gsomix gsomix n4nd0 --- Log opened Mon Apr 23 00:00:19 2012 it's time to sleep 00:03 good night, guys 00:03 good night gsomix 00:04 -!- puffin444 [230bf329@gateway/web/freenode/ip.35.11.243.41] has quit [Quit: Page closed] 00:06 -!- blackburn_ [d5578adc@gateway/web/freenode/ip.213.87.138.220] has quit [Ping timeout: 245 seconds] 00:41 good night guys. 00:56 -!- harshit_ [~harshit@182.68.43.52] has quit [Remote host closed the connection] 00:57 -!- av3ngr [~av3ngr@42.62.206.121] has quit [Ping timeout: 240 seconds] 00:59 -!- av3ngr [~av3ngr@42.62.206.121] has joined #shogun 01:00 -!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has quit [Read error: Connection reset by peer] 01:10 -!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] 01:13 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] 02:23 -!- av3ngr [~av3ngr@42.62.206.121] has quit [Ping timeout: 260 seconds] 02:43 -!- av3ngr [~av3ngr@42.62.206.121] has joined #shogun 02:45 -!- pluskid [~pluskid@173.254.214.60] has joined #shogun 03:42 -!- av3ngr [~av3ngr@42.62.206.121] has quit [Ping timeout: 260 seconds] 04:02 -!- xiangwang [~chatzilla@159.226.60.224] has joined #shogun 04:04 -!- xiangwang [~chatzilla@159.226.60.224] has quit [Client Quit] 04:05 shogun: Soeren Sonnenburg master * r719f903 / examples/undocumented/ruby_modular/modelselection_parameter_tree_modular.rb : disable printing of combinations - http://git.io/jAv3ew 06:16 -!- Netsplit *.net <-> *.split quits: pluskid 06:57 -!- Netsplit over, joins: pluskid 07:14 build #507 of ruby_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/507 07:30 -!- vikram360 [~vikram360@117.192.178.207] has quit [Ping timeout: 250 seconds] 07:37 -!- av3ngr [~av3ngr@122.110.219.42] has joined #shogun 07:54 -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun 08:26 -!- vikram360 [~vikram360@117.192.190.229] has joined #shogun 08:46 -!- vikram360 [~vikram360@117.192.190.229] has quit [Ping timeout: 246 seconds] 08:51 -!- vikram360 [~vikram360@117.192.190.169] has joined #shogun 08:52 pluskid: I was just reading your patch thx! 09:10 pluskid: to get the idea - you introduce multiclassstrategy that can be attached to any multiclass classifier and will generate configurations to be learned by some binary classifier / same with evaluation 09:16 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun 09:17 sonne|work, yes 09:18 pluskid: that makes a lot of sense - only thing that might be confusing to users is that we have *true* multiclass machines 09:19 sonne|work, the motivation is that blackburn suggest me to write a tree-structure training strategy 09:19 like KNN that will directly do everything 09:19 sonne|work, yes, like many MulticlassSVM 09:19 they handle training directly 09:20 I think we have only one multiclass svm - GMNP 09:20 libsvmmulticlass is also doing OvO 09:20 maybe LaRank? 09:20 yes 09:20 true 09:20 -!- emrecelikten [~Anubis@176.40.232.233] has joined #shogun 09:20 I think either way (the old and the new) of handling multiclass strategy is OK, if we only have OvO and OvR 09:21 but if we want to introduce more, this way is more convenient 09:21 especially when we want to attach some more data to the strategy 09:21 for example, in a tree-structure training, we need the user to define a tree and attach that to the strategy object 09:22 pluskid: no your way is much better 09:23 I mean you should read the ECOC JMLR paper / code 09:23 there are so many strategies to do so it is much nicer the new way 09:23 hmm, you are right 09:23 pluskid: only thing I don't know is how to make clear in the class hierarchy that we have true multiclass machines 09:24 not requiring any of this 09:24 yes, and training and applying can also be separated 09:24 yeah only training is true multiclass for many 'true' multiclass machines 09:25 some machines use their own training, but use the generic strategy for applying 09:25 yeah 09:25 not clear how to handle that yet 09:25 pluskid: anyway the best way to proceed is to implement other ECOC schemes from the paper - things will become clear then 09:27 pluskid: btw I added some (minor) comments to your patch - please address them and then it is ok to be merged 09:27 sonne|work, yes, practicing is the best teacher :p 09:27 OK, I'll look at them 09:28 -!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] 09:28 pluskid: thanks and please continue the great work! 09:32 -!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun 09:32 -!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] 09:32 -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun 09:32 sonne|work, thanks, I'll try my best :) 09:32 -!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun 10:00 -!- emrecelikten [~Anubis@176.40.232.233] has quit [Quit: Leaving.] 10:20 sonne|work: multiclassliblinear and multiclassocas are true too 10:24 haha, my laptop is *just* capable of compiling shogun, if I opened other big prog (like Matlab) simutaneously, the compiler will complain out-of-memory 10:26 pluskid: oh you are here ;) 10:27 you did great work 10:27 I haven't been thinking about it actually 10:28 blackburn, thanks, I'm making some modifications according to sonney2k's feedback 10:28 -!- emrecelikten [~Anubis@176.40.232.233] has joined #shogun 10:29 -!- pluskid [~pluskid@173.254.214.60] has quit [Ping timeout: 260 seconds] 10:34 -!- pluskid [~pluskid@111.120.43.103] has joined #shogun 10:52 sonne|work, blackburn I updated the pull request 10:52 yes thanks 10:52 pluskid: have you seen ECOC training code in scikits learn? 10:54 blackburn, no 10:54 pluskid: https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/multiclass.py 10:54 looks pretty easy actually 10:55 hmm 10:56 I think I understand the basics 10:57 after reading its comments 10:57 pluskid: yes rather easy - I am also having some long discussion with v. franc about training ECOC svm in true multiclass fashion 10:58 what does that mean? "in true multiclass fashion" 10:59 pluskid: in one optimization problem 11:00 implies better normalization 11:00 OK 11:00 pluskid: I mean in basic ECOC with given K bits 11:01 you would need to construct K classifiers 11:01 yes 11:01 right? 11:01 so you train the K classifiers in one time? 11:01 I've been thinking about SVM (looks like SO at some place) 11:01 actually I'd like to construct K classifier for each class 11:02 however K classifiers at the same time would work too may be 11:02 K classifier for each class? 11:02 hmm yes if I didn't get crazy 11:03 I didn't quite get you, so e.g. we have C class, K bits, then we have CK classifiers 11:03 with a test point x, how do we classify it? 11:04 max_{c\in C} \sum_{k\in K} 11:05 does it make sense for you? 11:05 I am unsure already 11:05 :D 11:05 let me think 11:06 it is not ECOC :D 11:06 daaamn I have to think about it 11:06 no, this has nothing to do with ECOC 11:07 ordinary one-vs-rest linear machine, haha 11:07 move the sum inside the inner product 11:07 not ovr 11:07 composition of K OvR - useless 11:08 hmm, but you can move the sum into the inner product, then \sum_{k\in K}w_c^k can be treated as a single w_c 11:09 ah right 11:09 pluskid: no, decision rule is wrong 11:10 blackburn, why? 11:10 should be 11:10 I don't understand 11:11 should be different I mean 11:11 :D 11:11 pluskid: nevermind this time 11:11 :) 11:11 shogun: Soeren Sonnenburg master * r1ccb9e4 / (21 files in 5 dirs): 11:17 shogun: Merge pull request #479 from pluskid/multiclass-strategy 11:17 shogun: Multiclass strategy refactoring (+14 more commits...) - http://git.io/Lf2ifA 11:17 pluskid: thanks for the quick updates! 11:17 sonne|work, :) 11:17 hope build-bot is happy with my patches, though I have already checked locally 11:18 pluskid: I hope so too :D 11:18 blackburn: how many exams left? 11:18 sonne|work: 3 11:18 sonne|work: but they are not related with optics 11:18 so I am available 11:19 :D 11:19 haha 11:19 sonne|work: do you think I should ask SLEP guys to GPL their code under their copyright? 11:20 blackburn: I think we should do the intrusive renaming changes rather soon 11:20 sonne|work: which ones? 11:20 blackburn: like moving around multiclass classifiers 11:20 and multitask / domainadaptation stuff 11:20 sonne|work: I still think multitask normalizers are better suited to be in transfer/multitask.. 11:21 then the only remaining intrusive stuff for this year is IMHO sgvector 11:21 maybe but the should be totally hidden from the user 11:22 sonne|work: but what to rename? 11:24 -!- pluskid [~pluskid@111.120.43.103] has left #shogun ["for supper"] 11:26 -!- pluskid [~pluskid@111.120.43.103] has joined #shogun 11:27 build #726 of cmdline_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cmdline_static/builds/726  blamelist: pluskid@gmail.com 11:33 blackburn: rename includes moving things around :) 11:34 right 11:35 blackburn: so? 11:35 sonne|work: so what? 11:36 actually you are right we need to move knn, etc to multiclass 11:36 build #704 of r_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/704  blamelist: pluskid@gmail.com 11:36 I am saying better do all the rename / move now than later when we run on full steam 11:37 ah yeah I'll do something about it 11:39 build #705 of octave_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/705  blamelist: pluskid@gmail.com 11:41 build #692 of python_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/692  blamelist: pluskid@gmail.com 11:45 see you later people! 11:46 see yopu 11:47 you 11:47 hehe 11:47 :D 11:47 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] 11:48 build #727 of cmdline_static is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cmdline_static/builds/727 11:49 build #705 of r_static is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/705 11:54 build #706 of octave_static is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/706 11:59 build #693 of python_static is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/693 12:04 build #517 of java_modular is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/517  blamelist: pluskid@gmail.com 12:04 build #540 of octave_modular is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_modular/builds/540  blamelist: pluskid@gmail.com 12:05 build #528 of csharp_modular is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/csharp_modular/builds/528  blamelist: pluskid@gmail.com 12:06 build #516 of lua_modular is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/lua_modular/builds/516  blamelist: pluskid@gmail.com 12:07 build #513 of python_modular is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/513  blamelist: pluskid@gmail.com 12:14 build #508 of ruby_modular is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/508  blamelist: pluskid@gmail.com 12:20 build #512 of r_modular is complete: Failure [failed compile]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_modular/builds/512  blamelist: pluskid@gmail.com 12:21 build #518 of java_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/518 12:31 -!- pluskid [~pluskid@111.120.43.103] has quit [Ping timeout: 246 seconds] 12:35 -!- pluskid [~pluskid@173.254.214.60] has joined #shogun 12:35 build #541 of octave_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_modular/builds/541 12:39 -!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has joined #shogun 12:41 hello everybodyh 12:44 build #529 of csharp_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/csharp_modular/builds/529 12:46 build #517 of lua_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/lua_modular/builds/517 12:53 >> Welcome again to the Machine learning class. 12:59 huh 12:59 hi all 12:59 hi gsomix 12:59 :p 12:59 gsomix, yes, a lot of coursera courses start today 13:04 :p 13:04 build #509 of ruby_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/509 13:06 build #513 of r_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_modular/builds/513 13:13 happiness again 13:22 blackburn: look at python_modular 13:27 sonne|work: yes 13:30 I'll fix once I get home 13:30 sonne|work: I think that's ok to patch true multiclass machines 13:35 to avoid this error we can add trained flag there 13:35 so once we train machine we store number of classes 13:35 blackburn: just think of doing x-validation with Mc machines 13:36 what is the problem? 13:36 for some subsets you observer x for some x-1 classes or less 13:36 same problem with sparse features btw 13:36 there could be 1000 dims or 100 13:36 hey 13:36 ?? 13:36 that is the reason to train on different number of classes? 13:36 no sense for me at all 13:37 -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] 13:37 -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun 13:37 blackburn: you said you want to *infer* number of classes based on labels 13:37 that is dangerous 13:38 yes train labels should contain all possible classes 13:38 I think specifying nr classes is better 13:38 blackburn: why should they? 13:38 if you do x-val you might not have some classes 13:38 bad x-val 13:38 not bad xval 13:38 but you will get wrong results! 13:39 sonne|work: how can you ignore some classes? 13:40 so if you skip class C you will have zero w_c, what makes you able to use it? 13:42 wiking: hey 13:43 I support specifying nr classes when constructing a classifier 13:43 just a matter of taste, however 13:43 pluskid: both ways are ok 13:43 however crossvalidation with class ignores makes no sense to me 13:44 things would be totally wrong if you skip some classes 13:44 I mean it would be easier to implement the code if we are given nr classes at first 13:45 then no need for a flag like m_trained or similar 13:45 pluskid: where is it set? 13:45 no in current code 13:45 but I think it might be needed 13:45 pluskid: yes that's I am asking 13:45 on which step? 13:46 ah 13:46 not in current? 13:46 for example, you do not know the number of machines, until it is trained 13:46 how does it work now 13:46 ? 13:46 for example, in MulticlassMachine 13:47 get_num_machines() 13:47 will return "wrong" value, until the machine is actually trained 13:47 We need to know the state of the machine to see whether some "knowledge" is already available 13:48 here the number of classes 13:48 the same problem happens in MulticlassStrategy 13:48 MulticlassStrategy::get_num_machines(int32_t num_classes) 13:48 one has to pass num_classes as an argument when calling the function 13:49 but if we know num_classes from the begining, there's no need for this argument 13:49 anyway, I think it is just a matter of taste, coz the current solution also works 13:49 pluskid: I think we can assume no ties in labels while training 13:50 however it looks like it would be better to store number of classes on training 13:50 and obtain it while applying 13:50 this way you can be sure it works with ties in test set 13:51 blackburn: ho :) 13:51 wiking: does it make sense to you to skip some classes in train set? 13:52 (that are not skipped in test one) 13:52 ?? 13:52 blackburn, I'm OK with this solution 13:52 blackburn: why would one do that? 13:52 wiking: x-val 13:52 though still prefer specifying nr classes in advance :p (my taste...) 13:53 pluskid: in advance? 13:53 I mean when constructing the classifier 13:53 when training or constructing? 13:53 pluskid: problem is that if you have ties in labels 13:54 like test set contains only 1,2, 4 instances 13:54 of classes 1,2,4 13:54 I mean 13:54 so get_num_classes would return 3 13:54 and you will run only 3 of 4 machines 13:55 skipping last one - it looks incorrect 13:55 I mean in the constructor 13:58 before training 13:58 what do you mean by "ties in labels"? 13:58 1,2,4 13:58 no 3 13:58 pluskid: but what if you set different labels? 13:58 but the number of labels won't change, will it? 13:59 number of classes 13:59 why not? 13:59 you can set features, kernel, labels - everything 14:00 You mean 1,2,4 or 2,3,4, or 1,2,3, etc? 14:00 just some missing class 14:00 OK 14:00 so we need set_num_classes, haha 14:00 pluskid: yes we need to save num of classes on training 14:00 we can't rely only on constructor - set_labels() would make no sense then 14:01 blackburn, OK, that makes sense 14:01 however this requires to set number of classes in strategy 14:02 in every true svm as well 14:02 true multiclass* 14:02 yes, that's the problem 14:02 not really 14:02 oneliner 14:02 blackburn: I think that is the same issue like having a linear classifier operating on sparse features - sometimes not the last dim in any of the sparse feature vectors is non-zero so you need to specify length of w of the sparse classifier or so 14:03 sonne|work: are you still sure train set can contain a few classes? 14:03 (not all classes) 14:04 sure 14:04 sonne|work: and how would you classify class X if you haven't trained a classifier for that? 14:05 it will never predict class X but operate correctly otherwise 14:06 sonne|work: it broke other class' results 14:06 break* 14:06 why? 14:06 sonne|work: hmm I assume it would be w=0 14:06 exactly 14:07 sonne|work: it can override some other class scores 14:07 ah damn 14:07 ? 14:07 0 always 14:07 hah 14:07 okay sorry 14:07 <0,x>=0 14:07 sonne|work: it complicates things pretty much 14:08 I think if the samples for a class is missing, the problem is not with predicting, but with training. You need a one-class algorithm to train a "classifier" :D 14:09 pluskid: I am not even sure that would work with current algorithms 14:09 like gmnp/etc 14:09 however, if this happens, it should be an error 14:10 so if we are specified nr classes, we might be able to detect this kind of error 14:10 be back in a while 14:10 we need map there 14:11 to avoid any errors 14:11 we can't keep things like that anymore 14:11 -!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] 14:57 -!- gsomix [~gsomix@188.168.5.255] has quit [Ping timeout: 265 seconds] 15:00 -!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has quit [Read error: Connection reset by peer] 15:09 -!- harshit_ [~harshit@182.68.43.52] has joined #shogun 15:27 -!- ted__ [2e1fd566@gateway/web/freenode/ip.46.31.213.102] has joined #shogun 15:28 hi all 15:28 what type of Features do I need to use (c++) if I want to train multiclass svm using string kernel with float array samples of different size ? 15:30 thanks for help 15:30 and which string kernel is suitable for samples of different sizes ? 15:31 -!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has joined #shogun 15:40 -!- gsomix [~gsomix@83.234.54.70] has joined #shogun 15:50 RE: hi there, what type of Features do I need to use (c++) if I want to train multiclass svm using string kernel with float array samples of different size ? and which string kernel is suitable for samples of different sizes ? Thanks in advance. T. 15:54 ted__, hi ,I havn't had much experience with strings in shogun but i think you need to use CStringFeatures as features . 16:04 ted__, have a look at examples/documented/libshogun/features_copy_subset_string_features 16:04 ted__, Also wait for a while till other members are online [blackburn, soeren] they will help you better 16:05 -!- blackburn [~qdrgsm@188.168.2.179] has joined #shogun 16:18 4hours40 :S 16:20 blackburn, hi , I think ted__  needs ur help :) 16:20 harshit_: yes I just read logs 16:21 -!- vikram360 [~vikram360@117.192.190.169] has quit [Ping timeout: 265 seconds] 16:21 PhilTillet, really nervous moments :/ 16:21 ted__: StringCharFeatures ? 16:23 harshit_, let's just get our body frozen for 5 hours 16:26 pluskid: I don't like dependency of multiclassmachine and multiclassstrategy 16:30 PhilTillet, haha .. 16:30 -!- pluskid [~pluskid@173.254.214.60] has quit [Ping timeout: 246 seconds] 16:33 blackburn, I think there might me some issue with svmlin.cpp present in shogun, I have compared the results of original svmlin and that of shogun's, they are close but not exactly similar . 16:39 blackburn, have a look : http://snipt.org/uhjhh3 16:40 are all parameters equal? 16:40 blackburn, In svmlin original library we can set lambda so I set it to 1 16:41 sonne|work, harshit_ : sorry, was away 16:41 harshit_, I think I'll drink vodka in the 17-18UTC and go to bed, aha 16:41 blackburn, but in shogun internaly lambda = 1/2c so I make value of C as 0.5 16:41 harshit_: but other params? are there any? 16:42 blackburn, only other parameter is epsilon which is 1e-5 in both 16:42 Let's resume: I have samples of different sizes as vectors of float , I am looking for best shogun feature structure and kernel to train multiclass svm. 16:43 I'll take a look at StringCharFeatures 16:43 what about kernels ? 16:43 ted__: vectors of float? 16:44 yes 16:44 of variable size? 16:44 gsomix, I think vodka is for blackburn, we should go for tequila ! 16:44 y 16:44 ted__: non-trivial.. how can you compare it? 16:44 I know, I'm guessing stringkernel could do the work as they implement smthg like dtw distance , 16:45 ? 16:45 harshit_, huh, I have never tried tequila. 16:46 It can be vectors of int but sizes are different. 16:46 g2g, blackburn i'll do some more debugging and will let you know about the status ! 16:46 harshit_: yeah thanks 16:46 ted__: yes makes sense but they are for more 'finite' features like char 16:47 gsomix, you should, maybe today !! 16:47 blackburn: and I dont want to use simple bag-of-words representation 16:47 ted__: what is the nature of your features? 16:47 blackburn: [w1,w2,w3,..] with w_iEUR[0..N] N could be 10, 100, 1000, or 10000 16:48 and size of the "string" is not defined 16:48 ted__: have you any idea of similarility function? you can create custom kernel actually 16:49 ted__, what I think is that you should first use string to manipulate the words and then prepare its corresponding VSM model which you can store in a normal CSimpleFeatures 16:49 -!- pluskid [~pluskid@li381-56.members.linode.com] has joined #shogun 16:50 harshit_: VSM ? 16:50 vector space model 16:50 in which you have words as features and documents as examples/datapoints 16:51 -!- harshit_ [~harshit@182.68.43.52] has left #shogun ["Leaving"] 16:52 I thought that "string kernel" could handle features of different size ? 16:52 ted__: yes if they are contained with chars 16:53 then clustered by [0,255] ? 16:54 yes would work 16:54 Then it could be same algo for string of cluster indices (indices btw 0 and 1000..) 16:55 Well, which shogun features/cluster could I overload if I want to do that ? 16:56 ted__: could you please clarify what exactly you need? 16:56 Well, which shogun features/kernel could I overload if I want to do that ? 16:57 repeating sentence is not a clarification I'd say :D 16:57 blackburn: sry, I just type cluster and meant kernel . 16:58 aha 16:58 ted__: you don't really need to overload features for that 16:59 blackburn: So. I have a set of indices vector of different size. Let's say Indices EUR [0,100]. Each vector is associated with a label. let's say we have 10 labels. 16:59 -!- pluskid [~pluskid@li381-56.members.linode.com] has quit [Quit: Leaving] 17:00 blackburn: I want to train multiclass svm of this data. Problem is that features have different size. 17:00 -!- pluskid [~pluskid@li381-56.members.linode.com] has joined #shogun 17:00 blackburn: indeed I have read about string kernel apply to string of char. I want to do the same with my problem. 17:01 ted__: you may try to construct StringFeatures with your features 17:01 and use any kernel 17:01 string kernel I mean 17:01 ok. And if I have more than 256 indices ? 17:02 ted__: you have vecotr x = [x1,x2,...] where xi \in [0,255]? 17:03 vector* 17:03 what do you mean more that 256 indices? wider range of xi values? 17:03 blackburn: I have vector x = [x1,x2, ..] where xi \in [0,1000] 17:05 -!- vikram360 [~vikram360@117.192.190.169] has joined #shogun 17:05 ted__: ok and what is the similarity measure you want to use? 17:06 I am asking because string kernel are mainly used by bioinformaticians to estimate similarities between some ACTCGCT and TCGCGA - not a big alphabet 17:07 I know 17:08 wiking,gsomix,pluskid,blackburn, do you have some homepage I could link to? 17:08 I haven't 17:08 sonney2k: mmm not really :( 17:08 but i'm planning to keep a weekly posting on maeth.com 17:08 sonney2k: http://pluskid.org 17:08 but currently it's personal rampage :) 17:08 shogun-buildbot, nope 17:09 sonney2k, nope 17:09 huh 17:09 Dunno exactly yet what type of  similarity measure. 17:09 Maybe for the beginning one you used with your already implemented string kernels which support features of diff size 17:10 LocalAlignmentString ? 17:10 Any advice ? :p 17:10 ted__: it is based on alignment - I am not sure it would work for your features 17:11 ted__: where these features come from? 17:11 pluskid, cool pictures :] 17:12 spatial arrangement of my vectors are important: features extracted from video sequence, then I would like to keep the temporal info (=as in ACTCGCT) 17:13 gsomix: you like them? :D 17:13 pluskid, yep 17:13 gsomix: thanks! haha, I like drawing 17:14 ted__: really non-trivial thing you do 17:14 I know but it seems to be a good way to keep temporal information which is lost in a bag-of-frame algo. 17:22 maybe I could try with a mapping to [0,255] and see the results 17:22 -!- pluskid [~pluskid@li381-56.members.linode.com] has quit [Quit: Konversation terminated!] 17:23 thanks for helping , I anythg else come in mind, I stay here for a couple hours more 17:23 -!- pluskid [~pluskid@li381-56.members.linode.com] has joined #shogun 17:23 pluskid: you forgot to ref multiclass strategy ;) 17:24 blackburn, oh, my fault! 17:25 pluskid:  I fixed 17:25 could you please help me to fix this? 17:25 will commit in a min 17:25 thanks! 17:25 pluskid: I am currently working on getting examples to work 17:25 because of that enum -> class change 17:25 blackburn: do you have an example using string kernel and multiclass svm ? 17:26 blackburn, I see, I fixed libshogun, but it seems I forgot python_modular 17:26 blackburn, thanks again! 17:26 -!- av3ngr [~av3ngr@122.110.219.42] has quit [Ping timeout: 260 seconds] 17:26 ted_: in C++? 17:26 blackburn:  yes please, with features of diff size 17:27 ted__: check examples/undocumented/libshogun/ for some examples 17:30 I don't think we have any complex C++ example of string features and kernel learning 17:30 arf 17:31 blackburn: I will try EAlphabet with RAW BYTE - [0-255] 17:32 sonney2k: why did you collected the time zones for all the candidates? Is there any online conferences? 17:32 pluskid: yes for kickoff meeting 17:32 when? 17:32 I already checked those examples, was hoping you had somthg more specific on your own machines. 17:32 ted__: I am sorry, I haven't worked with such features (yet) 17:33 blackburn: thanks anyway for all those info. Maybe anyone else have ? 17:34 ted__: yes sonney2k has some experience with strings but he is pretty busy 17:34 pluskid: got error :( 17:35 File "classifier_multiclassmachine_modular.py", line 25, in classifier_multiclassmachine_modular 17:35 mc_classifier.train() 17:35 SystemError: [ERROR] Not a two class labeling - no positively labeled examples found 17:35 blackburn: leave that for me 17:37 pluskid: ok I'll push some fixes 17:37 blackburn: thanks! 17:37 shogun: Sergey Lisitsyn master * reeaa056 / (5 files in 3 dirs): A bunch of fixes for multiclass machine - http://git.io/jfjX2g 17:38 blackburn: BTW, what is the kickoff meeting for? 17:38 pluskid: hmm I think we need to discuss our plans 17:39 it is already late there you are.. 17:39 yes 17:39 pluskid: will you be available at 19 UTC? 17:39 I mean, we don't need to be there when Google announce the result, do we? 17:39 yes that's not required 17:39 OK, cool 17:40 if it's 15:40 UTC now 17:40 I don't think I will be available at 19 UTC 17:40 unless required 17:40 have to sleep :p 17:40 pluskid: when will you wake up? 17:41 let me calculate 17:41 around UTC 00:30 maybe 17:41 there is a chance I will not be sleeping yet 17:41 a lot of things to do today 17:42 :) 17:42 a regular sleeping time is good for a programmer! 17:42 to live longer! :D 17:42 pluskid: it would be painful to not complete these things today :) 17:43 pluskid: okay just read logs once you wake up 17:43 blackburn: hmm, then wish you finish them soon 17:43 sure 17:43 sonney2k: I am looking for C++ example of string features (with different size) and (multiclass) kernel learning. I know you are pretty busy, I you have time to share an example it would be great. Thank you in advance. 17:52 If* 17:52 -!- blackburn [~qdrgsm@188.168.2.179] has quit [Ping timeout: 272 seconds] 17:52 -!- harshit_ [~harshit@182.68.43.52] has joined #shogun 17:57 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun 17:57 n4nd0, yo 17:58 gsomix: hey! 17:58 n4nd0, how are you? 17:58 gsomix: I am fine, what about you? 17:58 and hey to all people :) 17:59 n4nd0, fine. I'm waiting for 19UTC. 18:00 gsomix: I think we all are! 18:01 gsomix: are you nervous? 18:01 gsomix, join the club :p 18:01 I am getting more and more nervous :O 18:01 n4nd0, a little. 18:01 boys, keep relax, since no matter what we do *within the several hours*, we can not change the (currently unknown) result :p 18:03 :p 18:08 good night all, hope to see good news when I wake up 18:09 -!- pluskid [~pluskid@li381-56.members.linode.com] has quit [Quit: Leaving] 18:09 pluskid: good night and good luck! 18:10 sonney2k: Is there any example of using string features with Alphabet [0,255] ? 18:23 :>>> 18:36 =>> 18:36 gsomix, dont' wait code :) 18:52 sonney2k, moin. what do you mean? 18:54 gsomix, SGVector is waiting for you :) 18:55 sonney2k, ah, I know.  I'm doing now a small patches for wiking and blackburn. 18:56 shogun: Soeren Sonnenburg master * r9cd129d / src/shogun/multiclass/MulticlassOneVsRestStrategy.cpp : 18:56 shogun: Merge pull request #481 from pluskid/multiclass-fix 18:56 shogun: Fix a bug in OvR strategy: inc counter in wrong point - http://git.io/CEHNVw 18:56 shogun: Chiyuan Zhang master * r0814c83 / src/shogun/multiclass/MulticlassOneVsRestStrategy.cpp : Fix a bug in OvR strategy: inc counter in wrong point - http://git.io/B-fS_Q 18:56 gsomix, SGVector is more important I think - lets hope blackburn knows that 18:57 gsomix: \o/ 18:57 -!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun 19:06 sonney2k: yes but it needs clarification 19:08 hey guys less than two hours left 19:10 yes ^^' 19:11 -!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun 19:12 -!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds] 19:15 -!- wiking_ is now known as wiking 19:15 blackburn, clarification in which respect? 19:15 sonney2k: explain to gsomix more what to do 19:17 gsomix, I don't know what to explain actually ... is there anything in particular that is unclear? 19:18 ref/unref? 19:18 I mean just continue to use SGVector& in arguments 19:18 blackburn, but the SGVector& stuff is not yet complete or gsomix? 19:19 ah this transition should be pretty clear 19:19 huh, I had planned to discuss it tomorrow. :) 19:19 but ok 19:19 sonney2k: ok then 19:19 gsomix, so please explain :) 19:20 sonney2k, hmm, I already did almost work for SGVector -> SGVector& in arguments, isn't it? 19:21 I am still wondering how to return SGVector from methods. 19:22 gsomix, SGVector& except if you return a new sgvectro 19:22 SGVector 19:22 then just ^ 19:22 gsomix, ok? 19:23 sonney2k, it's clear for me. But I worry about how it is then transferred to ref/unref. 19:24 -!- harshit_ [~harshit@182.68.43.52] has quit [Quit: Leaving] 19:24 sonney2k, ok, I'll do it in the next few days. 19:25 gsomix, we do the ref/unref business after this is done 19:25 gsomix, note that this will likely cause some errors that we cannot fix before we have ref/unref stuff 19:26 gsomix, if you want to know about ref/unref stuff - let us just copy the functions we have in SGObject for now 19:26 sonney2k, well I think step-by-step strategy is correct. 19:27 later on we should investigate whether we add another base class to derive SGVector / SGObject from 19:27 gsomix, btw, the same task then for SGMatrix/SGStringList/SGSparseVector ... 19:27 SGNDArray .. you name 19:27 it 19:27 sonney2k: a question, why to return SGVector and not a reference? 19:32 I mean what you said to gsomix ^ 19:32 n4nd0, if he allocates a new vector within a function (as local variable) - he cannot return a reference to it 19:33 since the reference will be invalid 19:33 (object de-allocated when the function returns) 19:34 wouldn't it better to create the vector to return in the heap? 19:34 blackburn, check a git. 19:34 *github 19:34 some stuff ... = new SGVector<...more stuff 19:35 gsomix: nice 19:35 n4nd0, I think it's not good at now (without ref/unref). 19:37 sonney2k: well actually SGVector is kind of light, a pointer and a couple of more things so probably the return by value shouldn't be much worse than returning a reference 19:37 gsomix: do you prefer return by value as well? 19:37 shogun: Soeren Sonnenburg master * r3007eb5 / src/shogun/io/StreamingAsciiFile.cpp : 19:37 shogun: pure whitespace changes 19:37 shogun: - convert file to use tabsize 4 - http://git.io/Ud_lvg 19:37 shogun: Soeren Sonnenburg master * r098c0cc / src/shogun/lib/v_array.h : 19:37 shogun: delete v_array's begin on exit 19:37 shogun: This hopefully fixes the memory leak observed in 19:37 shogun: https://github.com/shogun-toolbox/shogun/pull/401 - http://git.io/gfTGeA 19:37 n4nd0, it is a necessity, I mean 19:38 gsomix: I wondered if return by reference was an option 19:43 n4nd0, yes the object is alloc'd on heap however just *once* since RVO is used https://en.wikipedia.org/wiki/Return_value_optimization 19:44 n4nd0, there is no option 19:44 n4nd0, see http://stackoverflow.com/questions/4643713/c-returning-reference-to-local-variable and the first answer 19:45 sonney2k: ok, thank you 19:45 sonney2k: any preliminary date for kickoff? 19:47 n4nd0, I tried to fix that leak in streaming features - can you check if it still occurs? 19:48 blackburn, some people have ECML deadline tomorrow morning so not before that :D 19:48 maybe wednesday or some other day this week 19:49 sonney2k: yeah chris has 19:49 yeah he is using shogun :) 19:49 what is the relation? 19:49 sonney2k: I'll check it 19:52 shogun: Evgeniy Andreev master * r15afb6f / (14 files in 5 dirs): knn and conjugate to multiclass - http://git.io/XbLKXg 19:59 shogun: Soeren Sonnenburg master * rc46d1c0 / (14 files in 5 dirs): 19:59 shogun: Merge pull request #482 from gsomix/to_multiclass 19:59 shogun: move KNN and ConjugateIndex to shogun/multiclass - http://git.io/uSr2zQ 19:59 I think qda could me moved to shogun/multiclass too ^ 20:00 be moved, sorry 20:00 -!- harshit_ [~harshit@182.68.43.52] has joined #shogun 20:01 n4nd0, send a pull request 20:02 n4nd0, btw didn't you also convert LDA to multiclass? 20:02 blackburn, that reminds me - didn't we want to have CMulticlassLabels? 20:03 sonney2k: no, I didn't do anything to LDA 20:03 sonney2k: once, even before QDA was done, it was commented though 20:03 sonney2k: yes 20:04 sonney2k: we will do it once 20:04 n4nd0, too bad - that's really something we should have if we have multiclass enabled QDA 20:05 sonney2k: just for completeness or is there something else? 20:06 n4nd0, multiclass LDA is certainly much more robust than QDA :) 20:07 so it is really a very useful baseline method 20:07 much faster than SVM on low-dim data sets and even optimal when you can assume gaussian distribs 20:07 sonney2k: btw have you seen that SLEP supports some logistic regression even? 20:09 SLEP? 20:09 yes that package I am going to integrate 20:10 but we have logistic regression in liblinear already?! 20:10 sonney2k: why not to provide alternative? 20:11 I just don't understand your excitement 20:11 sonney2k: it is library containing a lot of L1 and L1/L2 things 20:12 sonney2k: I quite don't understand why QDA may fail somewhere LDA successes 20:14 I thought of QDA as a generalization, within class covariances need not to be equal, so I assumed it would give better results 20:15 even if teh covariances are the same for all the classes, case where LDA would perform fine too 20:15 -!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] 20:16 n4nd0, estimating more parameters is difficult - so for noisy / not a lot of data LDA will work better 20:25 sonney2k: ok, intuituively I understand too that QDA may be more prone to overfitting 20:26 * sonney2k is bored 20:31 27min :] 20:32 sonney2k: double free or corruption in muticlasslinearmachine example, pretty ugly stack trace :S 20:33 n4nd0, paste :) 20:33 http://snipt.org/uhjja8 20:35 -!- nickon [~noneedtok@kotnet-147.kulnet.kuleuven.be] has joined #shogun 20:36 -!- vikram360 [~vikram360@117.192.190.169] has quit [Ping timeout: 260 seconds] 20:39 -!- blackburn [~qdrgsm@188.168.2.179] has joined #shogun 20:41 sonney2k: gdb trace is pretty cool too 20:41 full of 20:41 [New Thread 0xb6345b70 (LWP 22769)] 20:41 [Thread 0xb6345b70 (LWP 22769) exited] 20:41 15 in room 20:41 oh 20:41 blackburn: do you mean 15 people there? 20:41 yeap 20:42 blackburn: oooh! you have followers man 20:42 followers? 20:42 yeah, I don't know :P 20:42 I thought you were all waiting for GSoC's answer 20:43 n4nd0, GSoC answer? What are you talking about? :p 20:43 hahah the crowd gets excited ... ;) 20:43 what is GSoC? :o 20:43 i bet gsoc site will be flooded 20:43 like last year 20:43 but as far as I remember the result is also send by mail 20:44 hehehe i wonder why don't they do a caching 20:44 and just do a simple precompiled html and that's it 20:44 * emrecelikten starts biting nails 20:44 sonney2k: oh wait, maybe I was handling wrong SG_UNREF in my example 20:45 are you fskin' ready? ;) 20:49 f5 spree 20:50 no way to see that even after 5 minutes of deadline 20:51 haha 20:52 yes I know 20:52 :p 20:52 at first not all the names are loaded anyway 20:52 Is it gonna be same as the organizations names were released ? 20:53 no 20:53 in principle there's also an e-mail 20:53 email is pretty late 20:54 last year I had the e-mail 5mins before the deadline 20:54 :p 20:54 but I guess it was some bug 20:55 n4nd0, can you be more specific? 20:55 or attach a gdb bt 20:55 I cannot make sense of that 20:55 sonney2k: I think the double free was caused by SG_UNREF I am doing and shouldn't be done 20:56 which one? 20:56 in any case I think that to use more SG_UNREF shouldn't cause that 20:56 in the example classifier_multiclasslinearmachine 20:57 the one with which I detected those memory leaks 20:57 The emails are going out 20:57 well good luck to all guys! 21:00 ;) 21:00 good luck guys 21:00 good luck :) 21:01 sonney2k: you are free to announce I think ;) 21:01 site is down :( 21:02 hehe 21:02 "Error retrieving data: please refresh the list or the whole page to try again" 21:02 :> 21:02 yeey for google :) 21:02 :) 21:02 I'm in too from another project 21:03 Best of luck to you guys 21:03 :) 21:03 emrecelikten: congrats! 21:03 emrecelikten, congrats ! 21:04 Thanks guys :) 21:04 any of you received email? 21:04 - 21:05 I did 21:05 no 21:05 I didn't 21:05 - 21:05 heheh melange is really SLOW 21:05 yes 21:06 The servers experience something like delta(t) * e^(-at) now I guess 21:06 it loaded about 500 students (no shogun :p) and then crashed 21:06 PhilTillet, same here ! 21:06 same again 21:09 after 400, crash! 21:09 XD 21:09 heheh 21:12 they have only the first 100 loaded 21:12 oh mailing list will be full of crazy stuff again 21:13 :D 21:13 -!- karlnapf [~heiko@host86-180-223-101.range86-180.btcentralplus.com] has joined #shogun 21:13 blackburn, why ? 21:14 hi all 21:14 hi karlnapf  :p 21:15 harshit_: gsoc mailing list is full of stuff every year 21:15 PhilTillet, hi there 21:15 karlnapf: hey 21:15 blackburn, hi 21:15 karlnapf: received email? ;) 21:15 blackburn, no not yet, but just turned on computer 21:16 have been vomiting all day :( 21:16 are you ok? 21:16 not really 21:16 -!- vikram360 [~vikram360@117.192.162.41] has joined #shogun 21:16 I hope you will recover soon 21:17 blackburn, thanks 21:17 I can recall last year Soeren disappeared on the announcement too hehe 21:17 karlnapf: yeah, take care ;) 21:17 n4nd0, thanks 21:18 It might a common thing, because my mentor disappeared too 21:18 I only got a mail fro carol 21:18 karlnapf: which mail? 21:19 Hi GSoC Students, 21:19 As with every year around this time, we are about to add new students to this mailing list in a.... 21:19 ahaha 21:19 yeah 21:19 the melange site seems to be down :) 21:19 yes tends to 21:19 774 loaded and no shogun o_o 21:20 the first 500 does not contain shogun related proj 21:20 PhilTillet: :D 21:20 no shogun at all! 21:20 lol 21:20 Yeah 21:20 well there were 4k applications 21:20 i don't know how many of them are going to get accepted 21:21 PhilTillet: you can load quite a lot man, are you using GPU in the browser too? :P 21:21 wiking: ~1.1K 21:21 n4nd0, XD 21:21 but it stopped at 774 21:21 I have a different case all together, cant even open melange site :( 21:21 no more loading 21:21 harshit_: I am mostly there too 21:22 which site are you all trying to load? 21:22 http://www.google-melange.com/gsoc/projects/list/google/gsoc2012 21:22 build #516 of python_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/516 21:22 ^ 21:22 it should be a nice DDoS attack - to tell gsoc students accepted project are announced on site X 21:24 lol 21:25 :D 21:25 GSoC accepted projects are on universalmusic.com 21:25 :> 21:25 lol 21:25 once again, stopped at 774 21:25 and no shogun! 21:26 I have 100 still 21:26 I'm trying too, for you guys :) 21:26 PhilTillet: i guess they are inserting lines by lines 21:26 -!- puffin444 [230bf329@gateway/web/freenode/ip.35.11.243.41] has joined #shogun 21:27 wiking, well, seems like i've loaded everything though 21:27 PhilTillet: yeah 774 and it gets stuck 21:27 I think we'll have to wait some additional time 21:27 lets have some vodka instead 21:27 who's in? 21:28 lol, not me :) 21:28 I don't drink :p 21:28 PhilTillet: bad way 21:28 :) 21:29 Well this is quite suspensful 21:30 it's beyond suspens 21:31 ^^ 21:31 hey guys, i will check back later in an hour or so 21:31 Yeah it's sort of torturous :) 21:31 I wonder how google updates these. Do you think my dashboard/email would tell me first? 21:32 puffin444: IIRC that list is first 21:32 774 is the magic number :D 21:33 400 here 21:33 again 774 21:33 :p 21:33 How many in total is it this year? 1000? 21:33 774 + 9 21:33 XD 21:33 :)) 21:33 puffin444: should be ~1.1K again I think 21:34 774 , no shogun :( 21:34 Yep, I'm stuck at 774 too 21:34 I suspect they haven't released all of them at the same time for some reason. 21:35 They did the same thing with the sponsoring organizations 21:35 Some didn't pop up until days after the release data :o 21:35 puffin444: no with orgs the reason was different 21:35 organizations had to complete their profiles 21:35 I think it's just a bug of theirs 21:35 not a bug 21:36 a feature? 21:36 with orgs - yes 21:36 i was talking about the students :p 21:37 as for students no idea what is going on hehe 21:37 maybe it's a feature too 21:37 :p 21:37 786! 21:37 the number is evolving 21:38 XD 21:38 terribly slow thing 21:39 Did anyone receive that "none" email a few days back? 21:39 I didn't 21:39 I didn't either 21:39 I did 21:40 It was a positive one 21:40 811 loaded 21:40 no shogun ;) 21:40 haha blackburn, new record! 21:40 i'm refreshing again 21:40 maybe i'll get a better score 21:40 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: Lost terminal] 21:40 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun 21:41 emrecelikten: what do you mean be "positive"? 21:41 *by* 21:41 814 21:42 :) 21:42 827 :p 21:42 no shogun though 21:42 Do you know how many slots shogun had this year? 21:42 puffin444: There were two types of emails as far as I know. The "thank you" ones that corresponded to negative ones and the "congratulations" ones. Some guy claimed that he got "pick one of the projects" email 21:42 836 21:43 This happened a few days ago? 21:44 puffin444: Yes, Friday I think 21:44 846 21:44 855 21:45 :p 21:45 hehe it becomes faster 21:45 It's like an auction 21:45 http://google-opensource.blogspot.de/2012/04/students-announced-for-google-summer-of.html 21:45 look 1212 is the number 21:45 yep 21:45 ok so 1212 :> 21:45 big # 21:46 I just got a rejection email from another organization 21:46 fyi: 5 500 * 1 212 = 6 666 000 21:46 ;) 21:46 943 21:47 funny number... i wonder whether it was intentional ;) 21:47 no shogun 21:47 :p 21:47 wiking: in russia it would be 300-400 - you have to stole 60% :D 21:47 950 here 21:47 No shogun 21:47 blackburn: :DDD 21:48 955 21:48 982 :p 21:48 You guys seems to be the last :P 21:48 978 21:48 * wiking refreshing :) 21:49 Maybe I'll never see the decision. I'll just die from a heart attack first :P 21:49 same for me 21:49 This is cruel 21:49 emrecelikten: yeah it is 21:50 999! 21:50 no shogun 21:50 :p 21:50 PhilTillet: Same 21:50 1000 , no sign of shogun :( 21:50 1000 21:50 we are elite 21:50 XD 21:51 ok 21:51 they are just at R 21:51 so soon 21:51 1007 21:51 how do you know? XD 21:51 i've got some T loaded 21:51 and it will be hell cause Shogun's list will progressively appear 21:51 :p 21:51 PhilTillet: yeah 21:51 PhilTillet: but then now suddenly between 900-1000 21:51 new P and R projects got in 21:51 1049 21:51 ;) 21:51 no shogun 21:51 n4nd0: You're in 21:52 If you're Fernando Garcia 21:52 emrecelikten: really? 21:52 yeeey 21:52 Yep 21:52 hehe congrats 21:52 Congrats! 21:52 congrats n4nd0 21:52 pluskid in 21:52 well Iglesias should be in between 21:52 nice :), congrats 21:52 Sergey Lisitsyn in 21:52 hehe 21:52 Michal in 21:52 :D 21:52 plsukid is in 21:52 I wan to see that guys! 21:52 Jacob Walker in 21:52 YES 21:52 vigsterkr in 21:52 \o/ 21:52 Oh wow! 21:52 still 1024 here 21:53 sergey in 21:53 Just filter it ;) 21:53 maaaail 21:53 got a mail :D 21:53 fuck I was not selected for multiclass :O 21:53 just got the mail :) 21:53 :P 21:53 hehe got the mail 21:53 n4nd0: heheheh i'm more happy that you'll do that 21:54 GOT SO! 21:54 Congratulations to you guys! 21:54 got a negative mail 21:54 :( 21:54 PhilTillet: I am sorry 21:54 -ve :( 21:54 PhilTillet: :(((( 21:54 PhilTillet: Sorry PhilTillet :( :( 21:54 PhilTillet: Next year mate :) I was rejected last year too, I know how it feels 21:54 PhilTillet: sorry man 21:54 emrecelikten, I was rejected last year too :p 21:54 n4nd0: when are you going to start the SO? 21:55 PhilTillet: I was being rejected for two years :P 21:55 n4nd0: asap :) 21:55 ups, again talking to me 21:55 haha 21:55 n4nd0: hahahah 21:55 n4nd0: i'll need it 21:55 asap :D 21:56 wiking: cool, we can probably exchange some thoughts about it :) 21:56 yeah sure 21:56 are you going to use libqp? 21:56 PhilTillet: okay actually we are just not ready for gpuing.. 21:56 wiking: yes, that was the first option I thought of 21:57 aha 21:57 so then you are the third one 21:57 blackburn, why so? 21:57 since i'm going to use it and michal as well 21:57 harshit_: I am very sorry - the problem is that Soeren is not capable to mentor two students 21:57 so i've started to talk with sonney2k about how to refactor some stuff 21:57 PhilTillet: no real mentor for that 21:57 and no base for that actually 21:57 n4nd0: i'll push this week something so your comments are more than welcome 21:58 hmmm ok 21:58 wiking: thank you 21:58 blackburn, can understand :| 21:58 I really hope you are not upset with anything 21:58 wiking: I'll start to be much more active again after Saturday 21:58 damned exams 21:58 blackburn, I'm not ^^ 21:58 n4nd0: no worries 21:58 PhilTillet: that's nice :) 21:59 PhilTillet: we are still welcome your contributions 22:00 harshit_: your contributions are welcome too I hope you are not very frustrated! 22:01 blackburn, I'll probably do the project on my own :) that's why I sent only one application, there's no point in working in something i'm not fully interested in ; 22:02 :) 22:02 hey, congrats guys 22:02 blackburn, no, actually a lil. but still can work on NN for shogun , and my research 22:02 PhilTillet, harshit_ let me explain a bit 22:03 you all have done really a lot of work 22:03 blackburn, but summer could be better with gsoc 22:03 I am/was the only mentor for your projects and the one we selected gsomix for 22:03 I would have loved to take you both too but I simply cannot handle >1 student 22:04 so we decided for the task we needed most - and that is some restructuring / accessibility cleanups 22:04 there was nothing wrong with your proposal or involvement or anything 22:04 sonney2k, we know :) 22:05 it really is about being at our capacity limits 22:05 sonney2k, nothing prevents us from contributing without being paid I guess :) 22:05 sonney2k, no worries :) 22:05 and there is nothing I could do about it so late 22:05 in the end I *never* expected that we get 9 slots 22:05 if I knew that beforehand I would have asked a lot more people :) 22:06 :p 22:06 anyways whole wikimedia foundation has 8 (or 9?) slots 22:06 Wow 22:06 so just to give you some impression how many we received 22:06 we keep receiving strong gsoc support :) 22:07 PhilTillet, harshit_ of course we welcome your contributions 22:07 sonney2k, sure ..! 22:07 karlnapf, do you have a homepage I can link to 22:08 puffin444, ^ ? 22:08 and does any of the students here blog? I can put you on shogun planet if you want... 22:09 sonney2k: Not at the current moment. I might be able to make one if you need one though. 22:09 sonney2k: do you think introduction @ mailing list is a good idea? 22:09 not required - I just wanted to give it some traffic - hehe 22:09 blackburn, I think so 22:09 sonney2k: I want to take advantage of gsoc and start doing it 22:09 hey guys 22:10 blackburn, I totally forgot how we did it last year 22:10 sonney2k: just some introduction mail nothing special I think 22:10 blackburn: every student had to send some short email introducing himself + project right? 22:10 yeap 22:10 n4nd0: puffin444: wiking: karlnapf: gsomix: could you please do this ^? 22:10 karlnapf: however you are well-known 22:10 but still 22:11 yes there is a need to introduce your project 22:11 you can look up in the mailinglist archive how these things look like 22:11 all right 22:11 jsut some short introduction 22:11 blackburn: Yeah sure. 22:11 the other more annoying job is to send a weekly update (from the official beginning to end of gsoc) 22:11 and some feature plan would be nice 22:11 so another short email 22:12 *short* 22:12 cannot be said often enough 22:12 attach pics! 22:12 :) 22:12 blackburn: on shogun mailing list? 22:13 wiking: right 22:13 yeah 22:13 alrighty i'll do it tomo morning 22:13 wiking, thanks 22:13 nw 22:13 sonney2k, yes I just created a hp, but its not completely ready, will send you the link this week 22:13 blackburn, I will send an introducionary mail :) 22:14 karlnapf, your old blog is dead too it seems 22:14 sonney2k: do you understand things will go even more crazy next days? 22:14 I'll send too 22:14 blackburn, why should they? 22:14 sonney2k: once active development begins it would become crazy I am sure :) 22:14 sonney2k, yes i took it offline to replace it by something nicer 22:15 * wiking wonders about the average number of users on #shogun after a month ;) 22:15 :D 22:15 wiking, let me guess 8 :D 22:15 sonney2k: + the bots! 22:15 n4nd0: your task is really important and you will definitely collaborate with wiking and michal! 22:16 blackburn: sounds very exciting 22:16 yeah i'll most def will need a SO asap :>> 22:16 n4nd0, that is why you were chosen ... we have high trust in you 22:16 n4nd0: SO WORK!!! 22:16 wiking: :D 22:16 sonney2k: thank you!! 22:16 however I have high trust in pluskid too ;) 22:16 n4nd0: but anyhow when u start do define the new labels class let me know 22:17 wiking: ok 22:17 well actually we know that you all can do the task - that is why you got selected ;) 22:17 okay congrats you guys 22:18 I have to get back to my shity deals 22:18 blackburn: congrats to you too man 22:18 blackburn, vodka? 22:18 dealer? 22:18 sonney2k: not yet 22:18 sonney2k: yes I deal vodka 22:18 blackburn: :DDD what's your poison ?:) 22:18 vodka blackburn the true soul of shogun :D 22:19 wiking: my poison is water 22:19 :D 22:20 blackburn, prostavlyaisya'. :] 22:20 gsomix: you should 22:20 hehe 22:20 why? 22:20 So just to clarify: An introductory email about me and my project tomorrow on the Shogun Mailing List, and possibly a link to a homepage? 22:21 blackburn, you should in the past year.  so where? :] 22:21 puffin444: yes 22:21 gsomix: heh ok 22:22 ok. 22:22 okay guys congrats again and see you later 22:23 I should review accepted projects. 22:23 it's very interesting 22:23 all 1212? 22:23 will take a night 22:23 -!- blackburn [~qdrgsm@188.168.2.179] has quit [Quit: Leaving.] 22:23 blackburn, aha 22:23 wow, Category Theory Module' for SymPy. 22:27 gsomix: Nice 22:29 see you later guys 22:29 n4nd0: Bye! 22:29 build #731 of cmdline_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cmdline_static/builds/731  blamelist: sonne@debian.org, gsomix@gmail.com 22:29 :( 22:30 ok guys, you'll be hearing from me tomorrow. See you later. 22:32 shogun-buildbot, aww, I forget about it. 22:32 build #709 of r_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/709  blamelist: sonne@debian.org, gsomix@gmail.com 22:33 -!- puffin444 [230bf329@gateway/web/freenode/ip.35.11.243.41] has left #shogun [] 22:34 -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 246 seconds] 22:34 shogun-buildbot what's up :) 22:36 wiking, my bag 22:36 build #710 of octave_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/710  blamelist: sonne@debian.org, gsomix@gmail.com 22:37 gsomix, what goes wrong? 22:41 build #697 of python_static is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/697  blamelist: sonne@debian.org, gsomix@gmail.com 22:41 sonney2k, I forget about libshogun tests. 22:42 gsomix, these are not even tests but just examples 22:42 but I do not know what with classifier_multiclasslinearmachine 22:42 sonney2k, aha 22:42 they are supposed to be run as tests later but we didn't do it yet 22:43 -!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] 22:54 sonney2k, I corrected conjugate example. check github 22:58 -!- harshit_ [~harshit@182.68.43.52] has quit [Ping timeout: 244 seconds] 22:58 -!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has left #shogun ["Leaving"] 22:58 shogun: Evgeniy Andreev master * r51e0223 / examples/undocumented/libshogun/classifier_conjugateindex.cpp : small fix in example - http://git.io/Ci5UwQ 22:59 shogun: Soeren Sonnenburg master * r6ab1607 / examples/undocumented/libshogun/classifier_conjugateindex.cpp : 22:59 shogun: Merge pull request #483 from gsomix/to_multiclass 22:59 shogun: Small fix in example - http://git.io/Y19esw 22:59 sonney2k, I need to go. Thanks a lot for your choose. 23:01 good night guys 23:01 -!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun 23:04 -!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] 23:04 -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun 23:04 -!- harshit_ [~harshit@182.68.43.52] has joined #shogun 23:10 -!- harshit_ [~harshit@182.68.43.52] has quit [Client Quit] 23:15 nite everyone 23:19 -!- karlnapf [~heiko@host86-180-223-101.range86-180.btcentralplus.com] has quit [Quit: Leaving.] 23:26 -!- av3ngr [~av3ngr@106.70.3.222] has joined #shogun 23:50 -!- nickon [~noneedtok@kotnet-147.kulnet.kuleuven.be] has quit [Quit: ( www.nnscript.com :: NoNameScript 4.22 :: www.esnation.com )] 23:54 --- Log closed Tue Apr 24 00:00:19 2012