and unit tests22:43
nileshKfor these. Also is there any time when Soren is available?22:43
nileshKlisitsyn: Are you going to be around?22:45
-!- curiousguy13 [~curiousgu@] has joined #shogun22:49
@HeikoSnileshK: s?ren is not available atm23:09
@HeikoSyou can try to email him, but its unlikely that he will reply, he is too busy currently23:10
-!- pickle27 [] has joined #shogun23:10
@HeikoSnileshK: is always around23:10
@HeikoSpickle27: hi!23:10
@HeikoSpickle27: did you see our take on sphinx and the manual?23:10
@HeikoSpickle27: we are almost there, just some things missing, pull from the repo and run make preview in the sphinx dir23:10
@HeikoSif you have an idea on the html tags, that would be useful23:10
pickle27I did see, looks like its getting close23:11
pickle27I've never used sphinx so I cant really add much there23:11
@HeikoSpython docutils is the thing we need to understand23:14
@HeikoSgoing to bed now. see you!23:14
nileshKGn HeikoS23:14
pickle27okay! I'll try and see if I have any time to look at it - got deadlines at work atm23:15
@HeikoSnileshK: night!23:15
@HeikoSpickle27: same ;)23:15
-!- HeikoS [] has quit [Quit: Leaving.]23:15
--- Log closed Mon Mar 30 00:00:21 2015