--- Log opened Thu May 09 00:00:37 2013 | ||
-!- iglesiasg [d58f32bd@gateway/web/freenode/ip.213.143.50.189] has quit [Quit: Page closed] | 00:11 | |
@sonney2k | van51, yeah remove that legacy crap | 00:13 |
---|---|---|
van51 | sonney2k: ok, got it ;) | 00:16 |
van51 | on another note, is there a reason why operator[] isn't overloaded in CDynamicObjectArray? | 00:16 |
van51 | not that it's important, it just seems strange :p | 00:17 |
@sonney2k | van51, well it is usually new'd so you would have to (*array)[1] etc | 00:19 |
@sonney2k | but feel free :D | 00:19 |
-!- goldbug [~quassel@global-2-11.nat.csx.cam.ac.uk] has joined #shogun | 00:21 | |
van51 | hehe just for consistency :D | 00:21 |
van51 | sonney2k: actually, maybe all get_*_kernel() should go, and just have get_kernel(index_t idx) | 00:28 |
van51 | and maybe a get_last_kernel | 00:30 |
@wiking | sonney2k: do we assure alignedness with SG_MALLOC? | 00:33 |
@wiking | as i can see not :( | 00:46 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 00:49 | |
-!- goldbug [~quassel@global-2-11.nat.csx.cam.ac.uk] has quit [Remote host closed the connection] | 01:04 | |
-!- foulwall [~foulwall@2001:da8:215:503:74db:1c21:5626:64bc] has joined #shogun | 01:20 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 01:28 | |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * c5a21a8 / src/shogun/mathematics/Random.cpp,src/shogun/mathematics/Random.h: https://github.com/shogun-toolbox/shogun/commit/c5a21a80ea25b41d9d7fd2dde7d961c4da52f8fd | 01:28 |
shogun-notifier- | shogun: Implement Ziggurat normal distrib sampler | 01:28 |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * d52b026 / tests/unit/statistics/QuadraticTimeMMD_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/d52b0263f7dc13c01c92e68f0731348c156c6154 | 01:28 |
shogun-notifier- | shogun: Change to EXPECT_NEAR macro | 01:28 |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * a44102d / src/shogun/mathematics/Random.cpp,src/shogun/mathematics/Random.h: https://github.com/shogun-toolbox/shogun/commit/a44102d8b355733d8c8830094cfabf7037060035 | 01:28 |
shogun-notifier- | shogun: Random: allocate vectors on heap | 01:28 |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * 68a7ab2 / tests/unit/mathematics/Random_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/68a7ab28597cb674511f4c8acadbfb03a6b35e71 | 01:28 |
shogun-notifier- | shogun: Add unit test for normal_distrib | 01:28 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun | 01:33 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7005414 | 01:33 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun [] | 01:33 | |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * c4cfc1a / src/shogun/mathematics/Random.h: https://github.com/shogun-toolbox/shogun/commit/c4cfc1ae534f396eb2c55f956789be85cc7cd005 | 01:34 |
shogun-notifier- | shogun: Random: include shogun/lib/common.h for stdint.h | 01:34 |
@wiking | doh | 01:38 |
@wiking | why there's no UINT64_MAX on ubuntu? :) | 01:38 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun | 01:40 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7005551 | 01:40 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun [] | 01:40 | |
van51 | wiking: I am on ubuntu and I have it | 01:47 |
van51 | but for c++ I had to define __STDC_CONSTANT_MACROS first | 01:53 |
naywhayare | wiking: uint64_t is a C99 type not a C++ type, maybe this is helpful: http://lists.freebsd.org/pipermail/freebsd-bugs/2012-December/051270.html :) | 02:05 |
naywhayare | (can you tell I am procrastinating on my own work by randomly answering questions unnecessarily in IRC ??) | 02:06 |
shogun-buildbot | build #334 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/334 | 03:15 |
sijin | Could anyone tell me what will be the problem when travis could not compile? | 03:16 |
shogun-buildbot | build #326 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/326 | 03:23 |
shogun-buildbot | build #389 of nightly_default is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/389 | 03:37 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 04:34 | |
sijin | lisitsyn: It seems that tapkee checks the value of num_neighbors for every embedding method | 04:59 |
sijin | lisitsyn: kpca doesn't use this , then some input would make the default value of num_neighbor (10) "invalid" | 05:03 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 05:45 | |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 264 seconds] | 06:23 | |
-!- sijin [~smuxi@144.214.222.109] has quit [] | 06:32 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 06:59 | |
-!- foulwall [~foulwall@2001:da8:215:503:74db:1c21:5626:64bc] has quit [Remote host closed the connection] | 07:04 | |
-!- nube [~rho@116.90.239.3] has quit [Read error: Connection reset by peer] | 07:32 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 07:33 | |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun | 07:34 | |
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has joined #shogun | 07:40 | |
deerishi | @sonny2k : could you take a look at this https://github.com/deerishi/shogun/compare/m1 | 07:41 |
deerishi | i did what you said,cloned another repo ,and pasted the files. | 07:42 |
deerishi | i did not touch the data,tests files | 07:42 |
deerishi | should i send the pr? | 07:42 |
deerishi | @HeikoS:could you also take a look at that? | 07:42 |
gsomix | good morning | 07:53 |
-!- sijin [~smuxi@144.214.222.109] has joined #shogun | 08:04 | |
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has quit [Ping timeout: 245 seconds] | 08:47 | |
-!- mdragan [8d55007a@gateway/web/freenode/ip.141.85.0.122] has joined #shogun | 09:06 | |
@sonney2k | naywhayare, taht is the destiny of IRC users :) | 09:07 |
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun | 09:23 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 09:23 | |
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Quit: Page closed] | 09:29 | |
deerishi | @HeikoS: I have sent the pr | 09:43 |
deerishi | it does not contain any of the data/tests broken files | 09:43 |
deerishi | @sonny2k : also the number of commits is 1 | 09:44 |
gsomix | deerishi, hi | 09:49 |
deerishi | hi gsomix | 09:49 |
gsomix | deerishi, please, sent your PR into develop branch, not master | 09:50 |
gsomix | *send | 09:50 |
gsomix | deerishi, thanks! | 09:55 |
deerishi | :) | 09:55 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds] | 10:20 | |
@wiking | sonney2k: actually when we include stdint.h why dont we have a #define __STDC_LIMIT_MACROS | 10:29 |
@wiking | so that we have access to the *_MAX macros? | 10:29 |
@wiking | or just include <cstdint> instead.... | 10:29 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 10:33 | |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * c3e8f19 / src/shogun/lib/common.h: https://github.com/shogun-toolbox/shogun/commit/c3e8f19d8212eba84727c89df90052d48a9fc7e5 | 10:33 |
shogun-notifier- | shogun: Define __STDC_LIMIT_MACROS before including stdint.h | 10:33 |
@wiking | weelll that did not help :D | 10:35 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun | 10:39 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7014634 | 10:39 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun [] | 10:39 | |
@wiking | mmm 3rd try :P | 10:46 |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * ea4dceb / src/shogun/lib/common.h,src/shogun/mathematics/Random.h: https://github.com/shogun-toolbox/shogun/commit/ea4dceb5842790d9b3aba26ad024299137bdb8b7 | 10:46 |
shogun-notifier- | shogun: Remove __STDC_LIMIT_MACROS macro and use std::numeric_limits | 10:46 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has joined #shogun | 10:51 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7014835 | 10:51 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has left #shogun [] | 10:51 | |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * e3143c5 / src/shogun/mathematics/Random.cpp: https://github.com/shogun-toolbox/shogun/commit/e3143c569f604f2ef7e26b02386e53ed3711ba04 | 10:53 |
shogun-notifier- | shogun: Use std::numeric_limit instead of macros | 10:53 |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 11:02 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 11:04 | |
-!- lambday [3ba2174f@gateway/web/freenode/ip.59.162.23.79] has joined #shogun | 11:11 | |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun | 11:12 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7014931 | 11:12 |
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun [] | 11:12 | |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 11:17 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 11:17 | |
-!- lambday [3ba2174f@gateway/web/freenode/ip.59.162.23.79] has quit [] | 11:22 | |
@lisitsyn | sijin: good point, making these checks lazy is on my todo | 11:29 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun | 11:34 | |
deerishi | @sonny2k :i have resent the pr | 11:35 |
sijin | lisitsyn: ok, currently, i just set num_neighbor of kpca to some "safe" value. btw, my pr still fails in travis-build ( I have pulled the latest develop branch) | 11:38 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds] | 11:41 | |
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has left #shogun ["PING 1368094192"] | 12:09 | |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds] | 12:19 | |
shogun-notifier- | shogun: deerishi :develop * b0a4260 / examples/ (4 files): https://github.com/shogun-toolbox/shogun/commit/b0a4260edf0bfd7ae451aaf15bb31e19ec72d623 | 12:22 |
shogun-notifier- | shogun: kpca | 12:22 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 372889d / examples/ (4 files): https://github.com/shogun-toolbox/shogun/commit/372889d49545994b1ae4f98bbaa01a1ed50b785e | 12:22 |
shogun-notifier- | shogun: Merge pull request #1078 from deerishi/mk1 | 12:22 |
shogun-notifier- | shogun: | 12:22 |
shogun-notifier- | shogun: Kernel PCA Test/Check | 12:22 |
shogun-buildbot | build #938 of bsd1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/938 blamelist: deerishi <deerishi@gmail.com> | 12:29 |
shogun-notifier- | shogun: Viktor Gal :feature/PRNG * 801bb93 / src/configure,src/shogun/mathematics/Random.cpp: https://github.com/shogun-toolbox/shogun/commit/801bb935307de36df78abd37d881e8429fdb573c | 12:36 |
shogun-notifier- | shogun: Random use d/SFMT fill array functions only when memory is aligned | 12:36 |
-!- mdragan [8d55007a@gateway/web/freenode/ip.141.85.0.122] has quit [Ping timeout: 245 seconds] | 12:39 | |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has joined #shogun | 12:40 | |
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/7016474 | 12:40 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has left #shogun [] | 12:40 | |
@wiking | sonney2k: once somebody has the time, should implement memory alignment support in shogun :P | 12:41 |
@sonney2k | wiking, but this has some side effects no? | 12:42 |
@sonney2k | if we allocate lots of small chunks | 12:42 |
@sonney2k | ? | 12:42 |
@wiking | sonney2k: well it has a sideeffect of course | 12:43 |
@wiking | overhead | 12:43 |
shogun-buildbot | build #939 of bsd1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/939 | 12:43 |
@wiking | sonney2k: i'm just saying that SFMT can only fast fill array with sse2 instructions if the memory is 16byte aligned | 12:43 |
@wiking | obviously | 12:43 |
@wiking | so i had to make a check there | 12:43 |
@wiking | as in darwin malloc is already giving you back a 16 byte aligned memory | 12:44 |
@wiking | always | 12:44 |
@wiking | on other os of course it's not like this | 12:44 |
shogun-buildbot | build #773 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/773 blamelist: Soeren Sonnenburg <sonne@debian.org> | 12:44 |
@wiking | sonney2k: but maybe we should as well add support for non system memory managers | 12:45 |
@wiking | like jmalloc | 12:45 |
@wiking | i've prepared a patch for it | 12:45 |
@wiking | i just wonder where the fuck did i put that one :DD | 12:45 |
@wiking | i guess it's either a branch or in a stash | 12:45 |
@wiking | yeah i think i got it :) | 12:45 |
@wiking | anyhow the PRNG feature is about to be ready | 12:46 |
@wiking | jemalloc and hoard is much faster | 12:47 |
@wiking | then the usual gnu malloc on linux | 12:47 |
@wiking | sonney2k: http://www.hoard.org/ | 12:49 |
@wiking | http://plasma.cs.umass.edu/emery/screenshots | 12:49 |
@sonney2k | wiking, speed is always welcome ... aligned_alloc() | 12:49 |
@wiking | sonney2k: yeah so i'll push that configure patch for replacing the default system malloc | 12:50 |
@wiking | i just need to test them | 12:50 |
@wiking | but this prng should be finished as well | 12:51 |
@sonney2k | wiking, one thing at a time | 12:51 |
@wiking | yeah tell me about it | 12:51 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun | 12:52 | |
@sonney2k | wiking, if you do this with aligned_alloc - we need to tell gcc somehow that the memory is aligned | 12:52 |
@sonney2k | otherwise it won't fully benefit from it | 12:52 |
@wiking | sonney2k: posix_memalign | 12:52 |
@wiking | we should use that rather | 12:52 |
@wiking | btw how we should test the new PRNG? | 12:53 |
@wiking | i mean now i think it's there | 12:53 |
@wiking | it's ready | 12:53 |
@wiking | but we need to fix the integration tests | 12:53 |
@wiking | ../data/clustering/KMeans_EuclidianDistance_3.mERROR | 12:54 |
@wiking | libshogun (x86_64/git_801bb93_2013-05-09_12:36134331283) Copyright (C) 1999-2009 Fraunhofer Institute FIRST Copyright (C) 1999-2011 Max Planck Society Copyright (C) 2009-2011 Berlin Institute of Technology Copyright (C) 2012 Soeren Sonnenburg, Sergey Lisitsyn, Heiko Strathmann Written (W) 1999-2012 Soeren Sonnenburg, Gunnar Raetsch et al. ( configure options: "configure options --disable-optimization --interfaces=python_static" compile flags: "-fPIC -g | 12:54 |
@wiking | as u can see this fails already | 12:54 |
@sonney2k | wiking, yeah makes sense | 12:54 |
@sonney2k | k-means does random init | 12:54 |
@wiking | yep | 12:54 |
@wiking | i know | 12:54 |
@wiking | but we need to fix that | 12:54 |
@wiking | and as well other integration tests | 12:54 |
@sonney2k | well once you have unit tests we can just regenerate the expected data | 12:54 |
@wiking | unit tests? | 12:54 |
@wiking | they are done :) | 12:54 |
@wiking | https://travis-ci.org/shogun-toolbox/shogun/jobs/7016700 | 12:55 |
deerishi | finally merged 1 PR!!!!!!!!!!!!!! | 12:55 |
deerishi | feels gr8!!! | 12:55 |
@wiking | deerishi: yeah tell me about it | 12:55 |
@wiking | deerishi: my mailbox is full with your PR attempts ;P | 12:55 |
deerishi | haha | 12:56 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has joined #shogun | 12:56 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7016698 | 12:56 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has left #shogun [] | 12:56 | |
deerishi | will start working on global alignment kernels | 12:56 |
deerishi | that issue is still open rite? | 12:58 |
shogun-buildbot | build #1070 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1070 blamelist: deerishi <deerishi@gmail.com> | 13:05 |
@wiking | sonney2k: so, merge? | 13:09 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds] | 13:12 | |
shogun-buildbot | build #1071 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1071 blamelist: Soeren Sonnenburg <sonne@debian.org> | 13:14 |
shogun-notifier- | shogun: Viktor Gal :develop * c521a6d / .gitignore: https://github.com/shogun-toolbox/shogun/commit/c521a6d96b8abae3294aa45e98d0dc4a5007cf88 | 13:20 |
shogun-notifier- | shogun: Add .DS_Store to gitignore | 13:20 |
shogun-buildbot | build #774 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/774 blamelist: Viktor Gal <viktor.gal@maeth.com> | 13:32 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has joined #shogun | 13:38 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/7017472 | 13:38 |
-!- travis-ci [~travis-ci@ec2-23-22-231-136.compute-1.amazonaws.com] has left #shogun [] | 13:38 | |
-!- foulwall [~foulwall@li379-21.members.linode.com] has joined #shogun | 13:52 | |
shogun-buildbot | build #1072 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1072 blamelist: Viktor Gal <viktor.gal@maeth.com> | 13:57 |
@sonney2k | wiking, yeah once the current errors are fixed | 14:36 |
@sonney2k | hmmhh MCLDA cannot work | 14:36 |
@sonney2k | pickle27, didn't you do MCLDA :D | 14:37 |
@sonney2k | look at multiclass/MCLDA.cpp:282 | 14:37 |
@wiking | ok fix it fuck | 14:52 |
@wiking | :> | 14:52 |
@wiking | who broke it? :) | 14:53 |
-!- lambday [3ba21749@gateway/web/freenode/ip.59.162.23.73] has joined #shogun | 15:01 | |
lambday | hi HeikoS | 15:04 |
lambday | hello lisitsyn | 15:04 |
-!- abinash [75c8e12e@gateway/web/freenode/ip.117.200.225.46] has joined #shogun | 15:40 | |
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has joined #shogun | 15:46 | |
-!- foulwall [~foulwall@li379-21.members.linode.com] has quit [Remote host closed the connection] | 15:58 | |
-!- abinash [75c8e12e@gateway/web/freenode/ip.117.200.225.46] has quit [Ping timeout: 245 seconds] | 15:58 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 16:09 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 16:20 | |
@wiking | votjakovr: ping | 16:38 |
-!- lambday [3ba21749@gateway/web/freenode/ip.59.162.23.73] has quit [] | 16:38 | |
votjakovr | wiking: Hi | 16:39 |
@wiking | lisitsyn: ping | 16:40 |
@wiking | votjakovr: with one of your commit we've got some things broken | 16:40 |
@wiking | i mean tests | 16:40 |
votjakovr | wiking: Do you mean travis-ci fault? | 16:41 |
@wiking | votjakovr: yes | 16:42 |
@wiking | votjakovr: suddenly gaussian regression started to fail | 16:42 |
@wiking | as well as charp modular cannot be compiled | 16:42 |
@wiking | lisitsyn: do you know what's DSYEV failed with code 175 is? | 16:45 |
votjakovr | wiking: but on my machine gaussian regression test is OK | 16:45 |
@wiking | i mean what is actually error code 175 | 16:45 |
@wiking | votjakovr: not on travis | 16:45 |
@wiking | votjakovr: what's your machine btw | 16:45 |
votjakovr | wiking: intel core i7-860, Ubuntu 13.04 | 16:46 |
@wiking | mmmhm | 16:46 |
@wiking | and when u run make tests | 16:46 |
@wiking | the python gaussian regression runs fine | 16:46 |
votjakovr | wiking: I run tester.py and it is OK | 16:47 |
@wiking | votjakovr: in shogun/src | 16:48 |
@wiking | just type make tests | 16:48 |
votjakovr | wiking: It is failed on classifier_svmlight_string_features_precomputed_kernel and library_map libshogun examples | 16:51 |
-!- gsomix [~gsomix@188.168.3.6] has quit [Remote host closed the connection] | 16:52 | |
@wiking | mmm interesting | 16:54 |
votjakovr | wiking: And it stops on libshogun examples, that is why I run tester.py | 16:55 |
@wiking | votjakovr: well then obviously there's more problem | 16:56 |
votjakovr | wiking: these libshogun examples reports: undefined reference to symbol 'pthread_spin_init@@GLIBC_2.2.5' | 16:57 |
@wiking | aaaah | 16:57 |
@wiking | votjakovr: can u check configure.log | 16:58 |
@wiking | if u still have it | 16:58 |
@wiking | if not then run again ./congiure | 16:58 |
@wiking | and check configure.log for: spinlock | 16:58 |
@wiking | damn fuck.... i have to fix that :) | 16:59 |
votjakovr | wiking: 'Result is: yes' | 17:01 |
@wiking | votjakovr: yeah i figured | 17:01 |
@wiking | there's some linking problem there | 17:02 |
@wiking | as u supposed to have spinlock | 17:02 |
@wiking | in configure.log | 17:02 |
@wiking | what is the cc line? | 17:02 |
@wiking | sonney2k: darwin version of spinlock: http://developer.apple.com/library/mac/#documentation/Darwin/Reference/ManPages/man3/spinlock.3.html | 17:04 |
votjakovr | cc -I/usr/include/json -I/usr/include/libxml2 -I/usr/include/atlas -isystem /usr/include/eigen3 -fPIC -g -Wall -Werror-implicit-function-declaration -Wno-unused-parameter -Wformat -Wformat-security -Wparentheses -Wshadow -Wno-unknown-pragmas -Wno-deprecated -O9 -fexpensive-optimizations -frerun-cse-after-loop -fcse-follow-jumps -finline-functions -fschedule-insns2 -fthread-jumps -fforce-addr -fstrength-reduce -funroll | 17:05 |
votjakovr | -loops -mfpmath=sse -march=native -mtune=native -pthread ./configure-25615-10652.c -o ./configure-2697-10652.o -Xlinker --no-undefined -pthread -lm -pthread -lhdf5 -ljson -lxml2 -llapack -lcblas -llapack_atlas -larpack -lnlopt -lglpk -lm -llzo2 -lsnappy -lz -lbz2 -llzma | 17:05 |
pickle27 | sonney2k: oh god | 17:06 |
votjakovr | wiking: It is from spinlock checking section | 17:06 |
@wiking | votjakovr: yep | 17:06 |
pickle27 | sonney2k: what a terrible mistake my bad | 17:06 |
@wiking | votjakovr: i wonder why you get undefined reference to symbol 'pthread_spin_init@@GLIBC_2.2.5' | 17:06 |
@wiking | since with these flags you managed to compile that small example and actually call pthread_spin_init | 17:07 |
votjakovr | wiking: libshogun example also writes this: 'note: 'pthread_create@@GLIBC_2.2.5' is defined in DSO /lib/x86_64-linux-gnu/libpthread.so.0 so try adding it to the linker command line' | 17:08 |
votjakovr | wiking: and this: 'note: 'pthread_spin_init@@GLIBC_2.2.5' is defined in DSO /lib/x86_64-linux-gnu/libpthread.so.0 so try adding it to the linker command line' | 17:09 |
@wiking | aaahaaa | 17:10 |
@wiking | i wonder what is linked then | 17:10 |
pickle27 | sonney2k: sent a PR to fix that embarrassing mistake, I'll be looking into breaking up the train function, making a unit test and an Eigen3 version over the next bit | 17:12 |
votjakovr | wiking: Btw csharp_modular is OK on my last PR. | 17:19 |
-!- foulwall [~foulwall@2001:da8:215:c252:d0e6:bc9e:72f:96d9] has joined #shogun | 17:26 | |
votjakovr | wiking: When I added -lpthread and -L/lib/x86_64-linux-gnu options, these two examples was linked without errors. | 17:51 |
@lisitsyn | wiking: hey back | 17:51 |
@lisitsyn | hey, back I mean | 17:51 |
@lisitsyn | dsyev 175 probably means | 17:51 |
@lisitsyn | that 175th leading minor has negative det | 17:52 |
@lisitsyn | ahh no | 17:52 |
@lisitsyn | it is tridiagonalization error | 17:52 |
@lisitsyn | wiking: I guess it may happen if matrix is not symmetric | 17:53 |
@wiking | lisitsyn: any suggestions? :) | 17:54 |
@wiking | lisitsyn: i mean now with using PruneVarSubMean finally MulticlassLibLinear started to give me reasonable results | 17:54 |
@lisitsyn | wiking: assure it is symmetric | 17:54 |
@lisitsyn | let me get back to log | 17:54 |
@lisitsyn | where does it happen? | 17:54 |
@wiking | lisitsyn: well i load in the data | 17:54 |
@wiking | i mean data matrix -> realfeatures | 17:55 |
@wiking | and then i try to do the pca | 17:55 |
@wiking | then basically crashes with that assert | 17:55 |
@lisitsyn | ohh interesting | 17:55 |
@lisitsyn | wiking: http://starlogs.net/#shogun-toolbox/shogun | 17:57 |
pickle27 | starlogs is awesome! | 17:57 |
@lisitsyn | haha | 17:57 |
@lisitsyn | I've never seen it before | 17:58 |
pickle27 | there is also http://www.commitlogsfromlastnight.com/ | 17:58 |
@wiking | lisitsyn: hahaha | 17:59 |
-!- foulwall [~foulwall@2001:da8:215:c252:d0e6:bc9e:72f:96d9] has quit [Remote host closed the connection] | 18:23 | |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun | 18:53 | |
@lisitsyn | HeikoS: why clusters? | 18:55 |
@lisitsyn | I don't get your comment here :) | 18:55 |
@lisitsyn | about KNN and modsel | 18:55 |
@HeikoS | uh | 18:55 |
@HeikoS | wait | 18:55 |
@HeikoS | wasnt that KMeans? :) | 18:55 |
@HeikoS | damn | 18:55 |
@HeikoS | I always switch those | 18:55 |
@HeikoS | let me delete the comment quickly | 18:55 |
@lisitsyn | haha! | 18:56 |
@lisitsyn | I thought so | 18:56 |
@lisitsyn | HeikoS: I think the idea on that issue is that we just can compute nearest neighbors once for maximum K | 18:56 |
@HeikoS | lisitsyn: sorry :) | 18:56 |
@lisitsyn | and later we are much faster with doing classification | 18:56 |
@HeikoS | lisitsyn: explain | 18:57 |
@lisitsyn | HeikoS: if we computed neighbors in train set for test set | 18:57 |
@lisitsyn | with K set to maximum of some range | 18:57 |
@lisitsyn | we can do classification rather fast - we just select precomputed indices | 18:57 |
@lisitsyn | we can even go with binary search here | 18:58 |
@HeikoS | lisitsyn: I cannot find the issue | 18:58 |
@HeikoS | how to search issues | 18:58 |
@HeikoS | lisitsyn: I see | 18:58 |
@lisitsyn | HeikoS: https://github.com/shogun-toolbox/shogun/issues/468 | 18:59 |
@HeikoS | but who cares about knn? | 18:59 |
@lisitsyn | haha | 18:59 |
@lisitsyn | good question | 18:59 |
@HeikoS | good idea still | 18:59 |
@HeikoS | but not really relevant for anyone | 18:59 |
@HeikoS | ah, long day here :) | 19:01 |
@HeikoS | lisitsyn: currently working on something for NIPS. And the title is MCMC Hammer :D | 19:01 |
@lisitsyn | HeikoS: ha! I have holiday today :P | 19:01 |
@lisitsyn | hahahahaha | 19:01 |
@lisitsyn | you can't touch this | 19:01 |
@HeikoS | exactly :D | 19:01 |
@lisitsyn | tadatada | 19:02 |
@lisitsyn | you can't touch this | 19:02 |
@HeikoS | will be the best sampler ever | 19:02 |
@lisitsyn | can't touch this | 19:02 |
@HeikoS | lisitsyn: do you know the original | 19:02 |
@HeikoS | I like that even more | 19:02 |
@HeikoS | rick james - superfreak | 19:02 |
@lisitsyn | superfreak superfreak! | 19:03 |
@lisitsyn | she's superfreaky | 19:03 |
@lisitsyn | can't touch this! | 19:04 |
@HeikoS | haha :) | 19:04 |
@lisitsyn | HeikoS: my mind is in off state now | 19:04 |
@lisitsyn | wiking: u can't touch this! | 19:04 |
@lisitsyn | HeikoS: can't touch this markov chain | 19:05 |
@lisitsyn | I can't touch shogun | 19:06 |
@HeikoS | hehe, glad you like this :) | 19:07 |
@lisitsyn | HeikoS: u can't touch this paper | 19:07 |
@HeikoS | grove on blackburn! | 19:07 |
@lisitsyn | HeikoS: you may say that to your supervisor | 19:08 |
@lisitsyn | u can't touch this | 19:08 |
@lisitsyn | :D | 19:08 |
votjakovr | wiking: So what do you think about how to solve the problem with libshogun examples? | 19:10 |
@lisitsyn | votjakovr: u can't touch this! | 19:10 |
@lisitsyn | :D | 19:10 |
votjakovr | lisitsyn: Yeah :D | 19:11 |
@lisitsyn | HeikoS: ooh why oh why you did it to me | 19:11 |
@lisitsyn | MCMC Hammer rocks | 19:11 |
votjakovr | lisitsyn: great song! | 19:12 |
@HeikoS | just being evil :) | 19:12 |
@lisitsyn | votjakovr: you get it! | 19:12 |
@HeikoS | https://www.youtube.com/watch?v=Vv3f0QNWvWQ&feature=player_embedded | 19:13 |
@lisitsyn | HeikoS: hahaha | 19:13 |
@lisitsyn | HeikoS: btw give me any good reference on mcmc bitte schoen! | 19:16 |
@HeikoS | lisitsyn: like a tutorial? | 19:16 |
@lisitsyn | yes for an asshole like me | 19:16 |
@lisitsyn | :D | 19:16 |
@HeikoS | https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CDwQFjAB&url=http%3A%2F%2Fwww.people.fas.harvard.edu%2F~plam%2Fteaching%2Fmethods%2Fmcmc%2Fmcmc.pdf&ei=FNqLUeG4M6X40gWlwoGADw&usg=AFQjCNE13QZ-k-KHSmw_NiHzAB9Gf4hOzA&sig2=FOlbrgLY6HNNe1fdrXf8WQ&bvm=bv.46340616,d.d2k | 19:17 |
@HeikoS | I like this | 19:17 |
@lisitsyn | nice, thanks! | 19:17 |
vgorbati | lisitsyn: hi, I wanted to ask: you are creating DenseMatrix from FeatureCallback 2 times, and I also need to do this in manifold sculpting implementation - do you mind if I will extract these lines into a separate function? | 19:18 |
@lisitsyn | vgorbati: hmm where it is created 2 times? | 19:19 |
vgorbati | lisitsyn: I mean that there are 2 places where you do this - in embedPassThru and in embed for t-SNE | 19:19 |
@lisitsyn | ahh | 19:20 |
@lisitsyn | yes of course | 19:20 |
@lisitsyn | if you need it | 19:20 |
@lisitsyn | vgorbati: just for better understanding | 19:20 |
@lisitsyn | idea of using callbacks here is to give some flexibility | 19:20 |
vgorbati | lisitsyn: I am thinking to put it in utils/matrix.hpp, is it a reasonable place? | 19:20 |
@lisitsyn | vgorbati: no I think some other place would be better | 19:21 |
@lisitsyn | matrix.hpp has center thing | 19:21 |
@lisitsyn | it has nothing to do with callbacks | 19:21 |
@lisitsyn | some utils/features.hpp may be or so | 19:22 |
vgorbati | lisitsyn: ok. btw, I see centerMatrix calls in tapkee/routines/pca.hpp or tapkee/tapkee_methods.hpp, but there are no #include '.../utils/matrix.hpp' there - should it be so? | 19:24 |
votjakovr | HeikoS: Hi, I'd like to discuss something. Do you have any time? | 19:26 |
@HeikoS | votjakovr: sure, whats up? | 19:27 |
@lisitsyn | vgorbati: yeah they are just in the same compilation unit so it works | 19:28 |
@lisitsyn | but it should be here probably, yes | 19:28 |
@HeikoS | votjakovr: will be leaving work soon though, whats up? | 19:30 |
votjakovr | HeikoS: I'm not sure, but I think that there is a bug in GPML matlab toolbox in partial derivative by ell in covSEiso. It is computed like: dK(i,j)/dell = sf2 * exp(-(x(i) - x(j))^2/(2*ell^2)) * (x(i) - x(j))^2 / (ell^2), but I think it should be computed like: dK(i,j)/dell = sf2 * exp(-(x(i) - x(j))^2/(2*ell^2)) * (x(i) - x(j))^2 / (ell^*3*). What do you think about this? | 19:32 |
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has quit [Ping timeout: 245 seconds] | 19:32 | |
@HeikoS | votjakovr: let me check | 19:36 |
@HeikoS | votjakovr: whats the kernel itself? the functional form? | 19:38 |
votjakovr | HeikoS: K(i, j) = sf2 * exp(-(x(i) - x(j))^2/(2*ell^2)) | 19:40 |
@HeikoS | votjakovr: so the difference is the exponent in the ell right? | 19:43 |
votjakovr | HeikoS: yep | 19:43 |
@HeikoS | votjakovr: wow and in the GPML toolbox, there is a two, should be a bug then. I wonder why nobody discovered this so far | 19:43 |
@HeikoS | which file is it computed in? | 19:44 |
votjakovr | HeikoS: cov/covSEiso.m | 19:44 |
@HeikoS | and i is | 19:45 |
@HeikoS | the parmeter index of which the partial derivative should be computed? | 19:45 |
votjakovr | first | 19:45 |
@HeikoS | K = sf2*exp(-K/2).*K; | 19:45 |
votjakovr | yep | 19:46 |
@HeikoS | you could do a quick simple check: | 19:46 |
@HeikoS | plot the kernel for two fixed points as a 1D function of ell | 19:47 |
@HeikoS | and plot the derivative | 19:47 |
@HeikoS | seems weird that this bug is in there ... | 19:47 |
@HeikoS | votjakovr: wait | 19:49 |
@HeikoS | ah sorry | 19:50 |
@HeikoS | no I think its a bug, but make sure to plot it | 19:50 |
votjakovr | HeikoS: I think the same | 19:50 |
@HeikoS | if it doesnt make sense (and your correction K = sf2*exp(-K/2).*K/ell; makes sense), then we should report it | 19:51 |
@HeikoS | the same? | 19:51 |
@HeikoS | ah I see | 19:51 |
votjakovr | Yep, I think it should be computed like K = sf2*exp(-K/2).*K/ell | 19:52 |
@HeikoS | indeed | 19:52 |
@HeikoS | could you send me the plots of both versions once you created them? I am curious how the difference looks like. Also good for reporting the bug | 19:52 |
@HeikoS | so I guess the line was copied in shogun? :) | 19:53 |
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds] | 19:54 | |
votjakovr | Which line? | 19:54 |
@HeikoS | the kernel derivative | 19:55 |
votjakovr | So in shogun kernel derivative is computed by width = 2*ell^2 and computed correctly | 19:56 |
-!- FSCV [~FSCV@204.45.132.11] has joined #shogun | 19:56 | |
@HeikoS | okay | 19:56 |
@HeikoS | votjakovr: well, nice job that you spotted this! :) | 19:57 |
votjakovr | HeikoS: I mean in CGaussianKernel | 19:57 |
@HeikoS | votjakovr: ok saw it | 19:58 |
@HeikoS | votjakovr: so maybe you should implement a graphical example which selects the best kernel? | 19:59 |
@HeikoS | you could generate data from the GP with a known kernel size and then learn the kernel size via ML2 | 19:59 |
@HeikoS | and then plot things in python | 19:59 |
@HeikoS | this way we have 1) a nice example for modelselection with GP regression 2) evidence that things work out fine in shogun | 19:59 |
@HeikoS | how did you spot the bug btw? | 20:00 |
@HeikoS | when writing unit tests? | 20:00 |
@HeikoS | comparing against GPML? | 20:00 |
votjakovr | HeikoS: Yep :) | 20:00 |
@HeikoS | very cool that you do that! Otherwise we can never be sure whether things work. And even finding bugs in toolboxes that are referenced by thousands of people around the world. Awesome! :) | 20:01 |
@HeikoS | votjakovr: I gotta go home now, see you around! | 20:01 |
votjakovr | HeikoS: Thanks :) | 20:01 |
@HeikoS | bye! | 20:02 |
votjakovr | See you soon | 20:02 |
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has joined #shogun | 20:05 | |
@sonney2k | van51, could you please fix the build warning and split up your combine_kernels function into a couple of reasonable sub-functions? | 20:37 |
@sonney2k | the function is just to big to be parse-able | 20:37 |
@sonney2k | van51, this one here http://shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1072/steps/compile%20libshogun/logs/warnings%20%282%29 | 20:39 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 20:44 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 874f3f3 / src/shogun/multiclass/MCLDA.cpp: https://github.com/shogun-toolbox/shogun/commit/874f3f3d5cae12df993e84249129707f924bedb3 | 20:44 |
shogun-notifier- | shogun: MCLDA fixed covariance matrix normalization | 20:44 |
van51 | sonney2k: yea sure | 20:48 |
@sonney2k | votjakovr, you said it you fixed the c# issue already? | 20:48 |
@sonney2k | van51, thanks | 20:48 |
@sonney2k | wiking, whats up with the spinlock thing? | 20:48 |
van51 | soneey2k: btw as I am re-writing the CCombinedKernel to use CDynamicObjectArray, should I also change this function to return an array? | 20:49 |
van51 | or it doesn't matter? | 20:49 |
van51 | sonney2k* | 20:49 |
@sonney2k | van51, well yes but one step at a time? | 20:51 |
@sonney2k | lisitsyn, we need that spreadsheet | 20:51 |
votjakovr | sonney2k: nope, i said that my PR, it seems, didn't break c# modular compilation. | 20:52 |
@sonney2k | votjakovr, ok then let me do the fix | 20:52 |
@sonney2k | wiking, you want some defines for spinlock business & OSX? | 20:53 |
shogun-buildbot | build #775 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/775 blamelist: Soeren Sonnenburg <sonne@debian.org> | 20:57 |
votjakovr | sonney2k: Ok, i'd like to ask about failed integration tests in travis-ci. How do you think to fix this problem? | 21:04 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has joined #shogun | 21:05 | |
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: http://travis-ci.org/shogun-toolbox/shogun/builds/7030036 | 21:05 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has left #shogun [] | 21:05 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 7418ae1 / src/interfaces/modular/modshogun_ignores.i: https://github.com/shogun-toolbox/shogun/commit/7418ae1b6acd86203d958ec2ff0a960f2f6d6460 | 21:06 |
shogun-notifier- | shogun: fix c# build failure | 21:06 |
@sonney2k | votjakovr, which one? | 21:06 |
votjakovr | sonney2k: gp regression tests | 21:07 |
-!- gsomix [~Miranda@188.168.3.6] has joined #shogun | 21:07 | |
gsomix | good evening | 21:08 |
shogun-buildbot | build #776 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/776 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:09 |
votjakovr | sonney2k: I think, it works everywhere except travis | 21:09 |
@lisitsyn | HeikoS: sonney2k: please check the doc I shared | 21:14 |
shogun-buildbot | build #1073 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1073 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:21 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has joined #shogun | 21:28 | |
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/7030719 | 21:28 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has left #shogun [] | 21:28 | |
@sonney2k | votjakovr, I don't know either - travis is just some buildbot running some ubuntu virtual machine it seems | 21:43 |
shogun-buildbot | build #1074 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1074 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:46 |
@sonney2k | Huh?! | 21:49 |
@sonney2k | what is that | 21:49 |
@sonney2k | The following tests failed! | 21:49 |
@sonney2k | p | 21:49 |
@sonney2k | p ??? | 21:50 |
@lisitsyn | what? | 21:50 |
@sonney2k | http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1074/steps/test%20python_modular/logs/stdio | 21:50 |
@lisitsyn | sonney2k: I see no p in examples | 21:51 |
@lisitsyn | sonney2k: so how did that happen :D | 21:56 |
@sonney2k | no idea yet | 21:58 |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving] | 21:58 | |
-!- arcolife [arcolife@nat/redhat/x-tfkofwlsgeocvlqk] has joined #shogun | 22:06 | |
-!- iglesiasg [d58f329b@gateway/web/freenode/ip.213.143.50.155] has joined #shogun | 22:10 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 22:11 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * b12a2f1 / data: https://github.com/shogun-toolbox/shogun/commit/b12a2f1a57eaac68ecde3f449cfaa7bb86412028 | 22:17 |
shogun-notifier- | shogun: require new data version for kpca tests | 22:17 |
@sonney2k | lisitsyn, reason is kpca data not yet there and whatever mishandling in tester.py | 22:18 |
@lisitsyn | oh I see | 22:18 |
@sonney2k | lisitsyn, btw why do we have 2 kernelpca things? | 22:21 |
@sonney2k | preprocessor_kernelpca_modular.py | 22:21 |
@sonney2k | preprocessor_kpca.py ? | 22:21 |
@lisitsyn | what is preprocessor_kpca .. | 22:23 |
@lisitsyn | :D | 22:23 |
@lisitsyn | uh what a code | 22:23 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * f313ed6 / tests/integration/python_modular/tester.py: https://github.com/shogun-toolbox/shogun/commit/f313ed636e5707dacefb130fde2924c2d2272f40 | 22:23 |
shogun-notifier- | shogun: fix weird bug in tester.py occurring when no data is available | 22:23 |
@lisitsyn | sonney2k: it is you who merged it :D | 22:24 |
@lisitsyn | sonney2k: but you were under PR DoS attack | 22:24 |
@sonney2k | psst | 22:24 |
@sonney2k | don't tell anyone | 22:24 |
@lisitsyn | let me count | 22:24 |
shogun-buildbot | build #777 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/777 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:24 |
@lisitsyn | sonney2k: that guy sent you 13 pull requests | 22:25 |
@lisitsyn | no surprise you got confused | 22:25 |
@lisitsyn | :D | 22:25 |
@sonney2k | lisitsyn, do we have some more nice graphical examples now btw? | 22:25 |
@lisitsyn | sonney2k: yes of course | 22:25 |
@lisitsyn | sonney2k: graphical/inverse_covariance_estimation_demo.py | 22:26 |
@sonney2k | who did that? | 22:27 |
@lisitsyn | sonney2k: monica | 22:27 |
@lisitsyn | sonney2k: graphical/group_lasso.py | 22:27 |
@lisitsyn | sonney2k: that's cesar fuentes | 22:28 |
@sonney2k | looks very cool | 22:28 |
@sonney2k | and very short too! | 22:29 |
-!- arcolife [arcolife@nat/redhat/x-tfkofwlsgeocvlqk] has left #shogun ["Leaving"] | 22:30 | |
@wiking | fuck | 22:33 |
@wiking | that pr spammed my inbox! | 22:33 |
@lisitsyn | wiking: kpca? | 22:33 |
@wiking | y | 22:34 |
@wiking | okok | 22:34 |
@wiking | when can i merge? | 22:34 |
@lisitsyn | wiking: merge what? | 22:34 |
@wiking | prng branch | 22:34 |
@wiking | imo it's ready | 22:35 |
@lisitsyn | ah | 22:35 |
@wiking | if i see kpca one more time i'll fucking git rm it :D | 22:35 |
@lisitsyn | wiking: well it is just 13 attempts | 22:35 |
@wiking | hehehe | 22:35 |
@wiking | only | 22:35 |
@wiking | :D | 22:35 |
@lisitsyn | why so serious | 22:35 |
@wiking | well to be totally | 22:35 |
@wiking | honest | 22:35 |
@wiking | the PRNG branch is like 20+ commits | 22:36 |
@wiking | :DDDD | 22:36 |
@wiking | so it's even worse than that kpca PR attempt | 22:36 |
@lisitsyn | no, why? | 22:36 |
@wiking | too many commit :P | 22:36 |
@wiking | anyhow | 22:36 |
@wiking | when do we get develop branch into a working shape again? | 22:36 |
@lisitsyn | why not, if they are not breakers | 22:36 |
@sonney2k | wiking, I think I fixed all issues... | 22:37 |
@sonney2k | but lets wait for the BB | 22:37 |
@wiking | ok | 22:37 |
@wiking | sonney2k: if it goes OK | 22:37 |
@wiking | then let's merge prng | 22:37 |
@wiking | sonney2k: btw why spinlock is more preferred than simple mutex? | 22:38 |
@sonney2k | lisitsyn, what is an inverse covariance matrix btw? | 22:38 |
@sonney2k | wiking, faster! | 22:39 |
@lisitsyn | sonney2k: just inverse of covariance matrix | 22:39 |
@wiking | ok sonney2k i'm just reading wikipedia | 22:39 |
@sonney2k | ok what does a inverse covmatrix tell me? | 22:39 |
@sonney2k | wiking, busy loops! | 22:39 |
@lisitsyn | sonney2k: how features are related | 22:39 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has joined #shogun | 22:39 | |
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/7032639 | 22:39 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has left #shogun [] | 22:39 | |
@lisitsyn | sonney2k: lasso regularization pushes coefficients to zero | 22:40 |
@lisitsyn | so you get only strongly related features | 22:40 |
@sonney2k | wiking, so for e.g. SG_UNREF / SGVector passing around we need these to ensure the refcount is in sane state | 22:40 |
@sonney2k | lisitsyn, ohh nice | 22:41 |
@sonney2k | now the example makes more sense to me | 22:41 |
@lisitsyn | sonney2k: I wrote it one day | 22:41 |
@lisitsyn | but never had an example | 22:41 |
@lisitsyn | I just re-used some stuff from SLEP | 22:42 |
@lisitsyn | that's why | 22:42 |
@lisitsyn | :) | 22:42 |
@lisitsyn | sonney2k: group lasso is very cool too | 22:42 |
@lisitsyn | sonney2k: but this is something you understand already, yes? | 22:43 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 4a0582b / examples/undocumented/python_modular/graphical/inverse_covariance_estimation_demo.py: https://github.com/shogun-toolbox/shogun/commit/4a0582b9abce9743125a83e5cadcf389a8e59e67 | 22:43 |
shogun-notifier- | shogun: fix typo in inv cov estiamtion demo | 22:43 |
shogun-buildbot | build #778 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/778 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:45 |
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has quit [Quit: Leaving.] | 22:53 | |
shogun-buildbot | build #1075 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1075 | 22:56 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has joined #shogun | 22:57 | |
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/7032878 | 22:57 |
-!- travis-ci [~travis-ci@ec2-107-20-28-143.compute-1.amazonaws.com] has left #shogun [] | 22:57 | |
@sonney2k | wiking, it's your turn | 23:04 |
@sonney2k | wiking, buildbot is happy | 23:04 |
@sonney2k | lisitsyn, isn't group lasso just MKL? | 23:05 |
@lisitsyn | sonney2k: I can't cast it to MKL | 23:05 |
@lisitsyn | may be it is but I don't know | 23:05 |
@lisitsyn | sonney2k: L1/L2 is like | 23:06 |
@lisitsyn | sum of absolute L2 in groups | 23:06 |
@lisitsyn | so weights in each feature group is pushed towards each other | 23:06 |
@lisitsyn | while groups are pushed to zero | 23:07 |
shogun-buildbot | build #779 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/779 blamelist: Soeren Sonnenburg <sonne@debian.org> | 23:07 |
@sonney2k | yeah same for MKL | 23:08 |
@sonney2k | you have SVM (L2) in group | 23:09 |
@sonney2k | and L1 on top | 23:09 |
@sonney2k | and it is crap :D | 23:09 |
@lisitsyn | sonney2k: ahh so you mean like kernels that work just on one group of features | 23:10 |
@lisitsyn | this way yes | 23:10 |
@sonney2k | yes | 23:10 |
@lisitsyn | sonney2k: I did one experiment for georg last summer | 23:11 |
@lisitsyn | hierarchical regularization worked quite well | 23:11 |
@lisitsyn | well the data was hierarchical | 23:11 |
@lisitsyn | that's why I guess | 23:11 |
-!- travis-ci [~travis-ci@ec2-54-224-131-101.compute-1.amazonaws.com] has joined #shogun | 23:13 | |
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/7033420 | 23:13 |
-!- travis-ci [~travis-ci@ec2-54-224-131-101.compute-1.amazonaws.com] has left #shogun [] | 23:13 | |
@sonney2k | wiking, alright once you wake up do your merge | 23:45 |
@iglesiasg | good night guys | 23:47 |
-!- iglesiasg [d58f329b@gateway/web/freenode/ip.213.143.50.155] has quit [Quit: Page closed] | 23:50 | |
shogun-buildbot | build #1077 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1077 blamelist: Soeren Sonnenburg <sonne@debian.org> | 23:57 |
--- Log closed Fri May 10 00:00:38 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!