IRC logs of #shogun for Monday, 2013-11-25

--- Log opened Mon Nov 25 00:00:39 2013
-!- thoralf [~thoralf@91-66-33-4-dynip.superkabel.de] has quit [Quit: Konversation terminated!]00:12
shogun-buildbot_build #628 of nightly_default is complete: Failure [failed notebooks]  Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/62804:19
-!- sonne|osx [~sonne@89.204.153.54] has joined #shogun08:08
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun08:10
shogun-notifier-shogun: Thoralf Klein :develop * 7d36d76 / src/shogun/base/SGRefObject.cpp,src/shogun/base/SGRefObject.h: https://github.com/shogun-toolbox/shogun/commit/7d36d76abdfb1cc81d139c082131f14d23c155d608:10
shogun-notifier-shogun: Introduced SGRefObject (pulling refcounting out of CSGObject)08:10
shogun-notifier-shogun: Thoralf Klein :develop * f9f46bf / src/shogun/base/SGObject.cpp,src/shogun/base/SGObject.h: https://github.com/shogun-toolbox/shogun/commit/f9f46bf0b4b7ff093664371f1d3153f3ed01fd3c08:10
shogun-notifier-shogun: Putting refcounting from SGObject to new super class SGRefObject.08:10
shogun-notifier-shogun: Thoralf Klein :develop * 9893df5 / examples/undocumented/libshogun/library_gc_array.cpp: https://github.com/shogun-toolbox/shogun/commit/9893df5b2423b330782f0a9b02f335277b677c4008:10
shogun-notifier-shogun: Fix libshogun example (preventing segfault when shogun_exit() is called before08:10
shogun-notifier-shogun: kernel objects are destroyed)08:10
shogun-notifier-shogun: Soeren Sonnenburg :develop * 4412de8 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/4412de87b2d5921e1bf38e283c51cc2a2ff170c408:10
shogun-notifier-shogun: Merge pull request #1771 from tklein23/csgobject_refactoring08:10
shogun-notifier-shogun:08:10
shogun-notifier-shogun: CSGObject refactoring: Introduced SGRefObject08:10
shogun-notifier-shogun: Soeren Sonnenburg :develop * 8b4597d / NEWS: https://github.com/shogun-toolbox/shogun/commit/8b4597d5592547961e130073ffea0aa9c8bad8f208:19
shogun-notifier-shogun: update news08:19
-!- sonne|osx [~sonne@89.204.153.54] has quit [Quit: sonne|osx]08:26
-!- travis-ci [~travis-ci@ec2-54-226-108-81.compute-1.amazonaws.com] has joined #shogun08:55
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1447697508:55
-!- travis-ci [~travis-ci@ec2-54-226-108-81.compute-1.amazonaws.com] has left #shogun []08:55
-!- travis-ci [~travis-ci@ec2-54-242-253-145.compute-1.amazonaws.com] has joined #shogun09:26
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1447718709:26
-!- travis-ci [~travis-ci@ec2-54-242-253-145.compute-1.amazonaws.com] has left #shogun []09:26
shogun-buildbot_build #136 of clang34 - undefined behaviour analysis is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/136  blamelist: Thoralf Klein <thoralf.klein@zib.de>09:44
shogun-buildbot_build #1978 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/1978  blamelist: Soeren Sonnenburg <sonne@debian.org>09:47
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has joined #shogun10:08
shogun-buildbot_build #134 of clang34 - thread analysis is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/134  blamelist: Thoralf Klein <thoralf.klein@zib.de>10:09
-!- thoralf [~thoralf@enki.zib.de] has joined #shogun10:45
thoralfHo.10:45
shogun-buildbot_build #2053 of deb3 - modular_interfaces is complete: Failure [failed test csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2053  blamelist: Soeren Sonnenburg <sonne@debian.org>, Thoralf Klein <thoralf.klein@zib.de>10:48
shogun-buildbot_build #153 of clang34 - static analysis is complete: Failure [failed analyse]  Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/153  blamelist: Thoralf Klein <thoralf.klein@zib.de>10:52
sonne|workhey thoralf11:02
thoralfsonne|work: Seems my first patch works fine.  Now talking about the next one?11:04
sonne|workthoralf: yes sure11:04
thoralfhttps://github.com/shogun-toolbox/shogun/pull/177211:04
sonne|workI think we need DynamicRefObjectArray ...11:05
sonne|workotherwise you have the same issues again with you not knowing when to ref / unref11:05
* wiking is writing a new qp library11:12
sonne|workwiking: hurray another one :D11:13
@wikingdelibrately not making part of shogun so that ppl can use it externally11:13
@wikingbut the thing is that it'll be lgpl311:13
sonne|workwiking: libqp11:13
@wikingand will do a completely subtitute in sense of qp11:13
@wikingof mosek11:13
@wikingso finally we can do mosek like qp for structural learners11:14
@wikingsonne|work: libqp is almost good11:14
@wikingbut it's too strickt with the s.t. of the minimization11:14
@wiking*strict11:15
@wikingand another thing11:15
@wikingit's gonna have a general KKT solver using various factorization methods11:15
@wikingso if we manage to get this into shogun11:15
@wikingwe might consider to exchange the kkt solving in other qps11:16
@wikingusing these ones11:16
@wikingas they'll be superoptimized11:16
@wikingusing eigen11:16
@wikingand openmp11:16
@wikingbasically what i'm trying to implement is a quadrating cone program11:17
@wikingso it'll solve a pair of primal and dual quadratic cone programs11:18
thoralfwiking: Ehr.  Many people from my institute are developing for LP/QP/fooP-solver.  Sure you wanna start from scratch?11:18
@wikingthe idea came from CVXOPT11:18
sonne|workwiking: that is a massive effort... not sure you have the time...11:18
@wikingthoralf: ^11:18
@wikingsonne|work: well about kkt interfacing i'm sure that i dont have time for that11:19
thoralfwiking: fyi: http://www.zib.de/en/optimization/software.html11:19
@wikingi just want to give an eigen like (hpp only implementation) for KKT solving11:19
@wikingthoralf: license?11:19
@wikingor all gpl3 valid11:19
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]11:19
thoralfhttp://scip.zib.de/licence.shtml11:20
thoralfOne of them (ZIBopt)11:20
@wikingthoralf: yeah it's almost like svmlight license :S11:20
thoralfwiking: If you have concerns about the license, let me talk to my boss. ;)11:21
@wikingthoralf: yeah we already have issues with svmlight :S11:21
@wikingand basicaly it's almost like mosek in one way11:21
@wikingsonne|work: hehehe remember when u told me this "sonne|work> wiking: that is a massive effort..." about cmake? :)11:21
sonne|workwiking: yes and we are still not at the stage where cmake is working without a hassle11:22
thoralflol11:22
@wikingsonne|work: sorry :S11:22
thoralfThe Gallileo gambit.11:22
@wikingfor trying my best11:22
@wiking:P11:22
@wikingthoralf: http://pystruct.github.io/index.html11:22
@wikingu know this one... ?11:22
sonne|workwiking: so just dont' underestimate the work11:23
@wikingpretty cool about ssvm11:23
@wikingsonne|work: yeah sure note11:23
@wiking*not11:23
@wikingsonne|work: but we are getting there with cmake i think11:23
@wikingthoralf: btW: big up for the separation for structuralData from sgobjc11:23
thoralfwiking: Yeah, heard of, but Python is painful when you're having huge data... :(11:24
thoralfwiking: Even thought of using it.11:24
@wikingsonne|work: if i manage to get that KKT shit working and when we get gsoc slots it could be a good idea to dedicate a student to do that kkt interfacing for other qp solvers11:24
@wikingthoralf: sure, this is why i started this whole shit ... wanna do it in c++11:24
@wikingeigen style11:24
@wikingor tapkee style11:24
lisitsyn1hah11:25
lisitsyn1tapkee style come on yo11:25
@wikinglisitsyn1: heheh yeah if u r around i have questions11:25
lisitsyn1wiking: sure11:25
-!- lisitsyn [~lisitsyn@80.252.20.67] has quit [Disconnected by services]11:25
-!- lisitsyn1 is now known as lisitsyn11:25
@wikinglisitsyn1: do u know if for example in eigen the llt solver (cholesky factorization)11:25
@wikingif i give it a sparsematrix11:26
@wikingwill it use the sparse cholesky11:26
@wikingor it will use the dense one?11:26
lisitsynwiking: don't remember by heart actually11:26
lisitsynbut I tend to think it is dense11:26
@wikingas i've tested that it works11:26
thoralfsonne|work: Well, about the DynamicRefObjectArray thing: I already spend too much time on this issue.  I did got a working patch - and did even a refactoring of CSGObject.11:26
@wikingEigen::LLT<MatrixType> lltOfK11:26
@wikingwhere MatrixType = Sparse11:26
@wikingworks as well11:26
@wikingbut dunno if it's actually calling the sparse one11:27
-!- lisitsyn1 [~lisitsyn@80.252.20.67] has joined #shogun11:27
thoralfsonne|work: To make it short: I won't do this DynamicRefObjectArray class... but I could adapt my patch if you could provide it.11:27
@wikingor just does the dense on sparse matrix11:27
-!- mode/#shogun [+o lisitsyn] by ChanServ11:27
@wikingif so11:27
-!- lisitsyn1 was kicked from #shogun by lisitsyn [lisitsyn1]11:27
@wikinghow would u do this then11:27
@wikingsay template<typename MatrixType> class KKTCholeskySolver11:27
@lisitsynwiking: there is SimplicialLLT or so11:27
@wikinglisitsyn: yeah i know11:27
@wikingbut it would depend on the MatrixType which one to use11:27
@lisitsynthat works with sparse quite ok11:27
@lisitsyndo you want it generic?11:28
@wikingyes11:28
@wikingi want it to be able to do it both on sparse and dense11:28
sonne|workwiking: you were working on cmake since june this year ... and your estimate was 2 weeks max11:28
@lisitsynwiking: well template specialization ;)11:28
@wikingsonne|work: well i've got cmake working in 2 weeks no? :))11:28
@wikinglisitsyn: ok so that's the only way right?11:28
sonne|workwiking: lol11:28
@wikingi thought so....11:28
@wikingsonne|work: well we are really now better off with this than the ./configure i'd say11:29
@lisitsynwiking: you can actually support not only eigen but any other stuff11:29
@wikingbut we can always switch back to ./configure ;PP11:29
@wikinglisitsyn: yeah i dont want :D11:29
@wikinglisitsyn: just fucking have eigen11:29
@wikingit's really easy to have installed on a toaster even :PP11:29
@lisitsynwiking: with sfinae you could check if method supports some operator() :D11:29
@lisitsynplug n play haha11:29
@wikinghahaha p'n'p11:30
@wikingwhat was that shit about in the 90s11:30
@wiking:DD11:30
@lisitsynwiking: yeah I still remember even though I was like 5 in the middle of 90s haha11:30
@wikinglisitsyn: woah http://www.boost.org/doc/libs/1_55_0/libs/utility/enable_if.html11:31
@lisitsynwiking: yeah that's easy11:31
@wikinglisitsyn: do we have something similar in stdc?11:31
@lisitsynwiking: no, I don't think yet11:31
@lisitsynwiking: I usually rewrite sfinae haha11:32
@wikinghahahah11:32
@lisitsynbecause boost slows down compilation way too much11:32
@lisitsynand my autocompletion goes slower :D11:32
@wikingyeps11:32
@wiking:D11:32
@wikingso true11:32
@wiking:>11:32
@wikingoh woah11:33
@wikinghttp://nonchalantlytyped.net/blog/2012/06/27/yet-another-sfinae/11:33
@lisitsynwiking: yes canonical sfinae11:33
@lisitsynah11:33
@wikingok so actually it's doable11:33
@lisitsynone more sfinae11:33
@lisitsyn:D11:33
@wikingyes11:34
@lisitsynwiking: well yeah if you want to check if it is sparse it is easy11:34
@wikingcool i'll do it then laater11:35
@wikingfirst i want to have it ready for initial commit that it actually works11:35
@wikingand then i'll do writing11:35
@lisitsynwiking: if you want to choose solver you'd have to have two parts11:35
@lisitsyncompile time part which selects sparse/dense11:35
@lisitsynand runtime part which selects proper solver11:35
@wikingmmmhmmm11:35
@wikingwell we'll see how far i can get with it11:35
@wikingand then if u have time for comments i'll give u a link11:36
@wikingto check on the shit11:36
@lisitsynwiking: sure11:36
@wikingoooh maaan11:37
@wikinglisitsyn: this is really daily wtf11:37
@wiking11:34 < lmatteis> i have an Microsoft Access file database11:38
@wiking11:34 < lmatteis> what's the easiest way to query it using SPARQL11:38
@wikingthe first line11:38
@wikingMicrosoft Access file database11:38
@lisitsynwiking: haha11:38
@lisitsynwiking: I actually tried another approach11:39
@lisitsynlike11:39
@lisitsyntemplate <typename MatrixType, typename SolverType>11:39
@wikingyeah i thought o fthat11:39
@wikingbut what if the idiot user (see academic 99%)11:40
@wikinggives u11:40
@wikingDenseMatrix, SparseSolver11:40
@wiking:)))11:40
@wikingof course i can do template specialization11:40
@wikingbut i dont want11:40
@lisitsynwiking: you fail compile time then11:41
@wikinglisitsyn: heheh11:42
@lisitsynwiking: it is ideal to see no <> so you could use that builder like syntax11:43
@lisitsynto approach with inference11:43
@lisitsynso you do solve(matrix).using(LLT)11:44
@lisitsynagain compile-time failure when doing something dumb11:44
@lisitsynwiking: if you want to have fun of users I have one thing I didn't implement yet11:45
@lisitsynwiking: solve[method=LLT,matrix=X]11:45
@lisitsyn:D11:45
@lisitsynlegal C++11:45
shogun-buildbot_build #2054 of deb3 - modular_interfaces is complete: Failure [failed test csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2054  blamelist: Soeren Sonnenburg <sonne@debian.org>12:07
-!- sukalpo [8622b0e0@gateway/web/freenode/ip.134.34.176.224] has joined #shogun12:29
sukalpoHello...could someone refer me any proper documentation to implement kernel Caching for MKL ..I  am about to code my own MKL (wrapper approach )  ...the only issue I cannot figure out is how to deal with kernel matrices bigger than my memory...... I have all the other required algorithms though....Can anyone suggest me something12:31
-!- besser82 [~besser82@2a02:8108:8840:1800:e8b:fdff:fe16:bb33] has joined #shogun12:34
-!- besser82 [~besser82@2a02:8108:8840:1800:e8b:fdff:fe16:bb33] has quit [Changing host]12:34
-!- besser82 [~besser82@fedora/besser82] has joined #shogun12:34
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun12:59
-!- mode/#shogun [+o iglesiasg] by ChanServ12:59
-!- new_lido [~walid@41.218.174.238] has joined #shogun13:33
sonne|worksukalpo: just use a kernel from shogun and let the svms use their kernel cache13:55
sukalpoActually I would like to use my optimization routines and stuff like that14:10
sonne|worksukalpo: for solving the inner svm problem?14:16
sukalpoyes ....14:17
sukalpofor the outer loop as well14:17
-!- new_lido is now known as new_lido_bsy14:53
sonne|worksukalpo: well then you have to do the caching your own of course14:56
sukalpocould you suggest me any well documented algorithm?14:58
-!- new_lido_bsy [~walid@41.218.174.238] has quit [Ping timeout: 272 seconds]15:03
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Ex-Chat]15:14
-!- sukalpo [8622b0e0@gateway/web/freenode/ip.134.34.176.224] has quit [Quit: Page closed]15:45
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:08
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!]16:55
-!- thoralf [~thoralf@zib-52-158.zib.de] has joined #shogun17:21
thoralfHeyhey.17:21
-!- sonne|osx [~sonne@89.204.137.159] has joined #shogun17:32
-!- sonne|osx [~sonne@89.204.137.159] has quit [Quit: sonne|osx]17:47
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun17:47
shogun-notifier-shogun: Soeren Sonnenburg :develop * ecb17dd / src/interfaces/modular/SGBase.i,src/shogun/base/SGDynamicRefObjectArray.h: https://github.com/shogun-toolbox/shogun/commit/ecb17ddd8852359cb8ab2245eaea3929b1f82b1f17:47
shogun-notifier-shogun: - add dynamic array for SGRefObject's17:47
shogun-notifier-shogun: - make SGRefObject the class that is swig ref/unref'd17:47
-!- sonne|osx [~sonne@89.204.137.159] has joined #shogun17:48
sonne|osxthoralf: ok there you go!17:48
sonne|osxthoralf: base/SGDynamicRefObjectArray.h17:48
-!- sonne|osx [~sonne@89.204.137.159] has quit [Client Quit]17:49
shogun-buildbot_build #1979 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/197917:56
-!- lisitsyn [~lisitsyn@80.252.20.67] has joined #shogun18:13
thoralfHey Sergey18:17
-!- thoralf [~thoralf@zib-52-158.zib.de] has quit [Quit: Konversation terminated!]18:40
shogun-buildbot_build #2055 of deb3 - modular_interfaces is complete: Failure [failed test python modular test ruby modular test java modular test csharp modular test lua modular test octave modular test r modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2055  blamelist: Soeren Sonnenburg <sonne@debian.org>18:53
-!- travis-ci [~travis-ci@ec2-54-237-127-137.compute-1.amazonaws.com] has joined #shogun19:09
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1450036719:09
-!- travis-ci [~travis-ci@ec2-54-237-127-137.compute-1.amazonaws.com] has left #shogun []19:09
lisitsynah missed thoralf19:46
-!- new_lido [~walid@41.218.174.238] has joined #shogun19:55
-!- sonne|osx [~sonne@f053040204.adsl.alicedsl.de] has joined #shogun20:03
shogun-buildbot_build #308 of osx1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx1%20-%20libshogun/builds/308  blamelist: Soeren Sonnenburg <sonne@debian.org>, Thoralf Klein <thoralf.klein@zib.de>20:18
-!- new_lido_ [~walid@41.218.174.238] has joined #shogun20:29
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]20:47
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun21:18
shogun-notifier-shogun: Soeren Sonnenburg :develop * 9caa249 / src/ (3 files): https://github.com/shogun-toolbox/shogun/commit/9caa249776eb00d7d0b33fd59ae2291c60ddf67d21:18
shogun-notifier-shogun: fix typo and documentation of SGRefObject21:18
shogun-buildbot_build #357 of FC19 - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/357  blamelist: Soeren Sonnenburg <sonne@debian.org>21:35
shogun-buildbot_build #351 of FCRH - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/351  blamelist: Soeren Sonnenburg <sonne@debian.org>21:35
-!- new_lido_ [~walid@41.218.174.238] has quit [Quit: Leaving]21:40
-!- new_lido_ [~walid@41.218.174.238] has joined #shogun21:40
shogun-buildbot_build #309 of osx1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx1%20-%20libshogun/builds/30921:53
shogun-buildbot_build #2056 of deb3 - modular_interfaces is complete: Failure [failed examples and unit tests]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2056  blamelist: Soeren Sonnenburg <sonne@debian.org>22:23
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun22:38
-!- mode/#shogun [+o iglesiasg] by ChanServ22:38
@iglesiasggood evening guys22:39
lisitsynhey iglesiasg22:40
-!- travis-ci [~travis-ci@ec2-54-237-127-137.compute-1.amazonaws.com] has joined #shogun22:54
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1451046522:54
-!- travis-ci [~travis-ci@ec2-54-237-127-137.compute-1.amazonaws.com] has left #shogun []22:54
-!- thoralf [~thoralf@91-66-33-4-dynip.superkabel.de] has joined #shogun23:00
thoralfHey.23:00
thoralfsonney2k: Just fixing some compile errors... thanks for adding the class.23:01
-!- thoralf [~thoralf@91-66-33-4-dynip.superkabel.de] has quit [Ping timeout: 245 seconds]23:22
--- Log closed Tue Nov 26 00:00:40 2013

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!