Open in new window / Try shogun cloud
--- Log opened Tue Apr 02 00:00:18 2013
-!- FSCV_ [~FSCV@204.45.132.11] has quit [Quit: Leaving]00:23
-!- shogun-buildbot_ [~shogun-bu@7nn.de] has joined #shogun01:15
-!- blackburn1 [~blackburn@188.168.14.129] has joined #shogun01:16
-!- blackburn [~blackburn@188.168.14.129] has quit [Ping timeout: 272 seconds]01:21
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Ping timeout: 272 seconds]01:21
shogun-buildbot_build #341 of nightly_default is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/34103:38
-!- lambday [b613104d@gateway/web/freenode/ip.182.19.16.77] has joined #shogun03:42
-!- lambday [b613104d@gateway/web/freenode/ip.182.19.16.77] has quit []05:17
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun08:34
shogun-notifier-shogun: Soeren Sonnenburg :master * 2234df4 / src/shogun/kernel/string/OligoStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/2234df4636c1e6061fcc19209fe279850efa735408:34
shogun-notifier-shogun: likely fix oligokernel bug08:34
shogun-buildbot_build #850 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/850  blamelist: Soeren Sonnenburg <sonne@debian.org>08:41
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun08:49
@sonney2kx08:51
sonne|worky08:51
@sonney2kz08:51
sonne|workhaha!08:51
shogun-buildbot_build #977 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/977  blamelist: Soeren Sonnenburg <sonne@debian.org>09:17
blackburn1sonne|work: what is going on :D10:00
sonne|workblackburn1: I checked some old version of oligo and that was still giving reasonable results10:08
-!- blackburn1 [~blackburn@188.168.14.129] has left #shogun []10:29
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun10:50
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]11:34
-!- lambday [b613104f@gateway/web/freenode/ip.182.19.16.79] has joined #shogun14:10
-!- heiko [~heiko@027fc803.bb.sky.com] has joined #shogun14:58
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun15:10
-!- heiko [~heiko@027fc803.bb.sky.com] has quit [Quit: Leaving.]15:26
-!- lambday [b613104f@gateway/web/freenode/ip.182.19.16.79] has quit [Ping timeout: 245 seconds]15:57
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:27
-!- hoijui [~hoijui@141.23.93.47] has joined #shogun16:27
-!- heiko [~heiko@027fc803.bb.sky.com] has joined #shogun16:39
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Quit: Page closed]17:06
-!- ZeThomas [~thomaspr@2a02:2c40:100:a000:2810:8d1f:f379:1d82] has joined #shogun17:42
-!- ZeThomas [~thomaspr@2a02:2c40:100:a000:2810:8d1f:f379:1d82] has quit [Client Quit]17:42
-!- hoijui [~hoijui@141.23.93.47] has quit [Ping timeout: 264 seconds]17:47
-!- hoijui [~hoijui@141.23.93.47] has joined #shogun17:52
-!- blackburn [~blackburn@46.20.65.51] has joined #shogun18:33
-!- hoijui [~hoijui@141.23.93.47] has quit [Ping timeout: 252 seconds]19:05
-!- hoijui [~hoijui@141.23.93.47] has joined #shogun19:13
-!- hoijui [~hoijui@141.23.93.47] has quit [Ping timeout: 260 seconds]19:19
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun19:33
@sonney2kblackburn, I am making massive progress on oligo19:47
blackburnsonney2k: nice19:47
@sonney2know I have a matrix with NaNs instead of 1's19:47
@sonney2k:D19:47
@sonney2k:``-(19:47
blackburnhaha19:47
blackburnNaNs for the win19:47
* sonney2k eats tons of chocolate for peace of mind19:48
@sonney2kblackburn, btw similar to SparseUtils we should probably have a LabelFactory class for the casting business19:49
@sonney2kcreate_from_*19:49
blackburnsonney2k: I was thinking about one thing19:50
blackburnsonney2k: why do we have labels and features (separately)?19:50
@sonney2kblackburn, because labels might be anything, real numbers, ints, strucutres19:50
@sonney2kand features might be strings, etc19:51
blackburnsonney2k: yes but what is the reason to keep them that separate?19:51
@sonney2kif they were same type no need then19:51
@sonney2kI see so some Label entry inside Features19:51
blackburnyes19:51
blackburnI do not suggest it yet but am just trying to understand if there is a special capability19:52
@sonney2kblackburn, there is one CON - for unsupervised methods or dim red etc you have no labels19:58
blackburnsonney2k: yes19:58
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has joined #shogun19:59
-!- heiko [~heiko@027fc803.bb.sky.com] has left #shogun []20:02
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has joined #shogun20:03
@sonney2kblackburn, haha it gets worse and wors20:10
@sonney2ke20:10
blackburnsonney2k: crashed your kernel?20:10
@sonney2kno but parameters are all crap now20:12
blackburnsonney2k: in linux mean of kernel :)20:12
@sonney2kblackburn, got it to work :D20:13
blackburncool20:13
@sonney2kthere is some oddity when you use it from python20:14
@sonney2kif you do20:14
@sonney2kkernel = OligoStringKernel(size_cache=10, k=3, width=60)20:14
@sonney2kit will accept this20:14
@sonney2kbut the parameters for k, width etc are crap20:14
blackburnsonney2k: wait so we do support keywords?20:14
@sonney2kno but this is what n4nd0 wrote in his test script20:15
@sonney2kand python didn't complain20:15
@sonney2kit just called the default constructor (taking no args)20:15
@sonney2kI have no idea why how this can happen20:15
blackburnsonney2k: I expect so20:15
@sonney2kwhy does it not give an error?20:15
blackburnsonney2k: empty list and kwargs**20:15
@sonney2kblackburn, so argument list is empty but keyword args exist?20:16
blackburnsonney2k: yes20:16
@sonney2kbut why does swig allow that?20:16
blackburnsounds legit to me20:16
blackburnsonney2k: I think kws are always allowed20:16
blackburnit is just a matter if you handle them or not20:17
@sonney2kI mean then it should set kws properly20:17
@sonney2kbut it doesn't20:17
@sonney2kanyway results look very sane to me now20:22
-!- FSCV [~FSCV@206.225.135.24] has joined #shogun20:23
-!- FSCV [~FSCV@206.225.135.24] has quit [Ping timeout: 248 seconds]20:49
@sonney2kblackburn, OK I've thoroughly checked local alignment, oligo and locality improved kernel20:53
@sonney2kthey are ok20:53
@sonney2kthe simple locality improved kernel is not20:54
blackburnsonney2k: I see20:54
@sonney2kgunnars code is pretty hard to digest20:55
@sonney2kproblem is that the kernel sometimes gives OK results20:55
@sonney2ksometimes not20:55
@sonney2kI don't understand why20:55
-!- FSCV [~FSCV@187.210.54.166] has joined #shogun20:56
@sonney2kand of course valgrind is happy too...20:58
@sonney2kblackburn, while we are at it - we should get rid of all the bool init() functions20:59
@sonney2kbut use void() instead20:59
blackburnsonney2k: why?20:59
@sonney2kthere are very few functions where we actually return anything21:00
@sonney2kwe usually SG_ERROR if something goes wrong21:00
@sonney2kso the code looks like21:00
@sonney2kSG_ERROR / ASSERT21:00
@sonney2kand then return true;21:00
@sonney2kso return values are stupid21:00
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]21:03
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun21:11
shogun-notifier-shogun: Soeren Sonnenburg :master * de9a300 / src/shogun/kernel/string/ (5 files): https://github.com/shogun-toolbox/shogun/commit/de9a3005ebd65c53c16a24f40e75a2662a46c45d21:11
shogun-notifier-shogun: String kernel fixes21:11
shogun-notifier-shogun:21:11
shogun-notifier-shogun: - add reasonable sqrtdiag normalizer to LI kernel21:11
shogun-notifier-shogun: - add convenience constructor to the oligo kernel21:11
shogun-notifier-shogun: - add slow compute function back to oligo kernel21:11
shogun-buildbot_build #851 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/851  blamelist: Soeren Sonnenburg <sonne@debian.org>21:16
@sonney2kblackburn, found it21:19
@sonney2kblackburn, the SLIK kernel is not thread safe21:19
@sonney2kso when we call get_kernel_matrix21:19
@sonney2kit is done w/ multiple threads21:19
@sonney2k-> kaboom21:19
blackburnsonney2k: hah21:19
@sonney2kblackburn, it would be actually but it contains the init code in the compute function again - wtf!?21:25
@sonney2kjust some left over code21:25
@sonney2kso it will be quicker once removed and work too..21:25
blackburnsonney2k: that's why we should call all the things as generic as possible21:26
blackburnI mean in kernel we should only define init21:26
@sonney2kblackburn, well it is some left over code obviously sb. optimized the code but forgot to remove it from comput()21:28
@sonney2kbtw oligokernel looks like it gives best results here21:30
@sonney2kblackburn, what do you mean w/ only define init?21:31
blackburnsonney2k: it should be called in CKernel21:31
shogun-notifier-shogun: Soeren Sonnenburg :master * cf38f2c / src/shogun/kernel/string/SimpleLocalityImprovedStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/cf38f2c9f2c6a2737b76e9fb95d0e7c2cdfaf37d21:32
shogun-notifier-shogun: Bugfix for SLIK kernel21:32
shogun-notifier-shogun:21:32
shogun-notifier-shogun: - remove left over init code in compute function of SLIK kernel21:32
shogun-notifier-shogun: - this init code caused an unstable pyramid array, making21:32
shogun-notifier-shogun: the kernel fail when #threads > 121:32
@sonney2kblackburn, in constructor you mean?21:32
blackburnsonney2k: yeah probably21:33
blackburnsonney2k: I know it can't be virtual but we should find a way to avoid such errors21:33
@sonney2kblackburn, that is not possible... we dont' know the kernels' features yet and we need to construct some array based on these features (to speed up compute)21:33
@sonney2kthis here was really just some stupid bug21:34
@sonney2kso let me regenerate data for these kernels to get tests to pass21:34
blackburnalrigh21:35
@sonney2kno idea how to do that though for static hmmm21:35
@sonney2khmmhh probably generate_testdata.py21:36
@sonney2kkaboomn21:37
@sonney2kthis old test generator is just so obsolete...21:39
@sonney2kblackburn, however it is our only chance for testing what we have in static interfaces :/21:40
blackburnyeah21:40
shogun-buildbot_build #852 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/852  blamelist: Soeren Sonnenburg <sonne@debian.org>21:41
shogun-notifier-shogun: Soeren Sonnenburg :master * 73f200d / tests/integration/generator/dataop.py: https://github.com/shogun-toolbox/shogun/commit/73f200df9b88813c5740feaae2e43c833180a4a221:49
shogun-notifier-shogun: fix data generator for kernel integration tests21:49
shogun-notifier-shogun: Soeren Sonnenburg :master * 06939d9 / data: https://github.com/shogun-toolbox/shogun/commit/06939d951d1aa1efc592034373dbb2493ceaf94721:49
shogun-notifier-shogun: suck in new integration test data21:49
@sonney2kblackburn, I looked at all the HMM related stuff21:49
@sonney2kresults are really crap21:49
@sonney2ksome inf's in there21:50
blackburnsonney2k: whoops21:50
@sonney2kso it is only HMM_1en10_60_4_1_0_6_3_False_CUBE21:51
@sonney2kfailing in tests now AFAIK21:51
shogun-buildbot_build #978 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/978  blamelist: Soeren Sonnenburg <sonne@debian.org>21:52
@sonney2kthat shouldn't be too hard to fix - I mean it worked at some point :D21:52
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has quit [Ping timeout: 245 seconds]22:03
shogun-buildbot_build #853 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/853  blamelist: Soeren Sonnenburg <sonne@debian.org>22:19
shogun-buildbot_build #979 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/979  blamelist: Soeren Sonnenburg <sonne@debian.org>22:22
shogun-notifier-shogun: Soeren Sonnenburg :master * 8ffd610 / data,tests/integration/generator/kernel.py: https://github.com/shogun-toolbox/shogun/commit/8ffd610d0fc72be7e1ce93db147dab98fe23407422:23
shogun-notifier-shogun: 1e-9 accuracy for LA kernel22:23
shogun-buildbot_build #854 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/854  blamelist: Soeren Sonnenburg <sonne@debian.org>22:25
shogun-buildbot_build #855 of deb2 - static_interfaces is complete: Failure [failed test python_static]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/855  blamelist: Soeren Sonnenburg <sonne@debian.org>22:31
shogun-buildbot_build #980 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/980  blamelist: Soeren Sonnenburg <sonne@debian.org>22:58
shogun-buildbot_build #981 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/981  blamelist: Soeren Sonnenburg <sonne@debian.org>23:00
@sonney2kblackburn, you don't happen to recall when the tests worked last time for HMM?23:20
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has quit [Ping timeout: 252 seconds]23:21
-!- heiko [~heiko@027fc803.bb.sky.com] has joined #shogun23:49
--- Log closed Wed Apr 03 00:00:18 2013