Open in new window / Try shogun cloud
--- Log opened Mon Sep 16 00:00:01 2013
wikingi wonder what is the equivalent of FILE* in python00:00
wikingiglesiasg: maybe you need to implement00:00
wikingCSVFile(int fd)00:00
wikingand that'll solve your problem00:00
@iglesiasggsomix, what do you think? ^00:01
wikingyeah i think that's the problem00:01
wikingthat u need a file descriptor ctor00:01
wikingand not a FILE*00:01
gsomixiglesiasg, I'm ok with it.00:01
wikingfor this00:01
wikingiglesiasg: just create such a ctor00:01
wikingand just convert fd -> FILE*00:01
wikingin the ctor00:02
wiking( FILE *00:02
wiking     fdopen(int fildes, const char *mode);00:02
wiking^ is the standard c funct for this00:02
@iglesiasgwiking, and from python the int fd will be the output from, right?00:02
wikinghope so00:02
@iglesiasgisn't it a bit weird since then the file would be like opened twice?00:03
wikingyou are passing an FD00:03
@iglesiasgand then in the constructor00:03
wikingin ctor you open an fd00:03
wikingi mean you pass an fd00:03
wikinghence it's open00:03
@iglesiasgall right00:03
wikingyou just ready from the fd starting from that moment in CSVFile00:03
wikingu just need00:03
wikingfdopen(int fildes, const char *mode);00:04
wikingto convert from fd -> FILE*00:04
wikingas CSVFile is writter for using FILE*00:04
gsomixhm, are you sure that returns fd?00:04
wikinggsomix: most probably....00:04
gsomix>> This is a shorthand for GzipFile(filename, mode, compresslevel).00:04
gsomix>> class gzip.GzipFile([filename[, mode[, compresslevel[, fileobj[, mtime]]]]])00:04
gsomixbut class GzipFile class have  def fileno(self): method.00:07
* gsomix hates python's doc00:08
@iglesiasggsomix, any hack required to make a CSVFile constructor visible from interfaces?00:11
@iglesiasgI just compiled and installed and the same error persists00:12
@iglesiasgbut in the list of CSVFile constructors printed in the error the CSVFile(int) constructor I created does not appear00:12
@iglesiasgall right, cleaning and re-compiling...00:14
gsomixiglesiasg, I'd like to help, but... z-z-z00:20
@iglesiasggsomix, no worries00:20
gsomixiglesiasg, so if ctor works you can use fileno method of GzipFile for getting fd00:23
@iglesiasggsomix, nice, thanks!00:23
@iglesiasgthat can be indeed useful00:23
wikinggta v is out00:27
@iglesiasggood bye guys, see you in three months again :P00:29
-!- hushell [] has joined #shogun00:30
wikingah ok00:30
wikingi dont have neither ps3 nor xbox00:30
-!- pickle27 [~Kevin@] has joined #shogun00:47
-!- pickle27 [~Kevin@] has quit [Client Quit]00:47
@iglesiasgwiking, one of my friends got a ps3 a couple of days ago for this lol00:51
@iglesiasgwiking, btw it seems that doesn't return an int00:51
@iglesiasgor a fd for that matter00:51
wiking00:22 < gsomix> iglesiasg, so if ctor works you can use fileno method of GzipFile for getting fd00:52
@iglesiasgwiking, bah it didn't work for some reason00:53
@iglesiasgwiking, I can call the constructor with CSVFile(gzipfile.fileno())00:53
wikinghow so?00:53
@iglesiasgbut the features object has 0 features and vectors00:53
wikingmmm iglesiasg test then first the ctor00:54
wikingi.e. write a small c++ code00:54
@iglesiasgwiking, this is the constructor00:54
@iglesiasg+CCSVFile::CCSVFile(int fd) :00:54
@iglesiasg+       CFile(fdopen(fd, "r"), NULL)00:54
@iglesiasg+       init();00:54
wikingthat open a csv file with int fd = open('filename', 'r')00:54
@iglesiasg+       init_with_defaults();00:54
@iglesiasgI might have very well screwed00:54
wikingand then pass that fd to the new ctor00:55
@iglesiasgjust looked up what the other constructors do00:55
wikingCFile(fdopen(fd, "r"), NULL) -> CFile(fdopen(fd, rw), name)00:57
wikingand he ctor shouuld be00:57
@iglesiasgwiking, rw is "rw"?00:58
@iglesiasgI just didn't add the arguments for name and permissions at first00:59
wikingCSVFile(int fd, const char* name)00:59
wikingCSVFile(int fd, char rw, const char* name)00:59
wikingbut yeah00:59
wikingother than that that should work00:59
wikingThe mode of the stream must be compatible with the mode of the file descriptor.01:00
wikingmaybe it has different modes?01:00
wikingu should really first test this ctor from c++ code01:00
@iglesiasgwiking, how do you suggest me to try it?01:05
wiking00:54 < wiking> that open a csv file with int fd = open('filename', 'r')01:07
@iglesiasgwiking, all right, let's do that01:07
@iglesiasgsorry for making you repeat stuff :S01:08
@iglesiasg wiking, so I think the problem is that opening the file with gzip is not doing anything about uncompressing it01:20
@iglesiasgthe method does it01:20
@iglesiasgbut if we just open it and give it to CSVFile, then it is not uncompressed01:20
-!- shogun-notifier- [] has quit [Quit: transmission timeout]01:28
-!- iglesiasg [] has quit [Ping timeout: 264 seconds]01:32
-!- van51 [] has quit [Quit: Leaving.]02:13
-!- iglesiasg [] has joined #shogun02:19
-!- mode/#shogun [+o iglesiasg] by ChanServ02:19
-!- sonne|osx_ [] has joined #shogun03:13
-!- sonne|osx [] has quit [Ping timeout: 256 seconds]03:13
-!- sonne|osx_ is now known as sonne|osx03:13
-!- gsomix [~gsomix@] has quit [Remote host closed the connection]03:55
-!- iglesiasg [] has quit [Quit: Ex-Chat]04:25
-!- zxtx [] has joined #shogun05:58
-!- sonne|osx [] has quit [Quit: sonne|osx]06:46
-!- besser82 [] has joined #shogun07:52
-!- besser82 [] has quit [Changing host]07:52
-!- besser82 [~besser82@fedora/besser82] has joined #shogun07:52
-!- sonne|osx [~sonne@] has joined #shogun07:58
besser82sonne|osx: Der neue build f?r rubygem-narray f?r FC19 ist unterwegs, sollte im laufe des heutigen Tages in den Repos sein :)08:01
sonne|osxbesser82: very good08:01
sonne|osxerr sehr sch?n :)08:01
besser82sonne|osx: F?r FCRH m?sste das update schon da sein :)08:01
besser82sonne|osx: uuups, was confused.  This is an English channel :)08:02
lisitsynbesser82: na ja das ist ok08:02
besser82OK. :D08:02
besser82Ihr seid fast alles deutsche, oder?08:02
lisitsynbesser82: well I am very basic with deutsch ;)08:03
besser82lisitsyn: I see :)08:05
lisitsynbesser82: but HeikoS is native and wiking speaks it ok IIRC08:06
besser82lisitsyn: HeikoS, I know about :)  Had a 1:1 with him when I intially joined this channel.  But it's good to know wiking has knowledge in German, too.  This may come in handy on squashig problem eith CMake :)08:07
-!- shogun-notifier- [] has joined #shogun08:08
shogun-notifier-shogun: Soeren Sonnenburg :develop * f56e63e / examples/undocumented/r_modular/classifier_svmocas_modular.R:
shogun-notifier-shogun: fix ocas R example08:08
lisitsynbesser82: and I cover russian side  :)08:09
besser82lisitsyn: Good to know. Some russian buddy of mine possibly wanted to comaintain SHOGUN on FC :)08:09
besser82lisitsyn: so we have very good lang-covering here ;)08:10
lisitsynbesser82: yeah08:11
-!- sonne|osx [~sonne@] has quit [Quit: sonne|osx]08:15
shogun-buildbotbuild #1742 of bsd1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Soeren Sonnenburg <>08:17
shogun-buildbotbuild #80 of precise - libshogun is complete: Success [build successful]  Build details are at
-!- gsomix [~gsomix@] has joined #shogun08:30
gsomixsonney2k, hey08:30
shogun-buildbotbuild #96 of FC19 - modular_interfaces is complete: Failure [failed configure]  Build details are at  blamelist: Soeren Sonnenburg <>08:31
-!- sonne|work [] has joined #shogun08:34
-!- gsomix [~gsomix@] has quit [Client Quit]08:34
shogun-buildbotbuild #1477 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>08:49
shogun-buildbotbuild #1824 of deb3 - modular_interfaces is complete: Failure [failed test python_modular test r_modular]  Build details are at  blamelist: Soeren Sonnenburg <>08:55
-!- travis-ci [] has joined #shogun09:09
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:
-!- travis-ci [] has left #shogun []09:09
-!- lambday [67157d37@gateway/web/freenode/ip.] has joined #shogun09:14
sonne|workwiking: awake?09:27
wikingyep yep09:28
sonne|workwiking: could you enable running examples and tests for the static interfaces?09:29
sonne|workwiking: you currently enabled for cmdline only09:29
sonne|workand no tests09:29
wikingit's not about enabling them09:29
wikingit's about implementing them..09:29
wikingand that's more time09:29
sonne|workerrm what changed?09:30
wikingnothing but it was never implemented09:30
sonne|workI mean how do you run *_modular examples?09:30
wikingit has nothing to do with it09:30
wikingi mean cmake needs to be created09:31
wikingthe files that actually will run the examples09:31
sonne|workso I guess later then :/09:32
sonne|workwiking: and what did you find out about the R dispatcher?09:32
sonne|workwiking: not sure if you have seen but we now have only 2 examples failing09:32
sonne|workand it seems like real bugs / memory corruption09:33
lisitsynsonne|work: is thursday 18:00 ok for you?09:33
lisitsynsonne|work: your09:33
sonne|workno time09:34
lisitsynI am having hard time with that doodle09:34
lisitsynsonne|work: do you see anything that fits your timespans?09:34
besser82sonne|work: with the recent git-checkout build fails again, when no .git-dir is present: src/shogun/base/Version.cpp:20:43: Fehler: too many decimal points in number: const int32_t Version::version_revision = VERSION_REVISION;09:34
besser82sonne|osx: src/shogun/base/Version.cpp: In statischer Elementfunktion >>static int32_t shogun::Version::get_version_revision()<<:09:35
besser82sonne|osx: src/shogun/base/Version.cpp:70:1: Warnung: Kontrollfluss erreicht Ende von Nicht-void-Funktion [-Wreturn-type]09:35
sonne|workbesser82: I didn't change a thing though, please paste the generated src/shogun/lib/versionstring.h09:35
sonne|worklisitsyn: looking09:36
wikingsonne|work: yeah i'll do the things...09:36
wikingsonne|work: but now i'm working on something else :)09:36
wikingfor the first time09:36
sonne|workwiking: what did you find out about R?09:36
sonne|worklisitsyn: times are in UTC?09:36
lisitsynsonne|work: berlin tz09:36
lisitsynsonne|work: argh09:37
sonne|worklisitsyn: so when I click 11PM then it is local time right?09:37
lisitsynsonne|work: at the top of the table09:37
lisitsynsonne|work: there is timezone thing09:37
sonne|workyes it says berlin09:37
lisitsynsonne|work: yeah then what you see is berlin tz09:38
besser82sonne|work: here you go:
lisitsynsonne|work: because it shows me moscow tz for a reason :D09:38
sonne|workbesser82: #define VERSION_REVISION 0x3.0.009:38
sonne|workso yes indeed wrong09:38
besser82sonne|work: how should it be supposed to look like? 3.0.0?09:39
sonne|workI guess I will have to strip the dots and fill in some 0's09:39
sonne|workmaybe 0x30000009:40
besser82sonne|work: looks better, meanwhile I'll use some sed-magic on it....09:41
sonne|worklisitsyn: ok filled in09:42
besser82sonne|work: sed-magick doesn't work file get's regenerated on `make`09:43
sonne|workbesser82: sure I will fix it relax09:45
besser82sonne|work: thanks  :)09:45
lisitsynsonne|work: ohh the best case is w/o pickle27 lambday hushell patrick and fernando09:46
sonne|worklisitsyn: what shall I say...09:47
-!- lambday [67157d37@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]09:49
sonne|workbesser82: any idea how to generate a regexp that adds a trailing 0 ? Or some cmake way to do the job 1 -> 01 ?09:57
besser82sonne|work: you are talking about PREpending a "0"?09:58
sonne|workbesser82: yeah that is what I meant09:58
besser82sonne|work: SET (var "0${var}")  should do :)09:58
sonne|workbesser82: well but var needs to be 0...9 then09:59
besser82sonne|work: or you can SET(var "0") ....  SET( var "${var}${new-value}")  ;)10:00
besser82sonne|work: if you need some more elaborated way, please point me to the code/script in question10:01
sonne|workbesser82: well version.cmake in cmake dir10:02
sonne|workbesser82: I am extracting a VERSION_MAJOR / MINOR now10:02
sonne|workfor getting the10:02
sonne|workI need to prepend a 0 if the MAJOR/MINOR is <1010:03
besser82if (${var} <= 9)10:04
besser82set(var "0${var}")10:04
besser82sonne|work:  ^^^10:05
lisitsynI am the worst manager10:09
besser82sonne|work: build of testsuit fails, too:  Linking CXX executable shogun-unit-test10:09
besser82CMakeFiles/shogun-unit-test.dir/ In function `Dummy_dummy_Test::TestBody()':10:09
besser82/home/besser82/shared/github/shogun/third_party/GoogleMock/gtest/include/gtest/internal/gtest-internal.h:443: multiple definition of `Dummy_dummy_Test::TestBody()'10:09
besser82CMakeFiles/shogun-unit-test.dir/ first defined here10:09
besser82CMakeFiles/shogun-unit-test.dir/ In function `Dummy_dummy_Test::TestBody()':10:09
besser82/home/besser82/shared/github/shogun/third_party/GoogleMock/gtest/include/gtest/internal/gtest-internal.h:443: multiple definition of `Dummy_dummy_Test::test_info_'10:09
besser82CMakeFiles/shogun-unit-test.dir/ first defined here10:09
besser82CMakeFiles/shogun-unit-test.dir/ In function `Dummy_dummy_Test::TestBody()':10:09
besser82/home/besser82/shared/github/shogun/third_party/GoogleMock/gtest/include/gtest/internal/gtest-internal.h:443: multiple definition of `Dummy_dummy_Test::TestBody()'10:09
besser82CMakeFiles/shogun-unit-test.dir/ first defined here10:09
besser82CMakeFiles/shogun-unit-test.dir/ In function `Dummy_dummy_Test::TestBody()':10:09
besser82/home/besser82/shared/github/shogun/third_party/GoogleMock/gtest/include/gtest/internal/gtest-internal.h:443: multiple definition of `Dummy_dummy_Test::test_info_'10:09
besser82CMakeFiles/shogun-unit-test.dir/ first defined here10:09
besser82CMakeFiles/shogun-unit-test.dir/ In function `Dummy_dummy_Test::TestBody()':10:09
besser82sonne|work:  ^^^10:10
sonne|workbesser82: no idea about that one - that is sth for wiking...10:10
besser82sonne|work:  kk, will ping him about ;)10:11
besser82wiking: can you have a look, please.  ^^^^10:11
wikingmmm will do10:11
wikingbesser82: send me an email10:11
besser82wiking: your adress?10:12
besser82wiking: thx!10:13
besser82wiking: should be there in a sec  ;)10:14
-!- lisitsyn [] has quit [Quit: Leaving.]10:15
-!- lisitsyn [] has joined #shogun11:05
-!- shogun-notifier- [] has quit [Quit: transmission timeout]11:08
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:f40b:9348:29da:4f3b] has joined #shogun11:44
-!- mode/#shogun [+o iglesiasg] by ChanServ11:44
-!- lisitsyn [] has quit [Read error: Connection reset by peer]12:05
-!- HeikoS [] has joined #shogun12:12
-!- mode/#shogun [+o HeikoS] by ChanServ12:12
-!- lambday [67157e37@gateway/web/freenode/ip.] has joined #shogun12:15
@HeikoSlambday: hi!12:17
lambdayHeikoS: hi12:17
lambdayhehe I pinged at the same time12:17
@HeikoSjust checked your PR, still building in travis12:17
@HeikoShaha :)12:17
lambdayHeikoS: with 10 log-det samples, I get pretty good accuracy12:17
@HeikoSso how are things?12:17
lambday>>> print sum(probing_estimates)/len(probing_estimates) 1422.85679236 >>> print actual 1422.8440472412:17
lambdayworking with python now12:17
@HeikoSthis is the sparse matrix we had?12:18
lambdayI took A'*A+eye12:18
@HeikoSok cool12:18
@HeikoSis this dataset public open?12:18
lambdayHeikoS: I guess so12:19
lambdayscipy is an angel12:19
@HeikoSlambday: yep its very cool12:19
lambdaygreat! So I can use this for all the examples, right?12:19
@HeikoSyeah, we just have to find the license12:20
@HeikoSI dont know whether we can just copy it into our gpl repo12:20
lambdayHeikoS: how about I store the A'A+I one in the repo, that too would be causing license issues?12:21
@HeikoSI dont know12:21
@HeikoSbetter check12:21
@HeikoSIf it is one, we can download it from the python example or so12:21
@HeikoSThe original data sets and descriptions are from the   NIST web site and were not covered by an explicit copyright.   Modifications for R data sets are covered by GPL (>= 2).12:22
@HeikoSso thats the way we should do it12:22
@HeikoSshould add some comment stating where the data is from and that its not under gpl, but the changes are12:23
@HeikoSthats how R does it12:23
@HeikoSlambday: your latest patch, does that work with python modular?12:23
lambdayHeikoS: locally it works12:23
@HeikoSok cool,12:24
@HeikoSthere were other problems too12:24
@HeikoSdid you rebase recently?12:24
lambdaybut on travis some examples/tests fail12:24
@HeikoSok I will have a look once they are done12:24
lambdayHeikoS: got an weird swig related error for JobResultAggregator, left that one for the time being from adding to modular interface...12:24
lambdaythe java modular failure was being caused by that12:25
lambdayoh btw I converted this to ascii for libshogune examples as well, but some changes in the load_svmlight_file in SGSparseMatrix lately and it can't read12:25
lambdayHeikoS: weirdly enough, I just set the max iteration limit to 10000 and it converges properly12:26
lambdayfor CG-M12:26
lambdaymax/min eigs are similar to the one you had for your ipython12:26
lambdaynot same but of similar range12:27
@HeikoSlambday: ah annyoing with the changes12:27
@HeikoSlanczos might converge to different things in different runs12:27
@HeikoSmake sure the examples do not take too long to run12:27
@HeikoSthe python ones are part of the test-suite, so should be in less than say 2-3seconds12:28
@HeikoSthe ipython can take slightly longer (maybe 30s max)12:28
@HeikoSbut shorter is better12:28
@HeikoSrather add a comment for the users that they can increase this or that to get "proper" results12:28
lambday2-3 secs!12:29
lambdayI have to do with just 1 log-det sample then12:29
lambdayyeah that sounds good12:29
lambdayso, the matrix (mat and ascii) of A'A+I, where would it go?12:30
@HeikoSok thats ok12:30
@HeikoSinto the data repository, in fact put the original dataset there12:30
@HeikoShave a look at the other files12:30
@HeikoSand add some comment with the readme12:30
lambdayalright checking12:30
@HeikoSyou can compute A^T A yourself then12:30
lambdayummm... yeah, but for the libshogun ones, it would be easier if we have the final matrix too, otherwise again gotta use eigen312:31
lambdaybut for the time being only adding python examples12:32
@HeikoSlambday: thats fine isnt it?12:32
lambdaywill add libshogun ones when load_from_svmlight works12:32
lambdayyeah, should be fine12:32
@HeikoSI mean eigen needs to be there anyway12:32
@HeikoSlambday: yep ok12:32
@HeikoSlet me know how it goes12:32
-!- lisitsyn [] has joined #shogun12:32
lambdayHeikoS: alright12:33
@HeikoSlambday: I told some people about this stuff and they liked it a lot. In particular that you did this c-implementation12:33
lambdayI hope this becomes useful... but we sure need to add preconditioners12:34
@HeikoSlambday: yep we need those12:34
lambdayfeels good that things work properly12:36
lambdaycocg too works now12:36
@HeikoSlambday: yep very good!12:36
@HeikoSprobably takes much longer right?12:36
lambdayI checked up to rational approximation of log(A) using unit basis vectors - they work as good as triangular solvers12:37
lambdayfor that test, they don't take long, converges within a few iterations12:38
besser82sonne|work, wiking: /home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(23,17): error CS0246: The type or namespace name `SVRLight' could not be found. Are you missing a using directive or an assembly reference?12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(24,17): error CS0841: A local variable `svr' cannot be used before it is declared12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(26,17): error CS0841: A local variable `svr' cannot be used before it is declared12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(29,67): error CS0841: A local variable `svr' cannot be used before it is declared12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(29,53): error CS1502: The best overloaded method match for `LabelsFactory.to_regression(Labels)' has some invalid arguments12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/build/src/interfaces/csharp_modular/modshogun.dll (Location of the symbol related to previous error)12:40
besser82/home/besser82/rpmbuild/BUILD/shogun-f56e63e62ca6a3f6f252e50543ec5b21bd676e8a/examples/undocumented/csharp_modular/regression_svrlight_modular.cs(29,53): error CS1503: Argument `#1' cannot convert `object' expression to type `Labels'12:40
besser82Compilation failed: 6 error(s), 0 warnings12:40
besser82make[2]: *** [examples/undocumented/csharp_modular/csharp-regression_svrlight_modular] Error 112:41
besser82sonne|work, wiking:  ^^^ happenes when -DUSE_SVMLIGHT=OFF.12:41
@HeikoSbesser82: this looks like forgotten guards for SVMLight12:41
besser82HeikoS: who in charge for this?12:42
@HeikoSbesser82: the one who finds it :D12:42
@HeikoSI think in the example, there just needs to be a check whether USE_SVMLIGHT is defined12:42
@HeikoSand if not, not execute things12:43
@HeikoSor not compile to be more exact12:43
besser82HeikoS: I'm no official SHOGUN, I'm afraid...  Nor do I know about C#....12:43
@HeikoSlike #ifdef HAVE_BLABLALBA ..... #endif12:43
@HeikoSbesser82: yeah was just a joke, could you fill in an issue12:43
@HeikoSbesser82: I dont know who knows c# here12:44
besser82HeikoS: kk, will do :D12:44
lambdayman shogun-data is *huge*12:46
@HeikoSlambday: yep12:46
lambdayHeikoS: still downloading... oh and it doesn't have any README, so should I just add a separate directory for logdet and add a Readme.txt there, where I say that this matrix from this and that and GPL thing12:47
besser82lambday: what are going to bundle into SHOGUN?12:48
besser82lambday: s/are /&you12:48
@HeikoSbesser82: thanks!12:49
lambdaybesser82: adding a .mat file basically12:49
@HeikoSlambday: no put it there directly12:49
@HeikoSlambday: dont put a mat file12:49
@HeikoSput a more standard format like csv, like the other files in there12:49
@HeikoSmaybe even zipped12:49
@HeikoSlambday: we can ask sonney2k about the license stuff later, just add for now12:49
lambdayHeikoS: alright.. so I have it in svmlight format12:50
@HeikoSthats good12:50
lambdayHeikoS: but for python, I was planning to use the mat/mtx one12:51
@HeikoSlambday: you can load the thing and convert or?12:52
lambdayHeikoS: ya, and then tocsc()...12:53
@HeikoSlambday: yeah thats good12:53
@HeikoScsc_matrix() from scipy.sparse12:53
@HeikoSalso does that12:53
besser82lambday: I thought I would have been another piece of code ;)12:53
besser82HeikoS: np. We all want to get that in shape, don't we?12:54
lambdayHeikoS: so, shall I add the mtx that I directly downloaded from the website, and then my converted ascii svmlight format one (for libshogun examples), tarzipped12:55
lambdaybesser82: :D12:55
@HeikoSbesser82: yes!12:56
@HeikoSbesser82: you could also try to compile with LAPACK not being installed12:57
@HeikoSI fixed like 50 issues around that a few weeks ago12:57
@HeikoSbut there are more12:57
lambdayHeikoS: shogun-data/testsuite/mathematics/west0479.tar.gz sounds good?12:59
-!- gsomix [~gsomix@] has joined #shogun12:59
gsomixsonney2k, sonne|work hey ho12:59
lambdaywest0479.mtx and west0479.txt inside12:59
@HeikoSlambday: yep!12:59
@HeikoSls /share/apps/lapack-3-4.1/13:00
@HeikoSliblapack.a  librefblas.a  libtmglib.a13:00
@HeikoSthis is what I have13:00
@HeikoSah sorry wrong window13:00
lisitsynwiking: ping13:06
-!- van51 [] has joined #shogun13:16
van51sonne|work: ping13:17
sonne|workvan51: hey!13:21
sonne|workgsomix: ho!13:21
van51sonne|work: hey! I've started a comment on github, so I'll finish it there :)13:21
gsomixsonne|work, thnx for comments!13:22
sonne|workgsomix: hope they were useful...13:23
sonne|workvan51: ok ping me once done13:23
lambdayHeikoS: changed... moved it under logdet/13:23
gsomixsonne|work, yep, of course. CProtobufFile is a good idea I think.13:23
sonne|workHeikoS: btw have you seen almost all R examples seem to work now...13:24
sonne|workHeikoS: r_modular that is13:24
@HeikoSsonne|work: its R modular13:24
@HeikoSyeah I have seen something13:24
@HeikoSis it wiking?13:24
sonne|workonly 2 failing13:24
@HeikoSsonne|work: is this with reference counting?13:24
sonne|workI fixed them...13:24
@HeikoSsonne|work: wow! great job!13:25
@HeikoSmany people will love you for that? :)13:25
sonne|workI would wish we could compare if they do the right job...13:25
sonne|workstill things are a bit awkward13:25
sonne|worke.g. there are some weird bugs13:25
@HeikoSsonne|work: yeah testing the interfaces somehow would be great13:25
sonne|worklike you do kernel.init() for linear kernel13:25
@HeikoSsonne|work: btw a question: where in shogun-data should sparse test matrices for log-det framework go, new folder logdet?13:25
sonne|workand for some reason kernel.init_optimization is attempted to be called13:26
@HeikoSsonne|work: ok ...13:26
van51sonne|work: done13:26
@HeikoSsonne|work: if we can get it working at least minimally13:26
@HeikoSthen people would already use it, find more problems and maybe some people will start fixing them13:26
@HeikoSsonne|work: just have been to a stats workshop and everyone is using R13:26
@HeikoSits so weird13:26
sonne|workHeikoS: well it is it seems
@HeikoSthey dont know anything else, they dont know even what else exists13:27
sonne|workstatistics people...13:27
sonne|workvan51: ok will have a look13:27
@HeikoSbut would be good users ;)13:27
@HeikoSlambday: merged the data13:27
-!- shogun-notifier- [] has joined #shogun13:27
shogun-notifier-shogun-data: lambday :master * a12a28c / logdet/west0479.tar.gz:
shogun-notifier-shogun-data: Added matrix-data in mtx (original) and ascii (converted) (log-det)13:27
shogun-notifier-shogun-data: Heiko Strathmann :master * 7d37a77 / logdet/west0479.tar.gz:
shogun-notifier-shogun-data: Merge pull request #19 from lambday/master13:27
sonne|workgsomix: have you seen the <size> <protobufmsg> format suggestion?13:27
shogun-notifier-shogun-data: Added matrix-data in mtx (original) and ascii (converted) (log-det)13:27
lambdayHeikoS: thanks :)13:28
gsomixsonne|work, yep, I'm using this format for now.13:28
lambdayHeikoS: my PR gives green except python modular, that fails for a few tests13:28
lambday3 tests basically13:28
sonne|workgsomix: just make sure that the <size> thing is endian save13:29
shogun-notifier-shogun: lambday :develop * bc17983 / / (163 files):
shogun-notifier-shogun: refactored log-det and computation framework13:30
shogun-notifier-shogun: lambday :develop * d49c048 / / (16 files):
shogun-notifier-shogun: COCG linear solver fixed, unit-test modified (log-det)13:30
shogun-notifier-shogun: lambday :develop * 050f2a2 / src/ (5 files):
shogun-notifier-shogun: modular interface updated (log-det)13:30
shogun-notifier-shogun: Heiko Strathmann :develop * af2d9dc / / (175 files):
shogun-notifier-shogun: Merge pull request #1586 from lambday/feature/log_determinant13:30
shogun-notifier-shogun: Refactored log-det and computation framework13:30
@HeikoSlambday: yep thats not your fault, merged :)13:30
lambdayHeikoS: silly question - how shall I use this data for examples?13:30
@HeikoSlambday: well, import it13:31
@HeikoSlook at the other examples :)13:31
@HeikoSvery easy13:31
lambdayalright checking13:31
sonne|workvan51: can I reply here?13:31
sonne|workvan51: you can do w/o function pointers - I mean this is C++ so just an interface class would do13:32
sonne|workvan51: and you could add the abstraction to do the dot into this function13:32
shogun-buildbotbuild #2160 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: lambday <>13:32
sonne|workvan51: so some can call dense_dot some can do it better13:32
shogun-buildbotbuild #2162 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>, lambday <>13:34
van51sonne|work: ok sounds reasonable. I'll do it this way then :)13:35
van51I'll go grab a bite and get started13:36
sonne|workvan51: cool!13:36
-!- votjakovr [] has joined #shogun13:52
-!- travis-ci [] has joined #shogun14:27
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:
-!- travis-ci [] has left #shogun []14:27
@HeikoSvotjakovr: hi!14:28
@HeikoSvotjakovr: hope you are fine!14:29
@HeikoSvotjakovr: I thought that maybe for the notebook, one example could be on string data14:30
@HeikoSwhat do you think?14:30
votjakovrHeikoS: hi! yeah it will be cool, do you have any particular suggestions on dataset or i could choose the data by myself?14:37
@HeikoSvotjakovr: if you want, you can choose yourself, if not, I am just about to look something up14:38
@HeikoSthis one here is good14:38
@HeikoSI tried that before with an SVM14:38
@HeikoSthe string kernel they use is implemented in shogun14:38
@HeikoSvotjakovr: would be really cool to have a string example since most toolboxes dont have that14:39
votjakovrHeikoS: ok, cool, i'll add one :)14:40
@HeikoSvotjakovr: maybe you can also compare to an SVM in the notebook? (not in the examples)14:40
@HeikoSthis way, we could illustrate the certainties etc14:40
votjakovrHeikoS: yeah, great idea14:41
@HeikoSvotjakovr: if the dataset is too hard to import or so, you can choose another one, let me know ...14:42
@HeikoSactually, I think I might still have the plain sequence files somewhere14:42
@HeikoSIll look it up at home14:42
@HeikoSvotjakovr:  and as a standard numerical dataset, just pick something from UCI14:43
@HeikoSvotjakovr: like this one ^14:43
@HeikoSthere you can do window-glass against non-window-glass14:43
@HeikoSthat is very popular14:43
@HeikoSand no missing values etc14:44
@HeikoSjust import, normalise (remove mean and whiten or scale features to unit variance) and then it directly should work14:44
@HeikoSI tried it before with your binary logit classifier14:44
@HeikoSok going for lunch, see you!14:44
votjakovrHeikoS: Ok, thank you very much on datasets :) BTW Could you please look at some draft on gradients here:
@HeikoSok, will do after lunch if thats ok?14:45
votjakovrHeikoS: ok :)14:45
@HeikoScool! :)14:45
@HeikoSah cool you pthreaded it?14:45
votjakovrHeikoS: yeah14:46
-!- votjakovr [] has quit [Read error: Connection reset by peer]15:04
-!- votjakovr [] has joined #shogun15:19
@HeikoSvotjakovr: I like the derivative method!15:57
@HeikoSvotjakovr: any reason why you used pthread rather than openmp (which makes the code more readable, although thats not a problem here)15:57
@HeikoSwiking: ping15:58
votjakovrHeikoS: huh, good :) currently i'm using clang as a main compiler and clang doesn't support openmp yet, but we can change implementation in future, since it's pretty simple :-)16:03
@HeikoSvotjakovr: yeah no problem, just asking whether it was a conscious decision16:04
-!- az_de [82954e22@gateway/web/freenode/ip.] has joined #shogun16:10
-!- van51 [] has quit [Quit: Leaving.]16:15
-!- shogun-notifier- [] has quit [Quit: transmission timeout]16:30
-!- az_de [82954e22@gateway/web/freenode/ip.] has left #shogun []16:49
-!- shogun-notifier- [] has joined #shogun17:05
shogun-notifier-shogun: Soeren Sonnenburg :develop * 20a7eb6 / cmake/version.cmake:
shogun-notifier-shogun: wip commit17:05
shogun-buildbotbuild #2163 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Soeren Sonnenburg <>17:19
-!- foulwall [~zhengyang@] has quit [Remote host closed the connection]17:20
-!- sonne|osx [~sonne@] has joined #shogun17:24
-!- gsomix [~gsomix@] has quit [Quit: Leaving]17:27
shogun-buildbotbuild #152 of osx1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>17:39
-!- sonne|osx [~sonne@] has quit [Quit: sonne|osx]17:40
-!- lisitsyn [] has quit [Quit: Leaving.]17:41
lambdayHeikoS: I added two examples, one for CG and another for log-det17:49
@HeikoScool, where?17:50
@HeikoSis there a PR?17:50
lambdayHeikoS: but the data I put there is in tar.gz, so17:50
lambdayHeikoS: yeah just sending17:50
lambdayHeikoS: currently I loaded the matrix from the data as if its extracted in the same dir :-/17:51
@HeikoSlambday: are you working in python?17:51
@HeikoSsince in there you can easily extract compressed files17:52
lambdayHeikoS: two more python modular examples I thought of, one for COCG, and another is CG-M (because the rest of the solvers are pretty much common but I haven't seen these two anywhere)17:52
lambdayHeikoS: yes17:52
@HeikoSand the file reader fraction of shogun is working on zipped formats, too17:52
lambdayHeikoS: oh cool!17:52
@HeikoSlambday: ok cool17:52
@HeikoSsend the PR, Ill grab a coffee now but will look later17:52
lambdayHeikoS: but then I should modify to use the zipped one17:53
@HeikoSlambday: yeah with python thats easy17:53
@HeikoSlambday: getting coffee :)17:56
-!- sonne|osx [] has joined #shogun17:59
shogun-notifier-shogun: Soeren Sonnenburg :develop * 97d3a04 / cmake/version.cmake:
shogun-notifier-shogun: fix version18:04
-!- travis-ci [] has joined #shogun18:05
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:
-!- travis-ci [] has left #shogun []18:05
besser82sonne|osx: still fails with new commit: /home/besser82/shared/github/shogun-test/src/shogun/base/Version.cpp:20:43: Fehler: numerischer literaler Operator >>operator"" x<< konnte nicht gefunden werden18:09
besser82 const int32_t Version::version_revision = VERSION_REVISION;18:09
besser82                                           ^18:09
besser82/home/besser82/shared/github/shogun-test/src/shogun/base/Version.cpp: In statischer Elementfunktion >>static int32_t shogun::Version::get_version_revision()<<:18:09
besser82/home/besser82/shared/github/shogun-test/src/shogun/base/Version.cpp:70:1: Warnung: Kontrollfluss erreicht Ende von Nicht-void-Funktion [-Wreturn-type]18:09
besser82 }18:10
besser82 ^18:10
besser82make[2]: *** [src/shogun/CMakeFiles/shogun.dir/base/Version.cpp.o] Fehler 118:10
besser82sonne|osx: ^^^18:10
besser82sonne|osx:  ---> paste of versionstring.h18:11
besser82sonne|work:  ^^^^18:12
lambdayHeikoS: sending PR18:20
shogun-buildbotbuild #2164 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Soeren Sonnenburg <>18:20
-!- zxtx [] has quit [Ping timeout: 240 seconds]18:22
lambdayHeikoS: please have a look18:23
lambdayHeikoS: I am going for dinner.. will be back in an hour or so18:23
lambdayHeikoS: regarding plots, I am thinking of these 1) CG-M convergence for the log-det (residuals over the iterations), 2) Probing/Normal sampler estimates 3) matrix-power vs log-det-estimates (was trying with matrix power 2 and I think it can make a difference)18:26
lambdayoh and 4) number of shifts vs log-det accuracy18:27
lambdayHeikoS: with diagrams explaining sparsity, fill-in18:28
lambdayHeikoS: brb after dinner :)18:29
@HeikoSlambday: that sounds very nice!18:52
-!- travis-ci [] has joined #shogun19:26
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:
-!- travis-ci [] has left #shogun []19:26
-!- van51 [] has joined #shogun19:38
@HeikoSwiking: hi!19:44
wikinghey hey19:44
wikingi'm still not online actually19:44
@HeikoSthen nevermind :D19:44
@HeikoSI need a break anyways19:44
@HeikoSbe back in 1019:44
@HeikoSwiking: now?19:57
-!- lisitsyn [] has joined #shogun19:57
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:f40b:9348:29da:4f3b] has quit [Quit: Ex-Chat]19:59
@HeikoSlisitsyn: hi!20:00
lisitsynHeikoS: hey20:00
@HeikoSlisitsyn: I have more questions20:00
@HeikoSlisitsyn: so the cluster guy now installed lapack for me20:00
lisitsynHeikoS: .so?20:00
@HeikoSbut I still have problems20:00
lisitsynHeikoS: what's the problem?20:01
@HeikoSso first of all, cmake did not detect them even though they were in /usr/lib6420:01
@HeikoShowever -DBLAS_LIBRARIES=/usr/lib64 -DLAPACK_LIBRARIES=/usr/lib64 did that20:01
@HeikoSbut now unit tests dont compile20:01
lisitsynHeikoS: sorry will get back in ten minutes20:01
lisitsynHeikoS: re20:11
lisitsynHeikoS: did you revert your changes in cmakelists?20:12
@HeikoSlisitsyn: hi20:12
@HeikoSyes I did20:12
@HeikoSgit reset --hard should do it right?20:12
lisitsynHeikoS: heh yeah20:12
lisitsynHeikoS: alright20:13
@HeikoSso any idea?20:13
@HeikoSfirst weird thing: why doesnt cmake detect my files?20:13
lisitsynHeikoS: so the problem is that it doesn't work when you don't do /usr/lib6420:15
@HeikoS-- A library with BLAS API not found. Please specify library location.20:15
@HeikoS-- LAPACK requires BLAS20:15
@HeikoS-- A library with LAPACK API not found. Please specify library location.20:15
@HeikoSI did not touch LD_LIBRARY_PATH20:15
@HeikoS(for those things)20:15
@HeikoSsince it is /usr/lib6420:15
@HeikoSthat should not be the problem right?20:15
lisitsynHeikoS: I think we use default lapack finder20:15
@HeikoSbut ok20:16
@HeikoSI can specify the path and then it works20:16
lisitsynso that's out-of-box bug20:16
@HeikoSalso, I checked my files, they contain that symbol20:16
lisitsynHeikoS: so the main problem is that unit-tests do not compile I guess20:17
lisitsynthat's more interesting20:17
@HeikoSno other problem compiling libshogun20:17
@HeikoSthey should also fail if lapack is not there right?20:17
lisitsynHeikoS: I don't know if it is still needed20:19
@HeikoSah wait20:19
@HeikoS-- A library with BLAS API found.20:19
@HeikoS-- A library with LAPACK API found.20:19
@HeikoS-- Looking for clapack_dpotrf in /usr/lib64/atlas/
@HeikoSWARNING: Target "cmTryCompileExec2276832605" requests linking to directory "/usr/lib64".  Targets may link only to libraries.  CMake is dropping the item.20:19
@HeikoS-- Looking for clapack_dpotrf in /usr/lib64/atlas/ - found20:19
@HeikoS-- Found ATLAS: /usr/lib64/atlas/;/usr/lib64/atlas/;/usr/lib64/atlas/;/usr/lib64/atlas/;/usr/lib64/
@HeikoSthere is this warning20:19
lisitsynHeikoS: that sounds like cmake glitch20:19
lisitsynno idea what to do about it20:19
@HeikoSso any idea whether maybe the lib order in the build command causes this?20:20
lisitsynHeikoS: yeah may be20:21
lisitsynas your library files are here in the command line20:21
@HeikoSyeah and all those files exist20:24
@HeikoSlisitsyn: updated the issue20:26
@HeikoSthe files also contain the symbol that is needed20:26
@HeikoSso it *has* to be some homemade problem20:26
lisitsynHeikoS: weird20:27
@HeikoSah and annoying since I need that for work20:27
@HeikoSlisitsyn: so is first in the list20:28
@HeikoSmaybe it should be last?20:28
@HeikoSbut I dont really believe that20:28
lisitsynHeikoS: just try to rearrange them in the command20:28
-!- sonne|osx [] has quit [Quit: sonne|osx]20:29
@HeikoSlisitsyn: I asked wiking to add a bundle lapack option, that might help here20:33
wikingHeikoS: bundle lapack is impossible20:34
@HeikoSwiking: why?20:34
wikingHeikoS: if u can do it20:34
wikinggo for it20:34
@HeikoSI managed to compile it locally and create a shared library20:35
wikingit's just crazy shit20:35
@HeikoSmmh ok20:35
@HeikoSthere is this tool that helps creating the dynamic stuff20:35
@HeikoSwiking: do you have any idea about this problem I have?20:35
wikingmmm not yet20:36
wikingHeikoS: what's this distro?20:36
@HeikoSmmh, dont know, let me find out20:36
@HeikoSwiking: centos20:37
wikingHeikoS: which version?20:37
shogun-notifier-shogun: Soeren Sonnenburg :develop * dd41296 / cmake/version.cmake:
shogun-notifier-shogun: fix version (for real)20:37
@sonney2kbesser82, ok now!20:37
@HeikoScat /proc/version20:37
@HeikoSLinux version 2.6.18-308.13.1.el5 ( (gcc version 4.1.2 20080704 (Red Hat 4.1.2-52)) #1 SMP Tue Aug 21 17:10:18 EDT 201220:37
@HeikoS cat /etc/issue20:37
@HeikoSCentOS release 5.4 (Final)20:37
@HeikoSwiking: ^20:38
wikingi'll check it20:38
@HeikoScool man thanks!20:38
@HeikoSwiking: oh and btw, cmake doesnt detect the installed lapack20:38
@HeikoSI have to give the path by hand20:38
@sonney2kHeikoS, errm so it compiles or what?20:45
@HeikoSsonney2k: nope20:45
@sonney2kHeikoS, it fails compiling doing what?20:45
@HeikoS../../src/shogun/ undefined reference to `dstemr_'20:45
@HeikoSbut this symbol is there20:46
@HeikoSnm /usr/lib64/ | grep dstemr_20:46
@HeikoS00000000001e46d0 T dstemr_20:46
shogun-buildbotbuild #2165 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Soeren Sonnenburg <>20:46
@HeikoSand this file is also in the library list of the compile command20:46
wikingHeikoS: ldd ../../src/shogun/
@HeikoSldd ../../src/shogun/ =>  (0x00007fff8b7fd000)20:47 => /lib64/ (0x00002abdaf792000)20:47 => /usr/lib64/atlas/ (0x00002abdaf9ae000)20:47 => /usr/lib64/atlas/ (0x00002abdb00ab000)20:47 => /usr/lib64/atlas/ (0x00002abdb02ca000)20:47 => /usr/lib64/atlas/ (0x00002abdb04e8000)20:47 => /usr/lib64/ (0x00002abdb0ddd000)20:47 => /usr/lib64/ (0x00002abdb0ff1000)20:47 => /lib64/ (0x00002abdb12f2000)20:47 => /usr/lib64/ (0x00002abdb1575000)20:47 => /lib64/ (0x00002abdb1782000)20:47 => /lib64/ (0x00002abdb1991000)20:47
@HeikoS/lib64/ (0x000000315fe00000)20:47 => /usr/lib64/ (0x00002abdb1ce8000)20:47 => /lib64/ (0x00002abdb1f80000)20:47
@HeikoSah sorry20:47
@sonney2kHeikoS, so it compiled!20:47
@sonney2kso what is missing is a proper LD_LIBRARY_PATH=20:47
@sonney2kwhen running things20:47
wikingHeikoS: => /usr/lib64/atlas/
wikingthis does not have dstemr_  i guess20:48
@HeikoSnm says "no symbols"20:48
wikingHeikoS: ok so your problem is that the wrong is being used20:49
besser82wiking: It should have that symbol, since it's a rewritten lapack in c ;)20:49
@HeikoSsonney2k: shogun conmpiled, but its the unit tests ...20:49
@HeikoSwiking: ah ok20:49
@HeikoSso the one in lib64/atlas/ rather thanlib6420:49
@HeikoShow to change that?20:49
wikingas sonney2k said20:49
@HeikoSreally that easy? let me try20:50
@HeikoS export LD_LIBRARY_PATH=/usr/lib64/:$LD_LIBRARY_PATH  ?20:50
wikingtry that20:51
@HeikoSdoesnt work20:51
@HeikoSwhat about this list of .so files in the make command of the unit tests?20:51
@HeikoSin the gist20:51
@HeikoSsonney2k: to what do I have to set the LD_LIBRARY_PATH?20:51
besser82HeikoS, wiking: nm -D /usr/lib64/ | grep dstemr  ---> shows the symbol is there20:52
@HeikoSbesser82 yes it does, but => /usr/lib64/atlas/
@HeikoSand in there, it is not20:52
@sonney2kHeikoS, to /usr/lib64/atlas if it is there20:53
@HeikoSsonney2k: doesnt work either20:54
@HeikoSwiking, sonney2k cant I tell the compiler somehow to reference the right library somehow else?20:55
lisitsynI just wanted to note it is dstemr_ that is missed21:11
lisitsynmay be it is important here21:11
-!- zxtx [] has joined #shogun21:11
@HeikoSlisitsyn: ?21:12
lisitsynHeikoS: dstemr_ dstemr21:12
lisitsynHeikoS: do you have dstemr_ or dstemr?21:13
@HeikoSlisitsyn: I think it is as sonney2k and wiking said, references the wrong file, which is lib64/atlas/, but the file doesnt contain the symbol. However, lib64/ contains it, I just dont know how to make libshogun reference to it21:14
lisitsynHeikoS: ah21:14
lisitsynI see21:14
@HeikoSI tried prepending its path to LD_LIBRARY_PATH, the (re)-linking libshogun, but I still get the same error21:14
besser82HeikoS: lib64/atlas/ contains dstemr_ ---> nm -D /usr/lib64/atlas/ | grep dstemr21:15
lambdayHeikoS: back21:17
lambdayHeikoS: what's wrong with dstemr?21:18
@HeikoSlambday: hi! ah, dont worry about that21:18
@HeikoStrying to install things on weird machines21:18
@HeikoSlambday: I had some comments in the PR, minor things21:18
lambdayI added/used dstemr..  :-/21:18
lambdayHeikoS: yeah checking21:18
@HeikoSlambday: I cannot drop lapack unfortunately since I need the Lanczos on the cluster21:19
@HeikoSif only eigen could do that little thing ;)21:19
@HeikoSthen I wouldnt have these stupid problems with stupid lapack21:19
lambdayHeikoS: at times, I feel like adding that to eigen :-/21:20
lambdayHeikoS: lapack API is horrible.. I only succeded because lisitsyn explain each and every params to me one by one :D21:20
@HeikoSyeah I know that, he did that for me too a while ago21:20
lambdaylisitsyn: kudos, master yoda!21:21
@HeikoSwe should just get rid of lapack as soon as possible21:22
@HeikoSalmost all things can be replaced by eigen21:22
lambdayHeikoS: does lapack work faster than eigen3?21:22
@HeikoSit just burned 3 working days of mine, so no21:23
@HeikoScannot be that fast ;)21:23
@sonney2kHeikoS, well you suffer from bad ass ubuntu not lapack21:24
@sonney2kon debian you wouldn't have such trouble21:24
@HeikoSsonney2k: no its not ubuntu21:24
@sonney2kthen FC*21:24
@HeikoSwell I cannot do anything about this since its not my machine21:24
@HeikoSsonney2k: how to change this reference?21:24
lambdaysonney2k: you support lapack? :(21:25
@sonney2kHeikoS, what I would do is run make with verbose21:25
@sonney2kso you see the cmdline of linking libshogun21:25
lambdayusing it is so hard21:25
@sonney2kthen copy that line21:25
@HeikoSyes, its in the issue I made21:25
@sonney2kand then make sure it links to the libs you want by specifying -L/path21:25
@sonney2kand have *the same* libs in LD_LIBRARY_PATH21:26
@sonney2kthen it works21:26
@sonney2kHeikoS, where is the line?21:27
@sonney2kI only see attempts to link the example21:27
@HeikoSah sorry, will post21:27
@sonney2kHeikoS, and of course LD_LIBRARY_PATH won't help with *linking* an example21:28
@sonney2kyou need the proper -L / -l21:28
lambdayHeikoS: I have two files there, one mtx and another is libsvm.. I should add these individually zipped you meant21:28
@HeikoSlambday: why have two? why not one?21:29
lambdayHeikoS: the mtx one is for python modular examples, I intended to use the libsvm ones for libshogun examples21:29
@sonney2kHeikoS, try to add -/usr/lib64 -llapack to the end of the line21:29
@HeikoSsonney2k: ok thanks will try21:30
@HeikoSsonney2k: besser82 had another suggestion, comment out find_atlas in cmake, that in fact made cmake realize the proper .so file, currently compiling this way ...21:31
@sonney2kyou have most likely a clash between /usr/lib64/atlas/ and /usr/lib64/
@sonney2kHeikoS, that can work too21:31
@HeikoS /usr/lib64/atlas/ doesnt contain the symbol thats the problem21:31
@sonney2kHeikoS, then just don't link against it21:31
besser82sonney2k: the problem is CMakeLists.txt preferes ATLAS over FORTRAN-LAPACK21:32
@HeikoSsonney2k: besser82, doesnt work with disabling atlas, got even more undefined symbols21:32
@HeikoSsonney2k: I dont know how to not link against it21:32
besser82HeikoS: export LDFLAGS="-Wl,-l/usr/lib64/"21:33
besser82and try rebuilding with ATLAS re-enabled21:33
@HeikoSbesser82: what does that do?21:33
besser82adds original LAPACK to the linker-flags21:34
@HeikoSso before building I need to export right?21:34
@HeikoSbefore linking I mean21:34
besser82right, before configuring with comake ;)21:34
besser82cmake hardcodes ldflags to Makefile21:35
@HeikoSbesser that screws up the cmake totally21:36
besser82really? What's the output?21:36
-!- HeikoS [] has quit [Excess Flood]21:38
-!- HeikoS [] has joined #shogun21:39
-!- mode/#shogun [+o HeikoS] by ChanServ21:39
besser82wb, HeikoS21:41
@HeikoSsorry connection dropped21:42
@HeikoSnow compiling with verbose to see the link command of libshogun21:42
* HeikoS is dreaming of shogun in the cloud without installation21:44
* besser82 would say let's get started on openShift :)21:45
wikingHeikoS: well it needs to be installed once on the cloud as well21:45
besser82HeikoS: and will allow Software-as-a-Service ;)21:46
besser82HeikoS: easy deployment with containers and cartirdges ;)21:46
@HeikoSbesser82: yeah something like that21:46
besser82HeikoS: imageine a container as a vm21:46
@HeikoSan ipython server would actually be enough21:46
wikingbesser82: is it free? :D21:47
@HeikoScan write code in my browser then21:47
besser82wiking: openShift-community is free ;)21:47
wikingbesser82: yeah but not the service itself21:47
besser82wiking: and that's the EXACT codebase redhat uses ;)21:47
besser82wiking: a limited use is free as well21:48
wikingbesser82: i mean what we need is actually a machine where we can run it21:48
wikingbecause then we already have like virtualenv21:48
besser82amazon EC2?21:48
wikingbesser82: money? :)21:48
lisitsynwiking: CV help needed ;)21:49
lisitsynanyone else having a clue please comment too21:49
wikinglisitsyn: shoot although i'm like only 10% here21:49
besser82wiking: try creditcard, doesn't cost money, will be collect automatically ;)21:49
lisitsynmy problem is quite silly21:49
lisitsynI have bounding box where object is probably is21:49
@HeikoSsonney2k, wiking so the linking command of shogun contains /usr/lib64/atlas/ /usr/lib64/ in this order21:50
lisitsynbut I need to generate some candidates you know21:50
lisitsynhow not-brain-damaged-people-like-me do it?21:50
wikinghow do u assign probabilities?21:51
lisitsynwiking: no probabilities I just learned svm and have background model21:51
lisitsynso I just track background changes21:51
wikingok so u have the bo21:51
lisitsynyeah some box and I need candidates around it21:52
wikingwhat's candidates aroudn it?21:52
lisitsynwiking: well these regions I'll apply svm to21:52
lambdayHeikoS: there is a bug reagarding reading gzipped mtx file in python mmread :
lambdayHeikoS: I am adding the ImportErrors though, but not sure how to work with individual gzipped mtx files21:59
@HeikoScan you just only have libsvm format?21:59
lambdayHeikoS: yes, if I can read it from python, which I can I guess22:00
@HeikoSlambday: I am sure you can22:02
-!- lambday [67157e37@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]22:04
lisitsynHeikoS: is that lambday poweroff time?22:06
@HeikoSlisitsyn: maybe :)22:06
lisitsynHeikoS: ain't it insane?22:07
@HeikoShere is the output of the linking of shogun22:07
@HeikoSlisitsyn:  it is insane22:09
@HeikoSbut it really is like that oer there22:09
-!- lisitsyn [] has left #shogun []22:34
-!- iglesiasg [] has joined #shogun22:52
-!- mode/#shogun [+o iglesiasg] by ChanServ22:53
-!- HeikoS [] has quit [Quit: Leaving.]23:29
-!- votjakovr [] has quit [Quit: Good bye!]23:37
-!- shogun-notifier- [] has quit [Quit: transmission timeout]23:37
-!- besser82 [~besser82@fedora/besser82] has quit [Quit: Verlassend]23:43
--- Log closed Tue Sep 17 00:00:02 2013