Open in new window / Try shogun cloud
--- Log opened Tue Sep 03 00:00:42 2013
@iglesiasghushell, hey there00:02
-!- travis-ci [~travis-ci@ec2-107-20-114-62.compute-1.amazonaws.com] has joined #shogun00:03
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/1090780400:03
-!- travis-ci [~travis-ci@ec2-107-20-114-62.compute-1.amazonaws.com] has left #shogun []00:03
@iglesiasgwhat happened with travis, we have lost it :(00:06
hushelliglesiasg: :)00:20
hushelliglesiasg: have you tried Mosek in the new build system?00:21
@iglesiasghushell, so to set up Mosek with CMake00:21
@iglesiasghushell, no, I haven't00:21
hushellI can't find an option to set it up00:21
@iglesiasghushell, because there is not such an option00:22
@iglesiasgthe right way to do it would be adding a FindMosek.cmake file similar to the ones in the cmake/ directory00:22
hushellin CMake, we should have a FIND_PACKAGE(Mosek), right?00:23
hushellyes00:23
@iglesiasghushell, if you want to just use it for yourself00:23
@iglesiasgyou can probably set up something faster that works just locally00:24
hushellor I write a FindMosek.cmake?00:24
hushellokay, so I have to modify the CMakeList.txt?00:24
@iglesiasghushell, well, if you think you can do that easily, go ahead00:24
@iglesiasgotherwise, I would say you rather focus on the stuff for the SO project :)00:25
hushellI found the Makefile generated by CMake is hard to read00:25
@iglesiasgno idea how hard is to write a .cmake file00:25
@iglesiasgyeah sure, that Makefile is not to be read or edited hehe00:25
@iglesiasgthat's why of using CMake00:25
hushell:D00:25
@iglesiasgusing a couple of set(CMAKE_INCLUDE_PATH ...)00:26
hushellokay, I'll find a way to add -lmosek00:26
@iglesiasgpointing to your local dirs were your Mosek is placed should be enough to get it working locally00:26
@iglesiasghushell, however there is another thing that is more important00:26
@iglesiasgso as you are experiencing yourself, our parameter learning methods for SO suck00:27
@iglesiasgBMRM works fine for the stuff that Michal tested but it seems non-trivial to get it working with new stuff00:27
@iglesiasgand my PrimalMosekSOSVM relies on Mosek, which is a pain too00:27
hushelliglesiasg: I see00:27
@iglesiasgone important part of your project is focused on improving this00:27
@iglesiasgwith subgradient descent and the Frank-Wolfe00:28
@iglesiasgsubgradient descent should be easy, I think00:28
hushelliglesiasg: I thought about implementing the SGD in one day to check my FactorGraphModel code00:28
@iglesiasghushell, I think it would be a good idea to include a good SGD implementation00:28
@iglesiasgthat works with all the models00:29
hushellyeah, anything involves 3rd party code is painful00:29
@iglesiasgso I don't know00:29
hushelliglesiasg: my idea is try your Mosek solver first and then go to SGD00:30
@iglesiasgI'd prefer if you could implement that and get it stable in Shogun rather than fighting to get Mosek to work with it00:30
@iglesiasgbut as you wish00:30
hushellsince I understand your code more or less00:30
@iglesiasgfeel free to comment and modify as you find suitable so it gets more understandable than less :)00:31
hushelliglesiasg: I think your cutting plane implementation is good00:31
hushellyeah, the learning is more complex than inference, so if any part goes wrong, the whole story crashes00:32
@iglesiasgdo you think so about learning and inference?00:34
hushelliglesiasg: Thanks for the suggestions, maybe few hours everything be fine :)00:34
@iglesiasgI think that inference is the hard part00:34
hushelliglesiasg: yes, but learning calls inference all the time00:34
hushellthis is very risky00:34
@iglesiasghushell, why is it risky?00:35
hushelliglesiasg: I mean if anything is wrong in inference00:35
hushelljust a feeling :)00:35
@iglesiasghehe ok00:36
hushellokay, I go ahead to fix my problem.00:37
hushelliglesiasg: Good nite!00:37
@iglesiasghushell, good luck, bye!00:37
-!- travis-ci [~travis-ci@ec2-107-20-114-62.compute-1.amazonaws.com] has joined #shogun01:06
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/1090811701:06
-!- travis-ci [~travis-ci@ec2-107-20-114-62.compute-1.amazonaws.com] has left #shogun []01:06
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]01:11
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has left #shogun ["Good bye!"]01:42
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun []03:35
-!- sonne|osx_ [~sonne@f053046104.adsl.alicedsl.de] has joined #shogun03:37
-!- sonne|osx [~sonne@f053040230.adsl.alicedsl.de] has quit [Ping timeout: 264 seconds]03:39
-!- sonne|osx_ is now known as sonne|osx03:39
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun06:49
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun []06:49
-!- sonne|osx [~sonne@f053046104.adsl.alicedsl.de] has quit [Quit: sonne|osx]07:29
-!- sonne|osx [~sonne@89.204.130.92] has joined #shogun08:11
-!- sonne|osx [~sonne@89.204.130.92] has quit [Quit: sonne|osx]08:35
-!- gsomix [~gsomix@83.234.54.223] has joined #shogun08:51
gsomixgood morning08:52
besser82mornin'09:02
gsomixwiking, hey. I have a little problem related to cmake.09:14
gsomixwiking, look09:14
gsomix CMake Error: Attempt to add a custom rule to output09:14
gsomix "/Workspace/shogun/build/src/interfaces/python_modular/Multiclass_includes.i.rule" which already has a custom rule.09:14
gsomixafter any changes in config with cmake/ccmake09:14
thoralfHeyhey.09:20
* gsomix away in uni09:24
thoralfHow bad.09:29
* thoralf just burnt 180 CPU-hours just to make evaluation counts overflow. :(09:30
sonne|workthoralf: how that?09:31
thoralfsonne|work: I used int64_t for counting tp/fp/tn/fn09:31
sonne|workso no way that this overflows...09:34
thoralfYou're right.09:34
thoralfBut it did.  Already looking for some implicit conversion...09:35
thoralfsonne|work: tn = num_classes*num_labels - (tp+fp+tn); <-- only num_classes and num_labels are 32 bit.09:36
thoralfsonne|work: Is it possible that some implicit cast happened?09:37
thoralftn = int64_t(num_classes)*int64_t(num_labels) - (tp+fp+tn); <-- I dont want to write this.09:37
thoralfsonne|work: The result of (int32_t*int32_t) is int32_t, even if assigned to something bigger.09:49
sonne|workthoralf: sure09:50
sonne|workthat is why we should do int64_t(int32_t) before09:50
thoralfsonne|work: I could have seen this, but I was dump. ;)09:52
thoralfsonne|work: Lesson learned.09:52
thoralfBig data is not only about big machines. ;)09:53
thoralfsonne|work: Would you argue against (1l*int32_t*int32_t)?09:57
thoralfToo hacky? ;)09:57
thoralfHmm. 1l could depend on the platform.09:59
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun10:03
shogun-notifier-shogun: Fernando Iglesias :develop * ed8d605 / examples/undocumented/python_modular/metric_lmnn_modular.py: https://github.com/shogun-toolbox/shogun/commit/ed8d6056152596d692c21ca4bf8f2e58c44dce6310:03
shogun-notifier-shogun: LMNN example for Python modular10:03
shogun-notifier-shogun: Fernando Iglesias :develop * 4c689ad / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/4c689ad4bfbb79b7b05374e74d2d8c8d16aba00410:03
shogun-notifier-shogun: Update LMNN and KNN Python modular examples to use CSVFile.10:03
shogun-notifier-shogun: Soeren Sonnenburg :develop * 874d45b / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/874d45b0de75d461c1e2c0e6f85abcf3d34e82ba10:03
shogun-notifier-shogun: Merge pull request #1508 from iglesias/feature/lmnn_python10:03
shogun-notifier-shogun:10:03
shogun-notifier-shogun: LMNN example for Python modular10:03
shogun-buildbotbuild #1672 of deb3 - modular_interfaces is complete: Failure [failed compile libshogun]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1672  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>10:12
sonne|workthoralf: just use int64_t(i)*j10:14
shogun-buildbotbuild #1588 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1588  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>10:14
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 256 seconds]10:33
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun10:43
-!- lisitsyn [~lisitsin@mxs.kg.ru] has left #shogun []10:43
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun10:43
thoralfHey Sergey10:47
shogun-buildbotbuild #1370 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1370  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>10:48
-!- travis-ci [~travis-ci@ec2-174-129-83-121.compute-1.amazonaws.com] has joined #shogun10:57
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1092357610:57
-!- travis-ci [~travis-ci@ec2-174-129-83-121.compute-1.amazonaws.com] has left #shogun []10:57
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:7c32:76f3:b1c5:6634] has joined #shogun11:00
-!- mode/#shogun [+o iglesiasg] by ChanServ11:00
@iglesiasgsonne|work, hey there!11:01
@iglesiasgsonne|work, the semicolon is what is missing here, right? https://github.com/shogun-toolbox/shogun/pull/1506/files#r611938211:01
@iglesiasgI noticed that most of the examples do not have it though11:01
shogun-buildbotbuild #1589 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1589  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:02
shogun-buildbotbuild #1371 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1371  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:17
sonne|workiglesiasg: yes they all should11:20
sonne|workiglesiasg: you know that when you leave it out it will print tons of stuff11:20
@iglesiasgit doesn't print it out here actually11:20
shogun-notifier-shogun: Fernando Iglesias :develop * 8974b58 / examples/undocumented/octave_modular/metric_lmnn_modular.m: https://github.com/shogun-toolbox/shogun/commit/8974b58e97409165f3d40f380c2e849a850595c911:20
shogun-notifier-shogun: LMNN example for octave modular11:20
shogun-notifier-shogun: Soeren Sonnenburg :develop * 65d35cd / examples/undocumented/octave_modular/metric_lmnn_modular.m: https://github.com/shogun-toolbox/shogun/commit/65d35cdd4f141eddbfb5c1461a3c4244cbe603b511:20
shogun-notifier-shogun: Merge pull request #1506 from iglesias/lmnn_octave11:20
shogun-notifier-shogun:11:20
shogun-notifier-shogun: LMNN example for octave modular11:20
sonne|workiglesiasg: that is weird! tit hsould!11:20
sonne|workit should11:20
sonne|workiglesiasg: I couldn't reproduce the octave crasher either nor could pickle2711:21
sonne|workso I guess it is some local issue11:21
@iglesiasgoh, ok11:21
@iglesiasgit does not prevent me from working with octave so I can live with it11:21
sonne|workiglesiasg: you should try if a recent swig version would fix it and also that you don't have some clashes w/ other versions of libshogun11:22
sonne|workand that you did compile on *this* machine you run octave on11:22
@iglesiasgsonne|work, I compiled in the machine I am running octave, sure11:24
lisitsynsonne|work: iglesiasg: what would you use with boolean features?11:29
lisitsynsupervised classification11:29
@iglesiasgmmm let's see11:29
lisitsynsay 128 bits11:30
@iglesiasgwhy not an SVM?11:30
lisitsyniglesiasg: I didn't say anything about it so far ;)11:30
lisitsynlinear SVM?11:31
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:7c32:76f3:b1c5:6634] has quit [Quit: Ex-Chat]11:31
-!- iglesiasg_ [~iglesias@dyna218-109.csc.kth.se] has joined #shogun11:31
lisitsyniglesiasg_: linear SVM?11:31
iglesiasg_I pulled out the cable accidentally, sorry11:31
iglesiasg_I would try first with linear yeah11:31
lisitsyniglesiasg_: what if semi-supervised?11:32
-!- iglesiasg_ is now known as iglesiasg11:32
iglesiasglisitsyn, semi-supervised like there will be some examples in the training data that are not labelled, right?11:32
lisitsyniglesiasg: yeah11:33
lisitsyniglesiasg: well I have keypoints11:33
lisitsynSIFT/SURF whatever11:33
iglesiasglisitsyn, Nico mentioned in Berlin that he was working in a solver for that11:33
iglesiasgI am not sure but I think it was SVM too11:33
shogun-buildbotbuild #1590 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1590  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>11:40
sonne|worklisitsyn: binary classification -> SVM11:44
lisitsynsonne|work: haha11:50
lisitsynsounds like the only method?11:50
lisitsynwell that's what I was going to use indeed11:51
iglesiasglisitsyn, what do you think about some distance based method like KNN?11:51
lisitsyniglesiasg: need real-time11:51
iglesiasgusing trees it should be possible11:51
lisitsynso either linear explicit feature map or linear11:51
lisitsynI'd like to save some time there11:51
lisitsynsome preprocessing eats CPU too11:52
shogun-buildbotbuild #1673 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1673  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>11:58
shogun-buildbotbuild #1674 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1674  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>12:16
shogun-buildbotbuild #1372 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1372  blamelist: Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>12:18
-!- travis-ci [~travis-ci@ec2-174-129-83-121.compute-1.amazonaws.com] has joined #shogun12:26
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1092597512:26
-!- travis-ci [~travis-ci@ec2-174-129-83-121.compute-1.amazonaws.com] has left #shogun []12:26
-!- iglesiasg [~iglesias@dyna218-109.csc.kth.se] has quit [Quit: Ex-Chat]12:32
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:7406:eba4:9b94:874] has joined #shogun12:43
-!- mode/#shogun [+o iglesiasg] by ChanServ12:43
-!- HeikoS [~heiko@nat-163-201.internal.eduroam.ucl.ac.uk] has joined #shogun13:41
-!- mode/#shogun [+o HeikoS] by ChanServ13:41
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun13:43
shogun-notifier-shogun: Heiko Strathmann :develop * 4238c59 / tests/unit/mathematics/logdet/RationalApproximation_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/4238c59ef52fc4ef96fdf7b1547f1298486636a713:57
shogun-notifier-shogun: add test again and disable13:57
shogun-notifier-shogun: Heiko Strathmann :develop * 61a925b / tests/unit/mathematics/logdet/RationalApproximation_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/61a925b4c4b61d2a2c92fb9bf4e9bc9976a3a19213:57
shogun-notifier-shogun: Merge pull request #1518 from karlnapf/develop13:57
shogun-notifier-shogun:13:57
shogun-notifier-shogun: add test again and disable13:57
votjakovrHeikoS: ;)14:06
@HeikoSvotjakovr: I am sometimes very lazy and its good that people tell me :D14:06
shogun-buildbotbuild #1591 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1591  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:08
shogun-notifier-shogun: Heiko Strathmann :develop * 2814c54 / src/shogun/mathematics/logdet/NormalSampler.cpp: https://github.com/shogun-toolbox/shogun/commit/2814c548d1bfaeb846c3ac539f2f32e5cc1eeff914:11
shogun-notifier-shogun: remove dimension check for normal trace sampler14:11
shogun-notifier-shogun: Heiko Strathmann :develop * e23afd3 / src/shogun/mathematics/logdet/ProbingSampler.cpp: https://github.com/shogun-toolbox/shogun/commit/e23afd310be73e93e51b65105815a7714406bd1714:11
shogun-notifier-shogun: make index warning an assertion to prevent wrong samples14:11
shogun-notifier-shogun: Heiko Strathmann :develop * b9403e9 / src/shogun/mathematics/logdet/ (2 files): https://github.com/shogun-toolbox/shogun/commit/b9403e95d387a0f9a9e098f16207826e3f16bde714:11
shogun-notifier-shogun: Merge pull request #1519 from karlnapf/develop14:11
shogun-notifier-shogun:14:11
shogun-notifier-shogun: Remove some unnecessary warnings and add REQUIRE where its needed14:11
shogun-buildbotbuild #1592 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1592  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:14
shogun-buildbotbuild #1970 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/1970  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:17
gsomixRE14:17
@HeikoSsonne|work, iglesiasg the buildbot links are broken14:19
@HeikoSboth from the readme in git and the messages that appear here14:19
@iglesiasgHeikoS, it seems that the messages here are missing a slash ir orgbuilders14:20
shogun-buildbotbuild #1971 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/1971  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:20
@iglesiasgand we should update the README14:20
@HeikoSiglesiasg: any idea about those failing builds ^14:20
@HeikoSI have no idea what causes this14:20
@iglesiasgHeikoS, me neither, ensemble unit tests14:21
@HeikoSiglesiasg: and whats with the python integration tests failing all of a sudden?14:21
@HeikoSiglesiasg: thats since here14:21
@HeikoShttps://travis-ci.org/shogun-toolbox/shogun/builds/1089016514:21
@iglesiasgyeah I noticed that we lost travis yesterday14:22
@HeikoSstring file modular14:22
@HeikoSargh these integration tests cause nothing but annoyance14:22
@HeikoSany idea why it fails?14:22
@iglesiasgno idea at all14:23
@iglesiasgit makes no sense for me14:23
@HeikoSTraceback (most recent call last):  File "tester.py", line 182, in <module>    got=get_split_string(f[1])  File "tester.py", line 106, in get_split_string    strs.extend(e.replace('\\n','\n').splitlines())AttributeError: 'list' object has no attribute 'replace'14:23
@HeikoSno attribute replace?14:23
@iglesiasgincluding an octave modular example breaks python modular integration test14:24
@iglesiasgit is crazy14:24
@HeikoSprobably some python things changed in travis or so14:24
@iglesiasgso is it broken in all the builds from that one_14:24
@iglesiasg?14:24
@HeikoSFile "tester.py", line 106, in get_split_string14:24
@HeikoSstrs.extend(e.replace('\\n','\n').splitlines())14:25
@iglesiasgyou are right, it seems to be about Python14:25
@iglesiasglet me ask in travis IRC channel14:26
@HeikoShttps://github.com/shogun-toolbox/shogun/commit/42ca7de15c98c26cf75ad10e5b56109292058d6914:26
@HeikoSthis is the commit responsible14:26
@HeikoSsonne|work: any idea?14:27
@iglesiasgaaah I see14:27
@HeikoSalso, rather than using '\n', one should use os.linesep14:27
@HeikoSfor platform safety14:27
@iglesiasgHeikoS, mm but I don't think it is just that commit14:28
@iglesiasgthat was made 4 months ago14:28
@iglesiasgand travis have been green since that14:29
@HeikoSiglesiasg: I know, but that is the line14:29
@iglesiasggot it14:29
@HeikoSso something probably has changed in travis14:29
@HeikoSand now this idea doesnt work anymore14:29
@iglesiasgthat is it14:29
@iglesiasgI think your idea using os.linesep should be the right way14:30
votjakovrHeikoS: do you have any ideas, why we are using MMA (Method of Moving Asymptotes) instead of BFGS? NLopt provides both of them14:32
@HeikoSvotjakovr: for ML2?14:32
votjakovrHeikoS: yep14:32
@HeikoSvotjakovr: I have no idea why this decision was made ;)14:32
@HeikoSwhat is used in GPML and GPStuff?14:32
votjakovrHeikoS: BFGS in GPML14:33
@HeikoSvotjakovr: do you know about this mma method?14:33
votjakovrHeikoS: not yet :)14:33
@HeikoSdo you know how they compare to each other?14:33
@HeikoSvotjakovr: http://en.wikipedia.org/wiki/Method_of_moving_asymptotes14:34
@HeikoShaha14:34
@HeikoShttp://onlinelibrary.wiley.com/doi/10.1002/nme.1620240207/abstract;jsessionid=A02B054610BF4FAA529A002CFC023A35.d03t0414:34
votjakovrHeikoS: thanks, i'll check14:35
@HeikoSvotjakovr: I have no idea14:35
@HeikoSvotjakovr: is there any documentaion about the decision?14:35
votjakovrHeikoS: nope14:35
@HeikoSvotjakovr: is it easy to switch between the methods?14:35
@HeikoSvotjakovr: maybe we can add an enum that selects the method14:35
votjakovrHeikoS: yeah, just change enum14:35
@HeikoSvotjakovr: then we can offer both and set a default value14:36
@HeikoSvotjakovr: we can compare both methods on some easy and hard problems and then see14:36
votjakovrHeikoS: yeah sure14:36
@HeikoSvotjakovr: ok cool14:37
votjakovrHeikoS: http://ab-initio.mit.edu/wiki/index.php/NLopt_Algorithms14:37
votjakovrHeikoS: we also could use other algorithms14:37
votjakovrHeikoS: check local gradient-based optimization section14:38
@HeikoSvotjakovr: I dont know too much about these unconstrained optimization problems14:38
@HeikoSvotjakovr: so I would rather stick with the decision made in GPML14:38
votjakovrHeikoS: yeah, me too, i'd like to try to get the same result with GPML and then try some other methods and see the differences14:39
@HeikoSvotjakovr: sounds like a good plan?14:40
votjakovrHeikoS: yep, then i'll change default method to BFGS14:40
@HeikoSvotjakovr: ok, go ahead! btw are things already starting to work?14:42
votjakovrHeikoS: sorry, not yet, working on it14:42
@HeikoSvotjakovr: ok, let me know how things develop :)14:43
votjakovrHeikoS: btw i also changed constructor in gradient model selection, which allows not to pass the tree14:45
votjakovrHeikoS: and use all the parameters14:45
@HeikoSvotjakovr: nice!14:45
@HeikoSvotjakovr: does this "using all parameters" already work?14:45
votjakovrHeikoS: not yet :)14:46
shogun-buildbotbuild #1373 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1373  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:48
shogun-buildbotbuild #1676 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1676  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:49
@HeikoSvotjakovr: but thats a very useful feature :)14:49
votjakovrHeikoS: yeah, working on it... btw do you think that, we need to do this: http://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/modelselection/GradientModelSelection.cpp#L18714:50
@HeikoSwhat does it do? compare gradients against finite difference approximation?14:50
@HeikoSvotjakovr: I think this should be moved to unit tests, they did not exist last year14:51
votjakovrHeikoS: yeah, great, i think so14:52
@HeikoSvotjakovr: but we should do it in a gernal way14:52
@HeikoSvotjakovr: imagine people implement the gradients in a wrong way14:52
@HeikoSvotjakovr: we should check for this14:53
@HeikoSvotjakovr: so woul dbe good to have a class that does this for all classes that implement gradients14:53
votjakovrHeikoS: DifferentiableFunction ?14:53
@HeikoSvotjakovr: yes14:53
@HeikoSvotjakovr: what do you think?14:53
votjakovrHeikoS: yeah i think that is good, to move these functionality there14:54
votjakovrHeikoS: i mean there: http://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/evaluation/DifferentiableFunction.h14:59
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has joined #shogun15:09
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/1093034915:09
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has left #shogun []15:09
shogun-buildbotbuild #1374 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1374  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>15:12
shogun-buildbotbuild #1675 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1675  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>16:23
-!- travis-ci [~travis-ci@ec2-107-20-109-156.compute-1.amazonaws.com] has joined #shogun16:23
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/1093081616:23
-!- travis-ci [~travis-ci@ec2-107-20-109-156.compute-1.amazonaws.com] has left #shogun []16:23
-!- sonne|osx [~sonne@89.204.139.238] has joined #shogun16:58
shogun-notifier-shogun: Soeren Sonnenburg :develop * 7ee535c / examples/undocumented/r_modular/ (54 files): https://github.com/shogun-toolbox/shogun/commit/7ee535c03b6e24635ce469a66f558e056a4cdf4016:59
shogun-notifier-shogun: rename shogun -> modshogun16:59
-!- lambday [67157e4c@gateway/web/freenode/ip.103.21.126.76] has joined #shogun17:02
lambdayHeikoS: hi17:02
lambdayHeikoS: why it didn't call precompute? :(17:04
lambdayHeikoS: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/mathematics/logdet/LogDetEstimator.cpp#L8617:04
lambdaycalling sample on LogDetEstimator does that, right?17:04
lambdayHeikoS: alright moving the tests into different files17:05
shogun-buildbotbuild #1972 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/197217:07
shogun-buildbotbuild #78 of osx1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/osx1%20-%20libshogun/builds/78  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Soeren Sonnenburg <sonne@debian.org>, Fernando Iglesias <fernando.iglesiasg@gmail.com>17:08
shogun-notifier-shogun: Soeren Sonnenburg :develop * 6d8ae8a / README: https://github.com/shogun-toolbox/shogun/commit/6d8ae8a881f2e1e5bc3f97df26f1bda24c51590117:11
shogun-notifier-shogun: fix buildbot url17:11
shogun-buildbotbuild #1593 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1593  blamelist: Soeren Sonnenburg <sonne@debian.org>17:12
shogun-buildbotbuild #79 of osx1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/osx1%20-%20libshogun/builds/79  blamelist: Soeren Sonnenburg <sonne@debian.org>17:12
shogun-buildbotbuild #1973 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/1973  blamelist: Soeren Sonnenburg <sonne@debian.org>17:15
* gsomix away17:16
-!- sonne|osx [~sonne@89.204.139.238] has quit [Quit: sonne|osx]17:19
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Good bye!]17:20
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]17:32
shogun-buildbotbuild #1375 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1375  blamelist: Soeren Sonnenburg <sonne@debian.org>17:36
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun17:41
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun []17:41
shogun-buildbotbuild #1677 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1677  blamelist: Soeren Sonnenburg <sonne@debian.org>17:49
-!- travis-ci [~travis-ci@ec2-107-20-109-156.compute-1.amazonaws.com] has joined #shogun17:53
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1093749417:53
-!- travis-ci [~travis-ci@ec2-107-20-109-156.compute-1.amazonaws.com] has left #shogun []17:53
lambdayHeikoS: going for dinner.. will be back within an hour or so18:02
lambdayHeikoS: see you18:02
-!- lambday [67157e4c@gateway/web/freenode/ip.103.21.126.76] has quit []18:02
-!- travis-ci [~travis-ci@ec2-54-224-109-62.compute-1.amazonaws.com] has joined #shogun18:18
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1093796718:18
-!- travis-ci [~travis-ci@ec2-54-224-109-62.compute-1.amazonaws.com] has left #shogun []18:18
shogun-notifier-shogun: lambday :develop * c5ac03a / / (6 files): https://github.com/shogun-toolbox/shogun/commit/c5ac03ae86d512106d9d87a4b69bbe949d1d14f518:28
shogun-notifier-shogun: minor changes and added unit-test (log-det)18:28
shogun-notifier-shogun: Heiko Strathmann :develop * cdc196e / / (6 files): https://github.com/shogun-toolbox/shogun/commit/cdc196ebeb72c7df8233dab2609283b9b866ee5e18:28
shogun-notifier-shogun: Merge pull request #1517 from lambday/feature/log_determinant18:28
shogun-notifier-shogun:18:28
shogun-notifier-shogun: minor changes and added unit-test (log-det)18:28
shogun-buildbotbuild #1974 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/1974  blamelist: lambday <heavensdevil6909@gmail.com>18:30
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun19:02
@HeikoSthoralf: around?19:10
@HeikoSthoralf: did you eventually add your sanity checks?19:10
shogun-buildbotbuild #1975 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb1%20-%20libshogun/builds/197519:19
pickle27HeikoS: around? Uploaded my ECG notebook!19:20
shogun-notifier-shogun-data: Kevin :master * d416560 / ica/foetal_ecg.dat: https://github.com/shogun-toolbox/shogun-data/commit/d416560c1856bee95700a35d0aabb3dd3b3208ee19:21
shogun-notifier-shogun-data: added fetal ecg data19:21
shogun-notifier-shogun-data: Fernando Iglesias :master * 89e40ce / ica/foetal_ecg.dat: https://github.com/shogun-toolbox/shogun-data/commit/89e40ceba7bb7ae7908271d66cec93828c5faf4219:21
shogun-notifier-shogun-data: Merge pull request #14 from pickle27/master19:21
shogun-notifier-shogun-data:19:21
shogun-notifier-shogun-data: added fetal ecg data19:21
shogun-notifier-shogun: Kevin :develop * 47448f7 / doc/ipython-notebooks/ica/ecg_sep.ipynb: https://github.com/shogun-toolbox/shogun/commit/47448f7c1a6c07575404f35216c1edabe47e289819:21
shogun-notifier-shogun: added ECG source separation notebook19:21
shogun-notifier-shogun: Fernando Iglesias :develop * 541273c / doc/ipython-notebooks/ica/ecg_sep.ipynb: https://github.com/shogun-toolbox/shogun/commit/541273cf55d5c6a3d447867c3878245c932fa24d19:21
shogun-notifier-shogun: Merge pull request #1520 from pickle27/develop19:21
shogun-notifier-shogun:19:21
shogun-notifier-shogun: added ECG source separation notebook19:21
@HeikoSpickle27: hi!19:22
@HeikoSpickle27: cool! I will have a look soon19:22
shogun-buildbotbuild #1678 of deb3 - modular_interfaces is complete: Failure [failed compile libshogun]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1678  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:22
pickle27HeikoS: sounds good!19:22
@HeikoSpickle27: looking forward to see it!19:22
@HeikoSpickle27: if I find time, I will also play a bit with your ICA stuff19:23
pickle27HeikoS: sweet thanks!19:23
@iglesiasgpickle27, it looks nice!19:23
@iglesiasgI put the link in here in case you want to check HeikoS19:23
@iglesiasghttp://nbviewer.ipython.org/urls/raw.github.com/pickle27/shogun/47448f7c1a6c07575404f35216c1edabe47e2898/doc/ipython-notebooks/ica/ecg_sep.ipynb19:23
pickle27ah iglesiasg already merged it!19:23
@HeikoSthanks!19:23
shogun-buildbotbuild #1594 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1594  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:24
@iglesiasgpickle27, did you upload it with the figures already rendered?19:24
pickle27iglesiasg: nope no graphics in the upload19:24
@iglesiasgpickle27, all right, cool!19:24
@iglesiasgpickle27, I think there is a typo in the third block of code, the one that generates the first figures19:26
@iglesiasgAbdomonal -> Abdominal19:26
pickle27ahh indeed19:26
pickle27okay I will fix and make a PR soon, I asked Andreas to have a look so I'll wait for his feedback19:26
@iglesiasggood19:26
@iglesiasgpickle27, and the data file, is on purpose called foetal? I mean it because of the o19:27
pickle27iglesiasg: thats the exact name of the file downloaded from the website, I just left it19:27
@iglesiasggood then!19:27
@iglesiasgpickle27, apologies for being annoying :)19:28
pickle27iglesiasg: haha no I appreciate people taking the time to look through my stuff!19:28
@iglesiasgpickle27, it is nice19:29
@iglesiasgpickle27, how is it known that the baby's heart rate is twice the mother's?19:30
@iglesiasgjust out of curiosity hehe19:30
pickle27iglesiasg: I;m not sure but apparently it is a fact19:30
@HeikoSiglesiasg: my sister is pregnant and told me a few days ago ;)19:30
@iglesiasghehe ok, so it is something well known19:31
pickle27iglesiasg: Andreas did the analysis, Im not much of a bio guy19:31
@iglesiasgHeikoS, tell her that she can see her baby's heart rate using Shogun :)19:31
pickle27HeikoS: haha yes19:31
@HeikoShaha :D19:31
@HeikoSShe is a schoolbook saleswoman, dont know whether she appreciates that ;)19:32
@iglesiasghehe19:32
shogun-notifier-shogun: Heiko Strathmann :develop * 8402a3e / src/shogun/mathematics/logdet/NormalSampler.h: https://github.com/shogun-toolbox/shogun/commit/8402a3e150a6a5f0ed08a17c1768349756826d7019:33
shogun-notifier-shogun: add comment about ignoring the idx19:33
shogun-notifier-shogun: Heiko Strathmann :develop * 0095593 / src/shogun/mathematics/logdet/LogDetEstimator.cpp: https://github.com/shogun-toolbox/shogun/commit/0095593c7b93026e638457bcf0ab2b62a691727a19:33
shogun-notifier-shogun: some minor changes for readiability and style19:33
shogun-notifier-shogun: Heiko Strathmann :develop * d391c18 / src/shogun/mathematics/logdet/ (2 files): https://github.com/shogun-toolbox/shogun/commit/d391c18294564f8f79ce55fedc21897a0f83ab1219:33
shogun-notifier-shogun: Merge pull request #1521 from karlnapf/develop19:33
shogun-notifier-shogun:19:33
shogun-notifier-shogun: some minor changes19:33
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has joined #shogun19:36
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/1094130019:36
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has left #shogun []19:36
shogun-buildbotbuild #1595 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1595  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>19:36
shogun-buildbotbuild #1596 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1596  blamelist: Kevin <kevinhughes27@gmail.com>19:45
shogun-buildbotbuild #1376 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1376  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:51
shogun-buildbotbuild #1597 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/bsd1%20-%20libshogun/builds/1597  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:52
pickle27iglesiasg: hmm I think there actually was graphics in the ecg notebook19:57
pickle27iglesiasg: I'll fix that when I resubmit it19:57
@iglesiasgpickle27, ok, thank you19:57
shogun-buildbotbuild #1056 of rpm1 - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/rpm1%20-%20libshogun/builds/1056  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:59
-!- HeikoS [~heiko@nat-163-201.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]20:00
pickle27sonney2k: wiking iglesiasg did you guys see my PR for the ruby modular interface? Just wondering what people think20:04
@iglesiasgpickle27, I didn't check that one at all20:04
pickle27okay cool, any thoughts on updating to nmatrix from narray>20:05
@iglesiasgI have no idea about that, sorry :S20:08
@iglesiasgwhat did this guy in the mailing list that seemed to know about Ruby say?20:08
pickle27iglesiasg: not much, he was advocating a different but no longer maintained numerical lib20:09
@iglesiasgmm I see20:09
pickle27for me the whole point is that nmatrix is active20:09
pickle27in general python is much preferred to ruby for science so there isn't a great numerical lib20:10
@iglesiasgaham20:10
@iglesiasgI have no real preference about this20:10
pickle27but nmatrix and sciruby are active and trying to change that20:10
@iglesiasgand honestly are you saying, I don't think that will be used a lot since Python is preferred20:11
@iglesiasgit can be good then to move to nmatrix and sciruby20:11
pickle27iglesiasg: main reason I care is because my job next year will be in ruby20:11
pickle27so just playing around really20:11
@iglesiasgpickle27, will you be doing machine learning in Ruby you think?20:11
pickle27I do think it will be the right move eventually to go with nmatrix20:12
pickle27ummm yeah I expect to be doing some machine learning20:12
pickle27my job is web dev backend20:12
shogun-buildbotbuild #1377 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/cyg1%20-%20libshogun/builds/1377  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>, Kevin <kevinhughes27@gmail.com>20:14
@iglesiasgall right, time to go home, see you later!20:20
pickle27iglesiasg: later!20:21
@iglesiasgbye bye20:21
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:7406:eba4:9b94:874] has quit [Quit: Ex-Chat]20:21
shogun-buildbotbuild #1680 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1680  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Kevin <kevinhughes27@gmail.com>20:33
shogun-buildbotbuild #1057 of rpm1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/rpm1%20-%20libshogun/builds/105720:34
shogun-buildbotbuild #1679 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/deb3%20-%20modular_interfaces/builds/1679  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>20:36
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has joined #shogun21:18
travis-ci[travis-ci] it's Fernando Iglesias'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/1094320721:18
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has left #shogun []21:18
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has joined #shogun21:38
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun22:20
-!- mode/#shogun [+o iglesiasg] by ChanServ22:20
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has joined #shogun22:26
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/1094364322:26
-!- travis-ci [~travis-ci@ec2-54-226-106-89.compute-1.amazonaws.com] has left #shogun []22:26
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]22:33
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun []22:44
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Ex-Chat]23:26
-!- sonne|osx [~sonne@f053046104.adsl.alicedsl.de] has joined #shogun23:31
shogun-buildbotbuild #80 of osx1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.orgbuilders/osx1%20-%20libshogun/builds/8023:34
-!- hushell [~hushell@c-98-232-178-161.hsd1.or.comcast.net] has quit [Ping timeout: 240 seconds]23:49
--- Log closed Wed Sep 04 00:00:44 2013