IRC logs of #shogun for Saturday, 2014-03-08

« Friday, 2014-03-07 Index Sunday, 2014-03-09 »
--- Log opened Sat Mar 08 00:00:29 2014
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun00:06
-!- sonne|osx [~sonne@e178249159.adsl.alicedsl.de] has quit [Quit: sonne|osx]00:15
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds]00:30
-!- vadresh [~vadresh@62.109.10.44] has joined #shogun00:42
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun01:40
-!- abinash_panda [~abinash@1.38.23.53] has quit [Ping timeout: 240 seconds]02:24
shogun-buildbot_build #736 of nightly_default is complete: Failure [failed doc]  Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/73603:40
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Quit: Page closed]04:00
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Ping timeout: 264 seconds]05:47
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun05:48
-!- gif [~gif@130.195.253.57] has joined #shogun05:59
-!- gif [~gif@130.195.253.57] has quit [Remote host closed the connection]05:59
-!- gif [~gif@130.195.253.57] has joined #shogun05:59
gifHey guys ! I still get stuck on making the shogun on my mac os x mavericks laptop06:01
gifafter solving some compiling errors, here is the latest error messages that I got : http://codetidy.com/8341/06:04
gifany idea how to solve this?06:04
-!- gif [~gif@130.195.253.57] has quit []06:11
@sonney2kgif which version is this?06:14
-!- sonne|osx [~sonne@e179065162.adsl.alicedsl.de] has joined #shogun06:22
@sonney2kshogun-buildbot_, force build --branch=develop 'precise - libshogun'06:37
shogun-buildbot_build #477 forced06:37
shogun-buildbot_I'll give a shout when the build finishes06:37
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun06:47
shogun-notifier-shogun: Soeren Sonnenburg :develop * 762c126 / src/shogun/base/SGObject.h: https://github.com/shogun-toolbox/shogun/commit/762c12629477b4d9a4fd93d5cf164f9c8959f1f006:47
shogun-notifier-shogun: fix doc error06:47
shogun-buildbot_build #477 of precise - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/47706:55
-!- kislay [~kislay@93.114.45.13] has quit [Remote host closed the connection]06:59
-!- sonne|osx [~sonne@e179065162.adsl.alicedsl.de] has quit [Quit: sonne|osx]07:00
shogun-buildbot_build #521 of FC19 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/521  blamelist: Soeren Sonnenburg <sonne@debian.org>07:05
shogun-buildbot_build #478 of precise - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/478  blamelist: Soeren Sonnenburg <sonne@debian.org>07:20
-!- Saurabh7 [~Saurabh7@115.248.130.148] has joined #shogun07:32
-!- dhruv13 [~dhruv13@84.64.12.231] has joined #shogun07:55
-!- [1]Saurabh7 [~Saurabh7@public-anonymous-gw.mobilefree.jp] has joined #shogun08:09
-!- Saurabh7 [~Saurabh7@115.248.130.148] has quit [Ping timeout: 264 seconds]08:12
-!- [1]Saurabh7 is now known as Saurabh708:12
-!- kislay [~kislay@93.114.45.13] has joined #shogun08:14
-!- lisitsyn [~lisitsyn@80.252.20.67] has quit [Ping timeout: 264 seconds]08:28
-!- dhruv13 [~dhruv13@84.64.12.231] has quit [Ping timeout: 252 seconds]08:33
-!- [1]Saurabh7 [~Saurabh7@115.248.130.148] has joined #shogun08:59
-!- Saurabh7 [~Saurabh7@public-anonymous-gw.mobilefree.jp] has quit [Ping timeout: 252 seconds]09:03
-!- [1]Saurabh7 is now known as Saurabh709:03
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun09:36
-!- mode/#shogun [+o iglesiasg] by ChanServ09:36
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]09:47
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun10:25
shogun-notifier-shogun: lambday :develop * 402aeb0 / tests/unit/mathematics/linalg/RationalApproximationIndividualJob_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/402aeb0a4b7ad9c8bce9b6f39cbf823707a5fa4410:25
shogun-notifier-shogun: remove unwanted SG_SDEBUG messages from unit-test10:25
shogun-notifier-shogun: Viktor Gal :develop * 9e01d26 / tests/unit/mathematics/linalg/RationalApproximationIndividualJob_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/9e01d26bf7adf91068ac512bf8f2842d3c5e095910:25
shogun-notifier-shogun: Merge pull request #1960 from lambday/develop10:25
shogun-notifier-shogun:10:25
shogun-notifier-shogun: remove unwanted SG_SDEBUG messages from unit-test10:25
@wikingiglesiasg: yo10:31
@iglesiasgwiking, hey10:31
@wikingiglesiasg: keep the left,right... they are really from the previous implementations10:31
@iglesiasgwiking, all right10:31
shogun-buildbot_build #522 of FC19 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/522  blamelist: lambday <heavensdevil6909@gmail.com>, Viktor Gal <vigsterkr@gmail.com>10:41
shogun-notifier-shogun: pl8787 :develop * 4826bd5 / tests/unit/classifier/svm/LibLinear_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/4826bd5bfb06fd4596c4b6ee02767e8d53b4b0c010:42
shogun-notifier-shogun: Solve memory leak problem, but uninitial values jump problems didn't fixed. I need help~10:42
shogun-notifier-shogun: Viktor Gal :develop * 75ea8d3 / tests/unit/classifier/svm/LibLinear_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/75ea8d3e39239854a3ae549782e1109a21292a6310:42
shogun-notifier-shogun: Merge pull request #1957 from pl8787/develop10:42
shogun-notifier-shogun:10:42
shogun-notifier-shogun: Solve memory leak problem, but uninitial values jump problems didn't fix...10:42
shogun-buildbot_build #479 of precise - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/479  blamelist: lambday <heavensdevil6909@gmail.com>, Viktor Gal <vigsterkr@gmail.com>10:56
Saurabh7iglesiasg, hi10:59
Saurabh7for the kde is there any base class or template i should stick to?11:00
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Ping timeout: 240 seconds]11:19
shogun-buildbot_build #523 of FC19 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/52311:23
shogun-buildbot_build #480 of precise - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/48011:38
-!- dhruv13 [~dhruv13@84.64.12.231] has joined #shogun11:49
@wikingshogun-buildbot_: force build --branch=develop 'debian wheezy - memcheck'11:57
shogun-buildbot_build #211 forced11:57
shogun-buildbot_I'll give a shout when the build finishes11:57
-!- [1]Saurabh7 [~Saurabh7@public-anonymous-gw.mobilefree.jp] has joined #shogun12:24
-!- [1]Saurabh7 [~Saurabh7@public-anonymous-gw.mobilefree.jp] has quit [Client Quit]12:24
-!- Saurabh7 [~Saurabh7@115.248.130.148] has quit [Ping timeout: 240 seconds]12:26
-!- dhruv13 [~dhruv13@84.64.12.231] has quit [Ping timeout: 244 seconds]12:42
-!- sonne|osx [~sonne@e179065162.adsl.alicedsl.de] has joined #shogun12:51
-!- dhruv13 [~dhruv13@84.64.14.38] has joined #shogun12:55
-!- dhruv13 [~dhruv13@84.64.14.38] has quit [Client Quit]12:59
-!- cameo54321 [27297758@gateway/web/freenode/ip.39.41.119.88] has joined #shogun13:01
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun13:01
-!- mode/#shogun [+o iglesiasg] by ChanServ13:01
-!- dhruv13 [~dhruv13@84.64.14.38] has joined #shogun13:04
-!- travis-ci [~travis-ci@ec2-107-20-77-97.compute-1.amazonaws.com] has joined #shogun13:05
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/2033857813:05
-!- travis-ci [~travis-ci@ec2-107-20-77-97.compute-1.amazonaws.com] has left #shogun []13:05
dhruv13iglesiasg: regardig the kd tree implementation, bundling nanoflann is an option because it is a single header file with a BSD licence13:06
@iglesiasgdhruv13, ok13:07
@wikingdhruv13: what are the dependencies of nanoflann?13:07
@iglesiasgdhruv13, it sounds reasonable but what about deps?13:07
@wikingonly stdc++ or some other libraries?13:07
@iglesiasghaha yeah exactly13:07
dhruv13iglesiasg: stdc++13:07
@wikingiglesiasg: :D13:07
@wikingdhruv13: no boost or any other funky library right?13:07
dhruv13nope13:07
@wikingthat's good13:07
dhruv13only, it says that it is optimised for 2d and 3d clouds, though i dont know how exactly13:08
dhruv13still going through it13:08
dhruv13so maybe i should first try to get it to work with KNN, and then play around with it to make it more efficient?13:09
@iglesiasgdhruv13, I guess it is pretty efficient already13:10
@iglesiasgat least I hope so :)13:10
dhruv13iglesiasg: ha ha, no i meant i want to see why they say optimised for 2d and 3d clouds13:10
dhruv13iglesiasg: I'll study some more and get back :-)13:11
@iglesiasgdhruv13, all right, it sounds great!13:12
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer]13:15
-!- sonne|osx [~sonne@e179065162.adsl.alicedsl.de] has quit [Quit: sonne|osx]13:24
dhruv13wiking: about adding environment variable suppot for sgio, would the right place to do it be in the constructor of SGIO objects?13:34
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun13:35
-!- lambday_ [67157e54@gateway/web/freenode/ip.103.21.126.84] has joined #shogun13:36
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds]13:40
-!- lambday_ [67157e54@gateway/web/freenode/ip.103.21.126.84] has quit [Ping timeout: 245 seconds]13:40
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]13:42
-!- HeikoS1 [~heiko@p57972433.dip0.t-ipconnect.de] has joined #shogun13:43
@wikingdhruv13: well i'd rather do it as part of the init fucntion13:45
@wikingsrc/shogun/base/init.cpp13:45
-!- dhruv13 [~dhruv13@84.64.14.38] has quit [Ping timeout: 240 seconds]13:49
shogun-buildbot_build #211 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/21113:51
@wikingHeikoS1: you want to see some scary shit13:51
-!- dhruv13 [~dhruv13@84.64.14.38] has joined #shogun13:52
dhruv13wiking: okay13:52
-!- dhruv13 [~dhruv13@84.64.14.38] has quit [Ping timeout: 264 seconds]13:59
-!- kislay [~kislay@93.114.45.13] has joined #shogun14:02
kislayHeikoS1, I have sent a PR regarding the issue #187814:07
HeikoS1wiking:  yeah :)14:08
HeikoS1give it to me :)14:08
HeikoS1kislay: I will check soon! :)14:08
kislaythanks :)14:08
kislayHeikoS1,  if you can point me over some issue regarding image processing/m-l algorithm which i may take up.!14:10
HeikoS1kislay: maybe send a list to kevin (cc shogun-list) since he has some opencv applications in mind14:11
kislayokay!14:11
@wikingHeikoS1: http://pastebin.com/3AyS7HD214:17
@wikingHeikoS1: eigen solver: CPU vs GPU with magma14:17
@wikingHeikoS1: first column is the matrix size...14:17
@wikingCPU version is pure lapack...14:17
HeikoS1wiking: man!14:17
HeikoS1wiking: is that dense matrices?14:17
HeikoS1wiking: ultra cool! Maybe update the issue with that and make it more sexy?14:18
HeikoS1wiking: this in fact can also be very useful for the deep learning project14:18
@wikingHeikoS1: yeah does are dense matrices14:18
@wikingHeikoS1: heheh deep learning is totally different14:19
@wikingHeikoS1: but it should use some sort of opencl/cuda stuff for sure14:19
HeikoS1wiking: cool!14:19
HeikoS1wiking: man if we would use those solvers for all things in shogun that would be a blast14:19
HeikoS1imagine the nice notebooks comparing performance of say PCA14:20
HeikoS1if one change change the backed in one line14:20
HeikoS1amazing!14:20
@wikingHeikoS1: yeah i'm just trying to refactor a bit the eigensolver of lambday and then change everywhere to it where we do eigen solving...14:22
@wikingHeikoS1: that'll allow us to later have support for GPU/OpenCL based eigensolvers14:23
-!- witness___ [uid10044@gateway/web/irccloud.com/x-qjlutmnzduulfvgk] has joined #shogun14:23
@wikingHeikoS1: check the bottom of this page: https://github.com/bravegag/eigen-magma-benchmark14:27
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun14:27
shogun-notifier-shogun: lambday :develop * aea0036 / src/ (6 files): https://github.com/shogun-toolbox/shogun/commit/aea003659c2096277aba904d3293a078b7c6152814:27
shogun-notifier-shogun: add streaming mmd, refactor linear time mmd14:27
shogun-notifier-shogun: Heiko Strathmann :develop * 9c305cb / src/ (6 files): https://github.com/shogun-toolbox/shogun/commit/9c305cb657cf8c1df8758031ffd091a0b2c9fda414:27
shogun-notifier-shogun: Merge pull request #1959 from lambday/feature/btest14:27
shogun-notifier-shogun:14:27
shogun-notifier-shogun: add streaming mmd, refactor linear time mmd14:27
HeikoS1wiking: thats all great14:27
HeikoS1wiking: what do we need in order to make this work on all eigen operations?14:28
@wikingHeikoS1: well that's really tricky14:28
@wikingHeikoS1: i mean if ever this eigen-magma thing is going to be part of eigen then things are easier14:29
@wikingHeikoS1: but other than that we'll have to do all sorts of wrapping14:29
HeikoS1I see14:29
HeikoS1wiking: ok, maybe wait then, but we can start with the solvers14:29
HeikoS1wiking: besser82 once told me that once can just change one header and everything works out of the box14:30
HeikoS1but I guess that might be a bit too optimistic14:30
@wikingHeikoS1: mmm i dont know about such a library14:30
@wikingHeikoS1: i mean the thing is that there are several different libraries out there... not only eigen14:30
@wikingthat can do some basic math functions14:31
@wikingso this is why better do wrappings14:31
@wikingof these basic functionalities14:31
@wikingso that we can support different libraries... not only eigen14:31
@wikingthe problem is that eigen is spreading all around our code14:31
@wikingwhich is stupid14:31
@wikingbecause then we'll have to do similar thing like now with lapack->eigen14:32
@wikingit'd be better to start creating wrapper classes14:32
@wikingand do there library dependencies14:32
HeikoS1wiking: I see14:32
HeikoS1wiking: mmmh14:32
@wikinginstead of like whole functions like now for PCA or ICA14:32
@wikingthis is just calling for suicide in the long run14:32
HeikoS1wiking: I agree for operations like SVD/EVD14:32
HeikoS1wiking: but you think also for matrix*matrix and matrix*vector multiplication?14:33
HeikoS1wiking: cholesky is another thing14:33
HeikoS1wiking: maybe you are right14:33
@wikingwell there comes all the factorizations as well14:33
@wikingQR, cholensky etc.14:33
HeikoS1wiking: then it would be much easier to change the backend, if everything would be like in linalg14:33
HeikoS1yeah14:33
HeikoS1mmmh14:33
@wikingthere are like n different libraries that does it fast14:33
HeikoS1wiking: entrance task?14:33
@wikingone is opencl optimized14:33
HeikoS1wiking: yeah I totally agree14:33
@wikingthe other is CUDA optimized14:34
@wikingetc.14:34
HeikoS1wiking: thing is that one has to think of interfaces14:34
@wikingyeah these has to be all interfaced14:34
HeikoS1wiking: I did that for some lapack things, and its ugly14:34
@wikinglike in java14:34
HeikoS1all those optiones14:34
@wikingand do separate implementations of them14:34
HeikoS1wiking: what about doing this step by step14:34
@wikingi mean everybody should try to stick to SGMatrix and SGVector14:34
HeikoS1wiking: we can stard with EVD/SVD for now14:34
@wikingand pass aroudn that14:34
HeikoS1start14:35
HeikoS1see how it goes14:35
HeikoS1and then update things one-after another14:35
@wikingand then either it is going to be eigenized or some other funky library used to do that operation14:35
@wikingHeikoS1: yeah SVD is very similar14:35
@wikingthere should be an SVD solver class in shogun14:35
HeikoS1wiking: I agree14:35
@wikingand that should handle all the library dependency etc....14:36
HeikoS1wiking: so what about adding this stuff to the linalg framework?14:36
HeikoS1which was started by lambday?14:36
@wikingand then some global variable can be used to choose SVD engine on the library's level14:36
HeikoS1wiking: could you write a detaled intro task for that, maybe for SVD?14:36
@wikingsame with eigen14:36
@wikingyeah i'll do that14:36
HeikoS1fundamental multiplications can come later14:36
HeikoS1maybe start with a factorization14:36
@wikingyeah that's another story14:36
@wikingbut factorization, svd, eigen solving is really something14:37
@wikingthat should be pluggable14:37
@wiking*plugable14:37
@wikingmeaning if you have a new fast way to do that14:37
@wikingyou'll have to just create a wrapper for that inherited from the base class14:37
@wikingand then that is being used everywhere14:37
@wikingwithin shogun14:38
HeikoS1yeah14:38
HeikoS1great idea14:38
HeikoS1in fact stupid to do anything else ;)14:38
shogun-buildbot_build #524 of FC19 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/524  blamelist: lambday <heavensdevil6909@gmail.com>14:40
shogun-buildbot_build #525 of FC19 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/52514:44
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Quit: Leaving]14:47
shogun-buildbot_build #481 of precise - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/481  blamelist: lambday <heavensdevil6909@gmail.com>14:56
HeikoS1wiking: why is the GPU faster when there are more points. It takes less time, I dont get that15:07
shogun-buildbot_build #482 of precise - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/48215:14
-!- HeikoS1 [~heiko@p57972433.dip0.t-ipconnect.de] has quit [Ping timeout: 240 seconds]15:17
-!- HeikoS [~heiko@p57972433.dip0.t-ipconnect.de] has joined #shogun15:22
-!- mode/#shogun [+o HeikoS] by ChanServ15:22
-!- HeikoS [~heiko@p57972433.dip0.t-ipconnect.de] has quit [Ping timeout: 240 seconds]15:26
-!- travis-ci [~travis-ci@ec2-54-227-19-253.compute-1.amazonaws.com] has joined #shogun15:50
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/2034500015:50
-!- travis-ci [~travis-ci@ec2-54-227-19-253.compute-1.amazonaws.com] has left #shogun []15:50
-!- kjainGSoC [7838c402@gateway/web/freenode/ip.120.56.196.2] has joined #shogun16:04
kjainGSoCHello16:04
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer]16:23
-!- cameo54321 [27297758@gateway/web/freenode/ip.39.41.119.88] has quit [Ping timeout: 245 seconds]16:32
-!- lisitsyn [~lisitsyn@109-226-71-150.clients.tlt.100megabit.ru] has joined #shogun17:15
-!- kjainGSoC [7838c402@gateway/web/freenode/ip.120.56.196.2] has quit [Quit: Page closed]17:19
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]17:27
-!- dhruv13 [~dhruv13@84.64.15.124] has joined #shogun17:38
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun17:45
lambdaywiking: hola!17:45
@wikinglambday: yoyo17:45
@wikinglambday: so i've already done some changes locally17:46
@wikingi'll push soon the first commit to a feature branch17:46
@wikingbut let's finish up our discussion so i can make it right17:46
lambdaywiking: this is getting really exciting17:46
@wikingso about the naming17:46
@wikingdo you maybe have a better naming idea for DirectEigenSolver?17:46
lambdayyeah - you don't like dense?17:46
@wikingDenseEigenSolver?17:46
lambdayyeah direct as in triangular solver17:47
@wikingah ok17:47
lambdayummm... but that works for dense17:47
@wikingwell then let's keep it as is17:47
@wikingthat solver is going to be working only with17:47
@wikingCDenseMatrixOperator and SGMatrix17:47
lambdayI really like the passing linear operator on the fly idea - that's how things should be17:47
lambdaywiking: yeah!17:47
@wikingso basically now i've added17:47
@wikingvoid compute(SGMatrix<float64_t> m);17:47
@wikingto CDirectEigenSolver17:47
lambdaywiking: awesome!17:48
@wikinglisitsyn: ping17:48
@wikingi wonder what happend though when u pass SGMatrix<T> as is on the stack instead of reference/pointer17:48
lambdaywiking: so you need to change just a bit at RationalApproximation::precompute() where we call the compute function17:48
@wikingisn't the whole class is being copied to the stack?17:48
lambdaywiking: nah the underlying memory is on the heap17:49
@wikinglambday: yeah i was wondering if u can just pintpoint me the lines17:49
lambdayit increases the refcount17:49
@wikingwhat i should put to the compute() in RationalApproximation::precompute17:49
lambdayumm just a sec then :)17:49
@wikingk thnx17:49
@wikingif i have that i'll compile/test locally and if it goes ok i'll push in a sec the whole feature branch17:50
lisitsynwiking: pong17:50
-!- lambday_ [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun17:50
@wikinglambday: btw then LanczosEigenSolver should have compute(CLinearOperator<float64_t>* linear_operator), compute(SGMatrix<float64_t> linear_operator) and compute(SGSparseMatrix<float64_t> linear_operator) right?17:50
lambday_wiking: sorry got dc17:51
@wikinglisitsyn: see above my question regarding stack + sgmatrix17:51
@wikinglambday_: nw17:51
lambday_let me think17:51
lambday_yeah17:51
lisitsynwiking: so pass-by-value of SGMatrix?17:51
lisitsynit increases refcount17:51
lisitsynthen decreases17:51
@wikinglisitsyn: mmm good17:51
lambday_you internally create the appropriate linear operator17:51
lambday_and then do the rest17:52
lisitsynif compiler is valid (not visual studio) it shouldn't be optimized out17:52
lambday_lisitsyn: lol @valid17:52
@wikinglisitsyn: vs will kill it?17:52
@wikinglisitsyn: really? :)17:52
lisitsynwiking: no I am just unsure17:52
lisitsynI am unsure about everything related to vs17:52
@wikingah ok :> i thought u r talking out of experience17:52
lisitsynhah well out of other experience17:53
lisitsyne.g. covertree doesn't work with vs17:53
lisitsynbecause of I don't know what :D17:53
lambday_wiking: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/mathematics/linalg/ratapprox/opfunc/RationalApproximation.cpp#L11417:53
@wikinglambday_: ok so can u give me a small diff for CRationalApproximation, i.e. where do you have the LinearOperator?17:53
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds]17:53
@wikinglambday_: yeah my cursor is there in the editor17:53
-!- lambday [67157e54@gateway/web/freenode/ip.103.21.126.84] has joined #shogun17:54
lambdayargh this fucking freenode!17:54
@wikingheheh yeah it's hectic lately17:54
lambdaywiking: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/mathematics/linalg/ratapprox/opfunc/RationalApproximation.cpp#L11417:54
@wiking17:53 <@wiking> lambday_: yeah my cursor is there in the editor17:54
lambdaywiking: you should pass the linear operator (which is m_linear_operator)17:55
@wikingwhere's the linear operator that i can pass to compute?17:55
lambdaywiking: its in the base class of RationalApproximation, COperatorFunction17:55
@wikinglambday: mmm ok. where do u create the EigenSolver?17:55
lambdayprotected17:55
@wikinglambday: since we have to remove the linear operator from the ctor of EigenSolver17:56
lambdaywiking: things are created externally17:56
@wikinglambday: ah i see.. unit test?17:56
lambdaywiking: umm... there are few unit-tests17:56
@wikingor example?17:56
lambdayand examples, both17:56
lambday:(17:56
lambdaysorry :(17:56
lambdaywiking: can't we jointly contribute to your feature branch? then I can help17:56
lambdaybefore merging to develop17:57
lambdaywiking: do a quick grep :P17:57
@wikinglambday: mmm ok i can now push these changes into a feature branch17:57
@wikinglambday: RationalApproximation ?17:57
@wikinggrep for thi?17:57
@wikingthis17:57
lambdaywiking: nah EigenSolver17:57
lambdayrational approximation is just another base :P17:57
-!- lambday_ [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds]17:57
@wikinglambday: whooohooo... even in notebook :)17:58
lambdayshit17:58
@wikingno worries17:58
lambdaywiking: argh I had this in the linear solvers this way, why on earth didn't I think of this that time17:59
lambdaypass things on the fly I mean17:59
lambday(unless there was something that I am completely forgetting :-/)17:59
* lambday bites nails17:59
@wikinglambday: just some more seconds...18:00
@wiking;)18:00
-!- lambday_ [67157e54@gateway/web/freenode/ip.103.21.126.84] has joined #shogun18:01
@wikinglambday: ok so here18:01
lambday_wiking: where?18:01
@wikinghttps://github.com/shogun-toolbox/shogun/blob/develop/tests/unit/mathematics/linalg/LogDetEstimator_unittest.cc#L9818:01
@wikingis it safe just to remove op from the operator?18:01
@wikingright...? as we'll have reference on op within CLogRationalApproximationIndividual right?18:02
lambday_wiking: I think so, its SG_REF'ed anyway18:02
lambday_wiking: oh have it SG_REF'ed within the compute18:02
lambday_wiking: and SG_UNREF in the end18:02
@wikingsame goes for line 187, 272, 328, 394,18:03
@wikingright?18:03
@wikinglambday_: ok will do18:03
-!- lambday [67157e54@gateway/web/freenode/ip.103.21.126.84] has quit [Ping timeout: 245 seconds]18:03
-!- lambday__ [67157d51@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.81] has joined #shogun18:04
@wikinglambday_: have u seen this:18:04
@wiking18:02 <@wiking> same goes for line 187, 272, 328, 394,18:04
@wiking18:02 <@wiking> right?18:04
@wiking18:02 <@wiking> lambday_: ok will do18:04
lambday__wiking: irc is getting into my nerves18:04
lambday__checking... same file you mean?18:04
@wikinglambday__: yes18:04
@wikinghttp://github.com/shogun-toolbox/shogun/blob/develop/tests/unit/mathematics/linalg/LogDetEstimator_unittest.cc18:04
@wikinglambday__: and of course the same question goes for18:05
lambday__yeah, as long as its ref'ed already, I think we're cool18:05
@wikinglambday__: well i do an SG_REF+SG_UNREF within compute(CLinearOperator...)18:06
lambday__wiking: yeah!18:06
-!- lambday__ is now known as lambday18:06
@wikingok i'm fixing now the examples...18:07
-!- lambday_ [67157e54@gateway/web/freenode/ip.103.21.126.84] has quit [Ping timeout: 245 seconds]18:07
lambdaywiking: please let me know when you push things18:07
lambdaywiking: I will come back after dinner :)18:07
@wikinglambday: well now my only concern here of course18:07
@wikingcompute_all vs compute_extremal...18:07
@wikingas i really dont want to do these changes 2 times :)))18:07
lambdaylol18:08
lambday:D18:08
lambdaythe eigen solvers that we have are actually meant for extremals18:08
lambdaysee there are two data members for max and min eigenvalues :P18:08
lambdaywiking:which ideally should be removed alltogether18:09
lambdayand compute_extremal  should return a shogun_wrapper for std::pair18:09
lambdayand compute_all a SGVector18:09
-!- kislay [~kislay@93.114.45.13] has joined #shogun18:09
@wikinglambday: heheh yeah...18:09
@wikinglambday: ok you know what i'm pushing now these changes and then we'll discuss this later...18:10
lambdayalso, I guess only real eigenvalues are assumed18:10
-!- lambday [67157d51@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.125.81] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]18:10
-!- lambday [67157d4e@gateway/web/freenode/ip.103.21.125.78] has joined #shogun18:10
lambdaywiking: dc again18:10
@wikinglambday: no worries.. i'm already running make + tests18:11
lambdayanyway it was a nice discussion18:11
@wikinglambday: and we'll see the other changes...18:11
lambdaywiking: let me know if I can be of any help :)18:11
lambdaywiking: ciao man! brb after dinner :)18:12
lambdaywiking: bye18:12
@wikingcya18:12
-!- lambday [67157d4e@gateway/web/freenode/ip.103.21.125.78] has quit [Client Quit]18:12
-!- giridar [b49531e5@gateway/web/freenode/ip.180.149.49.229] has joined #shogun18:15
@wikinglisitsyn: fyi: https://github.com/shogun-toolbox/shogun/issues/1930#issuecomment-3709741718:16
-!- vadresh [~vadresh@62.109.10.44] has quit [Remote host closed the connection]18:24
-!- dhruv13 [~dhruv13@84.64.15.124] has quit [Ping timeout: 241 seconds]18:31
-!- giridar [b49531e5@gateway/web/freenode/ip.180.149.49.229] has quit [Quit: Page closed]18:32
-!- cameo54321 [2729cfbf@gateway/web/freenode/ip.39.41.207.191] has joined #shogun18:36
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 240 seconds]18:53
-!- kislay [~kislay@93.114.45.13] has joined #shogun19:06
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun19:08
shogun-notifier-shogun: Viktor Gal :feature/UnifiedEigenSolvers * cb82500 / / (12 files): https://github.com/shogun-toolbox/shogun/commit/cb825006fc161d03be09cf7a8b7b129ee3892cd719:08
shogun-notifier-shogun: Initial refactor of linalg/eigsolver classes19:08
shogun-notifier-shogun: starting work on issue #193019:08
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun19:28
-!- mode/#shogun [+o iglesiasg] by ChanServ19:28
-!- dhruv13 [~dhruv13@84.64.13.59] has joined #shogun19:47
-!- dhruv13 [~dhruv13@84.64.13.59] has quit [Quit: Leaving]20:01
-!- dhruv13 [~dhruv13@84.64.13.59] has joined #shogun20:01
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer]20:04
lisitsynwiking: wow this is faster than I can follow :D20:10
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Quit: Leaving]20:10
-!- kislay [~kislay@93.114.45.13] has joined #shogun20:20
@wikinglisitsyn: yeah should contrib your thoughts ;P20:27
@wikinglisitsyn: asdf dsygev dgremm20:27
@wiking:)20:27
lisitsynwiking: I am actually very afraid of doing this stuff20:27
lisitsynlike adding some wrapper code for linear algebra20:28
lisitsynalways better to outsource that :D20:28
@wikinghahah true that20:33
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun20:39
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Quit: Page closed]21:00
kislaywiking,  is there any way to push changes in the shogun/data folder without cloning everything present in it.21:05
@wikingkislay: you'll have to do the fork21:06
kislaythat i did.21:07
@wikingkislay: so they u fork it + add your changes21:07
kislayi am able to push my other commits, but changes in the data folder are not showing up21:07
kislayactually, i had to add my dataset.21:07
@wikingand do a PR against shogun-data21:07
@wikingkislay: nono21:08
@wikingkislay: that won't work21:08
@wikingdont do changes within the data submodule21:08
@wikingyou have to fork the data repo21:08
@wikingdo the changes21:08
@wikingcommit21:08
@wikingand then PR against the shogun-data21:08
kislayoh!!!21:08
kislayas simple as that21:08
kislaypfff21:08
@wikingyeah something like that21:08
kislaythanks.21:09
@wikingnw21:12
-!- travis-ci [~travis-ci@ec2-54-80-214-106.compute-1.amazonaws.com] has joined #shogun21:28
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/2035717421:28
-!- travis-ci [~travis-ci@ec2-54-80-214-106.compute-1.amazonaws.com] has left #shogun []21:28
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun21:32
lambdaywiking: having a fight - be back in a few mins - please note that the eigensolvers that I planned only works with SPD matrices21:33
@wikinglambday: yeah i saw that21:34
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds]21:37
-!- thoralf [~thoralf@p4FEC87BB.dip0.t-ipconnect.de] has joined #shogun21:37
thoralfHey!21:37
lisitsynthoralf in da house!21:38
thoralfYo :D21:38
thoralfCan da house tell me which indent style to use?21:38
thoralffor shogun?21:38
lisitsynthoralf: tabs/spaces?21:39
lisitsynthoralf: I see tabs in code mostly ;)21:39
thoralfIndentation, braces, etc.21:39
lisitsynthoralf: oh21:39
thoralfEvery time I send a PR, somebody complains.21:40
lisitsynif (shitz)21:40
lisitsyn{21:40
lisitsyn<tab>dosomething;21:40
lisitsyn<tab>dosomething;21:40
lisitsyn}21:40
lisitsynif (shitz)21:40
lisitsyn<tab>dosomething;21:40
thoralfNono, I mean, which indent style.21:41
thoralfindent <-- The command line tool.21:41
lisitsynthoralf: I have no idea what do you mean by that then :D21:41
thoralfindent <-- A unix tool to format code.  It has different styles. ;)21:42
lisitsynoh never used that :D21:42
lisitsynthoralf: okay then no idea ;)21:43
thoralfWhat are you using?21:44
lisitsynthoralf: I just write it formatted already21:44
thoralflol21:44
* thoralf writes in different languages - and every language has different styles.21:44
thoralfFor tabs/spaces, for braces, for everything.21:45
thoralfI'm always mixing this up, so I want my computer to fix that.21:45
lisitsynthoralf: use good ide ;)21:45
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]22:07
-!- pavelsom [~pavelsom@193.106.31.18] has joined #shogun22:09
-!- pavelsom [~pavelsom@193.106.31.18] has quit [Remote host closed the connection]22:15
cameo54321Hi wiking: Got a sec?22:18
@wikingshoot22:21
cameo54321I am trying to link stan with shogun22:22
-!- sonne|osx [~sonne@e179065162.adsl.alicedsl.de] has joined #shogun22:22
@wikingcameo54321: oh that's going to b fun22:22
cameo54321have been fiddling with cmake files of shogun for quite some time, but of no use..22:22
cameo54321haha22:22
@wikingcameo54321: but i'm afraid that is going to be rather hard...22:22
@wikingi mean rather hard to convince ppl here to be integrated22:23
cameo54321Very much... stan doesn't have a cmake kind of structure22:23
cameo54321And no configure scripts either22:23
cameo54321just a custom make file22:23
@wikingas it really dpeends on some heavy weith stuff22:23
@wiking*weight stuff like boost22:24
cameo54321Yeah.. By the way i got shogun to identify boost if installed on the system22:24
cameo54321I also went though the cmake files in cmake/external22:25
cameo54321tried to mimic the behavior there.. still no use..22:25
@wikingmmmm22:25
cameo54321Any ideas what can be the right approach to do that?22:25
@wikingcameo54321: why does stan depend on boost btw?22:25
cameo54321Good question.. I am unable to figure out yet.. but it seems it depends on boost quite heavily..22:26
@wikingcameo54321: can u give me a link to the library's src?22:27
@wikinglike github link or something22:27
cameo54321ya sure22:27
cameo54321https://github.com/stan-dev/stan22:27
@wikingoh man these guys are even crazier than we are... storying full libraries in their repo22:35
@wiking...22:35
cameo54321!22:36
@wikingshit seriously we do not want to depend on this library22:36
@wikingi mean i understand in a way22:36
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer]22:36
@wikingbut isnt there a better library22:36
@wikingthis one is crazy22:36
cameo54321Heiko pointed this issue... https://github.com/shogun-toolbox/shogun/issues/187522:37
cameo54321opened*22:37
cameo54321His suggestion was  that we can make cmake to download stan into external_libs dir just like colpack/eigen3(if nonavailable), etc22:39
@wikingyea that was not a good idea22:46
cameo54321Please see this: http://www.shogun-toolbox.org/irclogs/%23shogun.2014-03-07.log.html#t18:3622:46
@wikingcameo54321: i do understand that heiko wants to have this22:46
@wikingbut this is not a good idea22:46
@wikingim sorry for the contradiction22:46
@wikingbut integrating this especially as a bundle is really too much22:47
cameo54321The other way is that we can build these ourselves22:47
@wikingthis will bloat shogun like with another 500megs22:47
@wikingnobody wants that22:47
cameo54321Can there be a way if we just pick some code base classes from stan?22:48
@wikingwell that could be an option... but i've just checked some stuff and it's really heavily bundled with boost22:48
@wikingand we really would like to be dependent on boost22:49
@wikingi'm sorry i have to be like this22:49
@wikingi wouldn't rush myself into this if i were u22:49
cameo54321haha no prob22:49
@wikingimo heiko made a rush about this before checking out the status of stan22:49
cameo54321Well we'll develop stuff ourselves then..22:50
cameo54321May be get some inspiration as to how they structure some basic stuf... Class heirarchies for example..22:51
cameo54321but build stuff on our own, using boost...22:52
-!- kislay [~kislay@93.114.45.13] has joined #shogun22:52
cameo54321independently22:52
cameo54321wiking: thanks for the in time comments :-)22:54
-!- funcode [~funcode@117.55.241.55] has joined #shogun23:03
-!- cameo54321 [2729cfbf@gateway/web/freenode/ip.39.41.207.191] has quit [Ping timeout: 245 seconds]23:09
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 244 seconds]23:16
thoralfwiking: Hey.23:17
thoralfwiking: How would you read multilabel files in shogun?23:18
thoralfIt's about 3-4 lines in perl, but about 100 in C++.23:20
thoralfI'd be happy to have it in 20 or less lines ;)23:20
@wiking:D23:24
@wikingthoralf: official svmlight doesn't support it? or?23:24
thoralfNo, it doesn't.23:24
thoralfBut it's text-based.23:24
thoralf<hacky idea>We can put it in the first column without spaces.</>23:25
@wikingthe svmlight format does not support the multi-label data representation. You should convert your data into arff format.23:25
@wiking^ from a group discussion23:25
thoralfHmm.  But svmlight is so easy to generate. :)23:26
thoralfAnd it's <labels><space><data>23:26
thoralfAnyway, the file I'm reading in MultilabelLabels is simply "1 2 3 4 7 9" per line.23:27
thoralfNo features attached to this.23:27
@wikingyep  get it23:27
-!- tomy18 [~alex@79.114.164.19] has quit [Ping timeout: 260 seconds]23:28
thoralfFirst (bad) thing is that I'm reading the file two times.  First time to get number of lines, second time to parse it.23:29
-!- thoralf [~thoralf@p4FEC87BB.dip0.t-ipconnect.de] has quit [Quit: Konversation terminated!]23:57
--- Log closed Sun Mar 09 00:00:30 2014
« Friday, 2014-03-07 Index Sunday, 2014-03-09 »

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