IRC logs of #shogun for Tuesday, 2013-05-14

« Monday, 2013-05-13 Index Wednesday, 2013-05-15 »
--- Log opened Tue May 14 00:00:44 2013
@sonney2kHeikoS1, nope there still is some stuff not working00:11
-!- nube1 [~rho@36.253.43.151] has joined #shogun00:16
-!- nube [~rho@49.244.108.149] has quit [Ping timeout: 264 seconds]00:18
-!- nube1 [~rho@36.253.43.151] has quit [Quit: Leaving.]00:28
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]01:23
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has joined #shogun01:28
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has quit [Quit: Leaving.]02:20
shogun-buildbotbuild #394 of nightly_default is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/39404:38
-!- nube [~rho@116.90.239.3] has joined #shogun05:45
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 276 seconds]06:09
-!- nube [~rho@116.90.239.3] has joined #shogun06:14
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun06:54
-!- iglesiasg [~fjig@share-01.csc.kth.se] has joined #shogun09:15
-!- mode/#shogun [+o iglesiasg] by ChanServ09:15
@iglesiasghey guys, I am trying to use matlab static to run a couple of tests09:17
@iglesiasgI am getting this error though09:17
@iglesiasgInvalid MEX-file09:17
@iglesiasg/usr/local/etc/matlab/r2012a/bin/glnxa64/libstdc++.so.6: version `GLIBCXX_3.4.15' not09:17
@iglesiasgfound (required by /usr/local/lib/libshogun.so.1409:17
@iglesiasglet me know please if you happen to know something about it09:18
@wikingiglesiasg: which matlab?09:34
@wikingthere's a fix for this as far as i remember09:34
@iglesiasgwiking: 7.14.0.739 (R2012a)09:34
@wikingyeah09:34
@wikinggoogle that error09:34
@wikingand you'll find a fix at mathworks website09:35
@iglesiasgwiking: I found some people saying that sysem update helped09:35
@iglesiasgso I am trying with that09:35
@wikingiglesiasg: update of matlab to 2013a?09:35
@iglesiasgwiking: no, update of libstdc++09:35
@wikingnoooo09:36
@wikingiglesiasg: http://www.mathworks.es/support/solutions/en/data/1-EO8AAS/index.html09:36
@iglesiasgwiking: let me see, thanks you!09:36
sonne|work1iglesiasg: I posted a solution to this a couple of times on the mailinglist09:40
@iglesiasgsonne|work1: checking09:41
@wikingsonne|work1: put it in FAQ :)09:43
@wikingsonne|work1: http://shogun-toolbox.org/page/documentation/faq09:44
sonne|work1iglesiasg: please put it in FAQ when you're done :)09:45
@wikinglol09:45
@iglesiasgrecompiling atm...09:55
@wikingmmmm we have an error now in TParameter10:01
@wikingunit test10:02
@wiking[  FAILED  ] TParameter.equals_scalar_different110:02
@wiking[  FAILED  ] TParameter.equals_scalar_different210:02
@wiking[  FAILED  ] TParameter.equals_string_scalar_different10:02
@wiking[  FAILED  ] TParameter.equals_sparse_scalar_different10:02
@wikingmmm i guess it's only failing on my machine10:03
@wiking:D10:03
@iglesiasgwiking: I saw it on travis too10:04
@iglesiasgsonne|work1, wiking: I did a backup of the libstdc++* files in matlab sys/os directory and later symbolic linked the ones in /usr/lib/libstdc++* but the error is still there10:08
@iglesiasgI clean and recompiled matlab interface as well10:08
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has joined #shogun10:17
sonne|work1iglesiasg:  you removed this one here /usr/local/etc/matlab/r2012a/bin/glnxa64/libstdc++.so.6 ?10:21
sonne|work1just to be sure10:21
-!- sonne|work1 is now known as sonne|work10:21
sonne|workvan51: any progress?10:21
van51sonne|work: last night no, but I have an idea for debugging it10:22
van51I am working on it now10:22
@iglesiasgsonne|work: I made some progress when I removed the ones in bin/glnxa6410:28
@iglesiasgsonne|work: now I get version `GFORTRAN_1.4' not found (required by /usr/lib/liblapack.so)10:28
@iglesiasginstead of the GLIBC error10:29
sonne|workiglesiasg: now continue with the other libs in the glnxa64 dir until it works10:30
sonne|workvan51: very good10:30
@iglesiasgsonne|work: it seems to be working fine now10:37
sonne|workiglesiasg: -> FAQ :-)10:37
@iglesiasgIIRC stuff in the static interfaces is not directly available as in modular ones10:42
@iglesiasgthere is something to do in the ui directory10:43
sonne|workiglesiasg: yes sure10:46
sonne|workbut it would be worth having that compile error mentioned in the FAQ10:46
sonne|workit pops up pretty often10:46
sonne|workwiking: no this test should not fail10:46
@iglesiasgsonne|work: yeah, I am writting it10:46
@iglesiasgsonne|work: give 10 minutes to have it ready :)10:46
@iglesiasggive me*10:47
sonne|workwiking: it must be sth simple10:47
sonne|workwiking: how much does it (not) differ?10:47
@iglesiasgsonne|work: would it be crazy amount of work to get part of the SO framework to work in matlab?10:48
sonne|workno not crazy10:48
@iglesiasgsonne|work: what would it be in a few words? I have no idea how these classes in ui work or what do they have to do actually10:48
sonne|workwell you have to define appropriate commands10:50
sonne|workand then for each command add a function that does somthing for that command10:50
sonne|workso you define which arguments it has to have (input and output)10:50
@iglesiasgsonne|work: aham, I see10:50
sonne|workwhat is easy is to get the data / arguments into / out of shogun10:50
sonne|worke.g. to get a real number - just do get_real()10:51
sonne|workget_int() etc10:51
sonne|workfor matrix get_real_matrix() etc10:51
sonne|workhave a look at shogun/ui/SGInterface.cpp10:51
sonne|workI've created a couple of classes GUI* that are static helper things10:52
sonne|workotherwise this is stateless (!)10:53
sonne|workso these GUI* classes can store e.g. a CKernel10:53
sonne|workor features10:53
@iglesiasgsonne|work: ok, thanks!10:54
@iglesiasgsonne|work: FAQ updated11:06
sonne|workthanks!11:06
-!- iglesiasg [~fjig@share-01.csc.kth.se] has quit [Quit: leaving]11:08
sonne|workwell described!11:08
-!- iglesiasg [~fjig@share-01.csc.kth.se] has joined #shogun11:09
-!- mode/#shogun [+o iglesiasg] by ChanServ11:09
sonne|workiglesiasg: well described!11:18
@iglesiasgsonne|work: thanks! :)11:18
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun11:24
-!- mode/#shogun [+o lisitsyn] by ChanServ11:24
HeikoS1sonne|work: does the equals work now?11:24
sonne|workHeikoS1: Hey!11:32
HeikoS1hi11:33
sonne|workI managed to get say 50% of the tests to work11:33
sonne|workbut for the others it still fails11:33
HeikoS1cool11:33
HeikoS1what are the problems?11:33
sonne|workno idea11:33
HeikoS1debug outputs?11:33
sonne|workI used eps 1e-1511:33
sonne|workI can commit tester.py this evening11:34
sonne|workand comment the equals line to be used11:34
sonne|workso you could try yourself11:34
sonne|workbtw long double is really not needed :)11:34
sonne|work1e-300 or so is what double can do :)11:35
HeikoS1sonne|work, yes probably you are right :)11:35
HeikoS1ok cool, Ill have a look once you pushed the new tester11:35
@lisitsynI don't get that point11:35
@lisitsynsonne|work: what's the point with 1e-30011:36
sonne|worklisitsyn: no need to use long double11:36
sonne|workas eps precision11:36
@lisitsynahh11:36
@lisitsynfloating point is pain in the ass :D11:36
sonne|workHeikoS1: or do it yourself now - it is just this line here11:37
sonne|work    elif isinstance(a, modshogun.SGObject):11:37
sonne|work        return pickle.dumps(a) == pickle.dumps(b)11:37
sonne|workinstead of return pickle.dumps(a) == pickle.dumps(b)11:37
sonne|workdo return a.equals(b, 1e-15)11:37
HeikoS1ok will give it a try11:37
sonne|workHeikoS1: line 3011:37
HeikoS1thx11:38
sonne|worklisitsyn: yeah it is11:38
@lisitsynsonne|work: in windows kernel they do avoid floating and using fractions :D11:43
@lisitsyndid you know?11:43
sonne|workgood choice!11:43
@lisitsynsonne|work: thanks11:48
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.]12:00
votjakovrHeikoS1: Hi, i'm sorry, that was my mistake... I didn't think, that the derivative computed by log(ell).12:17
HeikoS1votjakovr: did carl answer to you too?12:18
votjakovrHeikoS1: Yep12:19
HeikoS1ok12:19
HeikoS1well, lesson learned :D12:19
HeikoS1so it was correct ....12:19
@lisitsynha!12:19
@lisitsynso you bugged carl?12:19
@lisitsyn:D12:19
HeikoS1never mess with the Cambridge guys :D12:19
votjakovr:)12:19
@lisitsynHeikoS1: one should always trust that curly guy12:20
HeikoS1votjakovr: I am still confused12:21
HeikoS1since he does ell=exp(hyp(1))12:21
HeikoS1so is the derivative wrt to log(ell)12:21
-!- iglesiasg [~fjig@share-01.csc.kth.se] has quit [Quit: Lost terminal]12:24
votjakovrHeikoS1: So if we want to compute derivative by ell, we should multiply his answer by derivative of log(ell), which equals to 1/ell12:27
-!- iglesiasg [~fjig@share-01.csc.kth.se] has joined #shogun12:28
-!- mode/#shogun [+o iglesiasg] by ChanServ12:28
votjakovrHeikoS1: But I little bit confused, why he use log(ell) as a hyper rather than ell.12:33
HeikoS1sonne|work: you broke the unit tests with the floatmax change btw12:33
HeikoS1sonne|work: which object had this num_element issue again?12:57
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun13:15
shogun-notifier-shogun: Heiko Strathmann :develop * 50e2efe / / (3 files): https://github.com/shogun-toolbox/shogun/commit/50e2efe9ccd2285683380efdd93b8322d13f19f513:15
shogun-notifier-shogun: -fixed floatmax_t issues from Sören's commit13:15
shogun-notifier-shogun: -moved num_lements check of DynamicObjectArray to CSGObject (cleaner)13:15
shogun-notifier-shogun: -added some new unit tests for num_elements and some NULL cases13:15
shogun-notifier-shogun: Heiko Strathmann :develop * 3105db8 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/3105db8aa1d7c7561a88383d28d503a18653ed1113:15
shogun-notifier-shogun: Merge pull request #1094 from karlnapf/develop13:15
shogun-notifier-shogun:13:15
shogun-notifier-shogun: a bunch of equals fixes13:15
shogun-buildbotbuild #1175 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/117513:18
sonne|workHeikoS1: you still have a SG_SPRINT("comparing \"%s\" to \"%s\"\n", get_name(), other->get_name()); in there13:20
sonne|workHeikoS1: how could the float64 change affect this?13:20
sonne|workHeikoS1: also you will need this num_elements hack for DynamicArray13:25
HeikoS1okay thats good to know13:28
HeikoS1will add...13:28
shogun-buildbotbuild #822 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/822  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:35
shogun-notifier-shogun: Heiko Strathmann :develop * f230b69 / src/shogun/base/SGObject.cpp: https://github.com/shogun-toolbox/shogun/commit/f230b69b7ca4434975aaf7fb88be92ab179e3a3313:36
shogun-notifier-shogun: added another num_elements check and changed print to debug print13:36
shogun-notifier-shogun: Heiko Strathmann :develop * a4bbf79 / src/shogun/base/SGObject.cpp: https://github.com/shogun-toolbox/shogun/commit/a4bbf79efc07375aca3d865793ba08100760d33813:36
shogun-notifier-shogun: Merge pull request #1095 from karlnapf/develop13:36
shogun-notifier-shogun:13:36
shogun-notifier-shogun: added another num_elements check and changed print to debug print13:36
sonne|workHeikoS1: does it work now btw for the tests?13:37
HeikoS1no13:37
HeikoS1and my terminal is spammed13:37
sonne|workahh yes there is a forgotten print msg at the bottom13:39
sonne|worklet me fix this13:39
sonne|workactually no13:41
sonne|workall intended13:41
sonne|workHeikoS1: so you deserve it :D13:41
sonne|workHeikoS1: what fails btw13:51
-!- travis-ci [~travis-ci@ec2-23-20-2-212.compute-1.amazonaws.com] has joined #shogun14:04
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/715264014:04
-!- travis-ci [~travis-ci@ec2-23-20-2-212.compute-1.amazonaws.com] has left #shogun []14:04
shogun-buildbotbuild #1115 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1115  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:07
van51sonne|work: I believe that the bug(if it's just one) is located in the CSVMLight::update_linear_component_mkl() method14:08
van51i'll try to find out exactly what it is now14:08
sonne|workvan51: could be... I wrote all that but hmmhh...14:08
van51I've re-introduced my changes gradually and the output was fine14:09
van51up until I changed that method14:09
-!- iglesiasg [~fjig@share-01.csc.kth.se] has quit [Quit: leaving]14:12
-!- Netsplit *.net <-> *.split quits: votjakovr14:16
van51sonne|work: found it! I hadn't changed a get_first_kernel() method to get_kernel(index)14:16
shogun-buildbotbuild #1116 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1116  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:17
sonne|workvan51: excellent!14:17
sonne|workvan51: so the output of the w matrix is now still the same?14:18
shogun-buildbotbuild #823 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/823  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:18
van51sonne|work: nope and I think it matches the original one14:19
van51but let me re-compile just to check14:19
-!- Netsplit over, joins: votjakovr14:21
-!- nube [~rho@49.244.11.195] has joined #shogun14:41
shogun-buildbotbuild #1117 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1117  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:44
-!- travis-ci [~travis-ci@ec2-23-20-2-212.compute-1.amazonaws.com] has joined #shogun14:48
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/715310214:48
-!- travis-ci [~travis-ci@ec2-23-20-2-212.compute-1.amazonaws.com] has left #shogun []14:48
sonne|workvan51: and?14:51
van51sonne|work: I recompiled the octave interface and the output seems ok to me14:53
van51i have kept it and will post a link to it on the PR page14:53
van51the thing is though, I did a ./configure && make now14:54
van51and it's stuck after a generating interface with SWIG.. Compiling interface14:54
sonne|workvan51: don't compile for *all* interfaces14:57
sonne|workthat can take ages14:57
sonne|work(1hr)14:57
sonne|workchoose your interface!14:57
van51sonne|work: I was just doing a ./configure before and then make and it was ok :P14:58
@lisitsynvan51: you have a lot of patience14:59
sonne|workyeah I would *never* do it like this14:59
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Ping timeout: 264 seconds]14:59
van51haha14:59
van51no really it wasn't taking that long14:59
@lisitsynvan51: what did you do waiting it to finish?15:00
@lisitsyn:D15:00
@lisitsynI think it takes one episode of some game of thrones15:00
@lisitsynor house m.d.15:00
@lisitsyn:D15:00
van51haha15:03
van51I was catching up on series15:03
van51yea15:04
van51:p15:04
van51or just wasted time there and there15:04
van51but I'm pretty sure it wasn't taking that long :D15:04
@lisitsynvan51: yeah when you are watching say 'breaking bad' time just flies away15:05
@lisitsyn:D15:05
@lisitsynvan51: ./configure --disable-optimization --interfaces=python_modular15:06
van51lisitsyn: now that you mention it, it could be :)15:06
@lisitsynvan51: ah and one more thing15:06
van51the only reason I noticed it now was because the computer became a bit unresponsive15:06
@lisitsynvan51: install ccache15:06
sonne|workvan51: yeah it needs 3GB or so to compile15:06
sonne|workand disable-optimization !15:07
sonne|workotherwise hmm15:07
@lisitsynand add the following lines to your bashrc15:07
@lisitsynCXX=ccache g++15:07
@lisitsynCC=ccache gcc15:07
@lisitsynthen reconfigure15:07
@lisitsynthis would save your time when you recompile the same thing twice15:07
van51lisitsyn: ah nice! thanks!15:08
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun15:10
-!- sijin [~smuxi@144.214.222.109] has joined #shogun15:21
sijinlisitsyn: are you there?15:24
@lisitsynsijin: yes15:24
sijinlisitsyn: I have written a test for kpca,15:25
@lisitsyngood15:25
@lisitsynwhere?15:25
sijinlisitsyn: I haven't commit it yet, I compare the result with matlab's kpca, I found the returned embedding of tapkee is the embbed data for input, which ,i think, can not be used as transition matrix15:26
@lisitsynsijin: yeah that might be a trouble15:27
sijinlisitsyn: embbedding_tapkee = diag * V', while I think the transition matrix should be 1/diag * V'15:28
sijinlisitsyn: therefore, we don't have  information of diag(eigen value of centered kernel matrix)15:29
@lisitsynsijin: alright got it15:31
sijinlisitsyn: I will submit the unit-test soon.15:36
sijinlisitsyn: can I just add a unit-test.cc in the unit/preprocessor's folder, do i need to modify another files?15:37
@wikingsijin: no15:37
@lisitsynyes just add a file15:38
@wikingsijin: just add the *_unittest.cc file there and that's it15:38
sijinlisitsyn: ok, thx15:38
-!- iglesiasg [~fjig@share-01.csc.kth.se] has joined #shogun15:39
sijinwiking: thx, i will use that naming15:39
@wikingwell if u dont use the right naming15:40
@wikingit wont work ;)15:40
sijinwiking: got it :)15:40
van51sonne|work: I pushed the fix along with the style changes you discussed on the PR16:08
van51g2g now16:10
van51cu later16:11
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has left #shogun ["QUIT :Leaving."]16:11
--- Log closed Tue May 14 16:37:59 2013
--- Log opened Tue May 14 16:38:09 2013
-!- shogun-t1olbox [~shogun@7nn.de] has joined #shogun16:38
-!- Irssi: #shogun: Total of 17 nicks [3 ops, 0 halfops, 0 voices, 14 normal]16:38
-!- Irssi: Join to #shogun was synced in 8 secs16:38
-!- Netsplit *.net <-> *.split quits: shogun-toolbox16:42
iglesiasgshogun-t1olbox ??16:43
sijinlisitsyn: I have sent a PR for KernelPCA_unittest17:00
-!- shogun-buildbot was kicked from #shogun by sonney2k [shogun-buildbot]17:04
-!- shogun-t1olbox was kicked from #shogun by sonney2k [shogun-t1olbox]17:04
--- Log closed Tue May 14 17:04:43 2013
« Monday, 2013-05-13 Index Wednesday, 2013-05-15 »

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