Open in new window / Try shogun cloud
--- Log opened Sun Dec 04 00:00:59 2011
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]01:17
-!- blackburn [~blackburn@] has joined #shogun10:23
--- Log closed Sun Dec 04 14:46:57 2011
--- Log opened Sun Dec 04 14:47:06 2011
-!- shogun-toolbox [] has joined #shogun14:47
-!- ServerMode/#shogun [+ns] by adams.freenode.net14:47
-!- Irssi: #shogun: Total of 1 nicks [1 ops, 0 halfops, 0 voices, 0 normal]14:47
-!- Irssi: Join to #shogun was synced in 6 secs14:47
-!- ServerMode/#shogun [-o shogun-toolbox] by adams.freenode.net14:52
! *** Notice -- TS for #shogun changed from 1323006426 to 130097232114:52
-!- ServerMode/#shogun [+ct-s] by adams.freenode.net14:52
-!- blackburn [~blackburn@] has joined #shogun14:52
-!- changed the topic of #shogun to: Shogun Machine Learning Toolbox | We participate in GSoC with 5 projects | Channel logs
-!- 15SAAI18M [] has joined #shogun14:52
-!- blackburn [~blackburn@] has quit [Remote host closed the connection]14:54
-!- shogun-buildbot [] has joined #shogun14:54
-!- sonney2k [] has joined #shogun14:54
-!- ServerMode/#shogun [+o sonney2k] by adams.freenode.net14:54
-!- blackburn [~blackburn@] has joined #shogun15:02
blackburnfucking shit18:02
blackburngod damned country18:02
blackburnsonney2k: strange you came on this video heh20:30
blackburnsonney2k: such an untypical process for russia, usually it comes backward20:31
blackburnsonney2k: more typical
@sonney2kI has it *lol*21:04
@sonney2kblackburn, did you make any progress on the tests?21:04
blackburnsonney2k: what kind of?21:10
shogun-buildbotbuild #76 of nightly_all is complete: Success [build successful]  Build details are at
blackburnsonney2k: we still have regression in HMM21:11
-!- in3xes [~in3xes@] has joined #shogun21:12
-!- in3xes [~in3xes@] has quit [Ping timeout: 252 seconds]21:35
blackburnsonney2k: which SVM test failures are ok?21:38
blackburnsubgradient, ocas, sgd are failing now21:38
@sonney2kblackburn, only HMM or more?21:58
blackburnsonney2k: nothing extra21:59
@sonney2kocas, sgd should not fail21:59
blackburnsonney2k: they are21:59
@sonney2kdid these fail in 1.0.0?22:00
@sonney2kwait these did fail in 1.0?22:01
blackburnsonney2k: I hate checking it but I will22:01
15SAAI18Mshogun: Sergey Lisitsyn master * rd320074 / (4 files): Added embed kernel for the kernel LLE -
blackburnthese elections results is fucking disappointing again22:04
blackburnI'm just like 'I haz it' russian22:04
shogun-buildbotbuild #361 of octave_static is complete: Failure [failed test_1]  Build details are at  blamelist: blackburn91@gmail.com22:09
shogun-buildbotbuild #360 of r_static is complete: Success [build successful]  Build details are at
blackburnsonney2k: yes it was failing in 1.0.022:20
blackburnsonney2k: while HMM was not22:21
blackburnkm_test: 1.648674e-02, km_train: 1.430996e-02 <--- accuracy: 1.000000e-0922:23
blackburnkm_test: 1.691602e-02, km_train: 6.647181e-04 <--- accuracy: 1.000000e-0922:23
blackburnkm_test: 1.489201e-01, km_train: 1.502606e-01 <--- accuracy: 1.000000e-0622:23
blackburnkm_test: 1.725301e-01, km_train: 1.051410e-01 <--- accuracy: 1.000000e-0622:23
blackburnstill here too22:23
blackburnkm_test: 4.217000e+00, km_train: 4.204000e+00 <--- accuracy: 0.000000e+0022:23
@sonney2kblackburn, you meant this
blackburnsonney2k: that's fucking unbelievable piece of shit23:23
blackburnin moscow exit-polls showed that this fucking party has ~30%23:23
blackburnbut results are >45%23:24
blackburntoo much people left russia already, I'm the one who will do the same in the next few years23:24
blackburnsonney2k: you said everybody cheats, but in some regions there is 99% rating of government party23:25
@sonney2kback to getting 1.1 ready - so we need to fix these 3-4 regressions and then we are OK right?23:26
blackburnsonney2k: HMM only23:27
@sonney2keven better - so what was the result of your git bisect?23:27
blackburnsonney2k: I failed to detect it23:28
blackburnand didn't managed to do this thing again23:28
@sonney2kblackburn, 99% in a election? not even nazis managed to get that...23:28
blackburnsonney2k: chechnya did23:28
blackburnsonney2k: statistician hell here: with 10% counted 'united russia' result was 45%23:29
blackburnnow with 50% counted it is >50%23:30
blackburnmassively cheating, really unbelievable23:30
blackburnactually they get 20-25%, not more23:31
@sonney2kblackburn, wait you said that HMM was working in shogun 1.0.0?23:33
@sonney2kbut isn't now?23:33
blackburnsonney2k: yes23:34
@sonney2kI am looking at the changelog23:34
@sonney2kand it is just this23:34
@sonney2kso I would suspect it either works in 1.0 and 1.1 or hell freezes ;-)23:35
blackburnsonney2k: I tested already three times or more23:37
@sonney2kor - we did regenerate tests?23:37
blackburnwe didn't23:37
@sonney2kbut how is this possible??23:37
blackburnsomething dependent changed23:37
@sonney2kbut that could only be some stringfeature related stuff23:38
blackburnsonney2k: can you test it anyhow?23:38
blackburnmay be old projects or so23:39
blackburnwhere you know what result should look like23:39
@sonney2knot now - I have to sleep now. I will try tomorrow...23:39
blackburnsonney2k: ok, thanks23:40
blackburnI'll try to fix customkernel later23:40
blackburnsee you23:40
--- Log closed Mon Dec 05 00:00:59 2011