--- Log opened Sat Mar 30 00:00:18 2013 | ||
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 01:21 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 01:28 | |
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has quit [Ping timeout: 260 seconds] | 02:11 | |
shogun-buildbot | build #338 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/338 | 03:56 |
---|---|---|
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun | 07:58 | |
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds] | 08:36 | |
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has joined #shogun | 09:08 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 09:39 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:18 | |
n4nd0 | sonney2k: good morning, around? | 10:19 |
n4nd0 | blackburn: ah you are also here :) | 10:19 |
blackburn | n4nd0: yes | 10:19 |
n4nd0 | blackburn: I started taking a look yesterday to these tests that fail | 10:19 |
n4nd0 | I ma not really sure about a couple of things | 10:20 |
n4nd0 | blackburn: so the likeliest is that the errors are inside the data files I understand, the ones that are written in matlab | 10:20 |
blackburn | n4nd0: most of errors are due to wrong parameter setting or introduced bugs | 10:21 |
n4nd0 | blackburn: and to find out if the data used makes any sense at all | 10:22 |
n4nd0 | one can plot it I guess | 10:23 |
blackburn | n4nd0: I don't think the data is broken | 10:23 |
blackburn | it is here from the very beginning | 10:23 |
n4nd0 | mm I understand | 10:23 |
@sonney2k | n4nd0, blackburn good morning! | 10:26 |
@sonney2k | about the failing string kernel tests: | 10:27 |
@sonney2k | it would already help to see if the kernels do sth reasonable | 10:27 |
@sonney2k | this could be done by generating some simple sequences say all contain letter 'A' | 10:27 |
@sonney2k | and are of lenght say 50 | 10:27 |
n4nd0 | ok | 10:28 |
@sonney2k | consider you have 20 sequences like this | 10:28 |
@sonney2k | you could introduce in 10 of them a couple of 'C's in the middle (or somewhere) | 10:28 |
@sonney2k | and then just compute the kernel matrix | 10:28 |
@sonney2k | you should see a block structure | 10:28 |
@sonney2k | with largest values on the diagonal | 10:29 |
-!- lambday [0e8b614d@gateway/web/freenode/ip.14.139.97.77] has joined #shogun | 10:32 | |
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has quit [Quit: Leaving] | 10:33 | |
lambday | for the log-determinant of sparse matrix, shall I take eigen's sparse matrix as input? I thought of using SGSparseMatrix but it seems like an adjacency list and I don't know how to get the dimensions from it | 11:39 |
n4nd0 | sonney2k: http://pastebin.com/L1iaNB20 | 11:39 |
n4nd0 | that is the test you have suggested above | 11:40 |
n4nd0 | I see block structure but no large values on the diagonal | 11:40 |
n4nd0 | lambday: SGSparseMatrix::num_vector SGSparseMatrix::num_features maybe help you to get dimensions?? | 11:41 |
n4nd0 | num_vectors* | 11:41 |
lambday | n4nd0: ummm... I think it would be tough.. say, size of a sparse matrix can be 100x100 and say only 10 columns have non-zero entries, so SGSparseMatrix would have only 10 num_vector entries I am guessing.. | 11:44 |
n4nd0 | lambday: no, I don't think so | 11:45 |
n4nd0 | for me it makes sense that num_vectors and num_features would both be 100 | 11:45 |
n4nd0 | then there will be only stored 10 numbers | 11:45 |
n4nd0 | it is like if you do size of a sparse matrix in matlab | 11:46 |
n4nd0 | I mean the other elements exist, they are just zero so they are not explit. stored in memory | 11:47 |
n4nd0 | anyway, I am not testing right now what I am telling you - so don't trust me 100% :) | 11:47 |
lambday | n4nd0: I' | 11:48 |
lambday | I'm testing... will let you know :) | 11:48 |
@sonney2k | n4nd0, could you put the 'A' strings all as the first 10 and the ones with C to the end? | 12:02 |
@sonney2k | and then do | 12:02 |
@sonney2k | from pylab import * | 12:02 |
@sonney2k | pcolor(kmatrix) | 12:02 |
@sonney2k | colorbar() | 12:02 |
n4nd0 | sonney2k: ok, doing right now | 12:05 |
@sonney2k | n4nd0, I am changing the code to use the sqrtdiagnormalizer by default | 12:05 |
@sonney2k | I think that makes a lot of sense | 12:06 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 12:07 | |
shogun-notifier- | shogun: Soeren Sonnenburg :master * d07ddb4 / src/shogun/kernel/PolyKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/d07ddb485b8c4fbde7bef821791fbdb88c87a77a | 12:07 |
shogun-notifier- | shogun: move normalizer assignment to init() to avoid code duplication | 12:07 |
shogun-notifier- | shogun: Soeren Sonnenburg :master * f33e4f3 / src/shogun/kernel/string/LocalAlignmentStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/f33e4f35d5786227dc7493988f803db75208ecb0 | 12:07 |
shogun-notifier- | shogun: normalize LA kernel to have norm 1 on the diagonal | 12:07 |
-!- travis-ci [~travis-ci@ec2-174-129-144-237.compute-1.amazonaws.com] has joined #shogun | 12:11 | |
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/5918779 | 12:11 |
-!- travis-ci [~travis-ci@ec2-174-129-144-237.compute-1.amazonaws.com] has left #shogun [] | 12:11 | |
shogun-buildbot | build #847 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/847 blamelist: Soeren Sonnenburg <sonne@debian.org> | 12:12 |
n4nd0 | sonney2k: https://dl.dropbox.com/u/11020840/km_train.png | 12:13 |
shogun-buildbot | build #1030 of deb1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1030 blamelist: Soeren Sonnenburg <sonne@debian.org> | 12:14 |
n4nd0 | sonney2k: I have no real idea how it should work but the blue part looks sort of weird, maybe it should be more uniform like the red?? | 12:16 |
@sonney2k | n4nd0, look at https://dl.dropbox.com/u/31480471/website-cbase.png | 12:17 |
n4nd0 | sonney2k: oh looking cool :) | 12:18 |
n4nd0 | but I guess we have to decide where to put it | 12:18 |
n4nd0 | we won't have the logo twice | 12:18 |
@sonney2k | n4nd0, regarding the inhomogenity of the blue part | 12:18 |
@sonney2k | could you put the 'C' just at one fixed position in the sequence | 12:19 |
@sonney2k | then it should go away | 12:19 |
@sonney2k | (I hope...) | 12:19 |
n4nd0 | sonney2k: for all the sequences? | 12:19 |
@sonney2k | n4nd0, for the second half only | 12:19 |
n4nd0 | yes | 12:19 |
@sonney2k | the other are still just 'A's | 12:19 |
n4nd0 | I mean if for all the sequences with two 'C' | 12:19 |
n4nd0 | these two 'C' in the same position among all those sequences | 12:20 |
@sonney2k | errm? | 12:20 |
@sonney2k | yes | 12:20 |
@sonney2k | n4nd0, regarding the website | 12:21 |
@sonney2k | maybe we could put c-base in the middle | 12:21 |
@sonney2k | and gsoc announcement as the long thing? | 12:21 |
n4nd0 | sonney2k: yeah, it sounds good | 12:21 |
@sonney2k | only issue is we don't have much room for a workshop advertisement | 12:22 |
n4nd0 | sonney2k: why not to advert the workshop in the long thing? | 12:22 |
n4nd0 | on the rhs of the logo on top | 12:22 |
@sonney2k | n4nd0, or that yes | 12:23 |
n4nd0 | sonney2k: fixing the 'C' ends with inhomogenity, I see only three colors now | 12:23 |
@sonney2k | and then gsoc | 12:23 |
@sonney2k | n4nd0, this kernel is really weird | 12:25 |
@sonney2k | looks like it is not positive | 12:25 |
@sonney2k | but the result makes sense | 12:25 |
@sonney2k | I assume that the red block are just the 'A's | 12:25 |
n4nd0 | yes | 12:25 |
@sonney2k | and the blue block where the ones with A and C | 12:25 |
n4nd0 | yes | 12:26 |
@sonney2k | and since it computes all possible local alignments between to sequences | 12:26 |
n4nd0 | and now there is a new block with darker blue that corresponds to only 'C's | 12:26 |
@sonney2k | we did have a quite inhomogenious blue block | 12:26 |
@sonney2k | the problem really is that 'AAA's align very well | 12:26 |
@sonney2k | so they will have the highest score | 12:27 |
n4nd0 | sonney2k: how do see that it doesn't look positive? (I guess positive semi-definite) | 12:27 |
@sonney2k | and there are not that many alignments of sequences that have a C | 12:27 |
@sonney2k | n4nd0, it's diagonal should have the largest values | 12:28 |
@sonney2k | in any case IMHO sanity check passed | 12:29 |
@sonney2k | just do the same with the other 2 string kernels | 12:29 |
n4nd0 | sonney2k: mmm OligoKernel gives me a completely uniform matrix | 12:33 |
n4nd0 | all values equal to 1 | 12:33 |
@sonney2k | investigating | 12:34 |
@sonney2k | n4nd0, do the other in the meantime | 12:34 |
n4nd0 | size_cache=10, k=3, width=1.2 | 12:34 |
n4nd0 | PolyKernel is the other one I guess | 12:34 |
n4nd0 | ah no | 12:34 |
n4nd0 | SimpleLocalityImproved | 12:34 |
@sonney2k | try a much larger width | 12:35 |
@sonney2k | say sth like length of the strings | 12:35 |
n4nd0 | still the same with width 50 | 12:36 |
@sonney2k | that's weird | 12:38 |
shogun-buildbot | build #974 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/974 blamelist: Soeren Sonnenburg <sonne@debian.org> | 12:41 |
n4nd0 | sonney2k: and the last one looks pretty weird too | 12:41 |
n4nd0 | maybe I set nonsense parameters: length=5, inner_degree=5, outer_degree=1 | 12:42 |
shogun-notifier- | shogun: Soeren Sonnenburg :master * dc0b603 / src/shogun/kernel/string/LocalAlignmentStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/dc0b60389353019bbb2454a4b287346dbf22b83f | 12:43 |
shogun-notifier- | shogun: add missing include to fix compile error | 12:43 |
@sonney2k | n4nd0, image? | 12:43 |
@sonney2k | could you try inner_degree=1 | 12:44 |
n4nd0 | https://dl.dropbox.com/u/11020840/2013-03-30-124548_1600x900_scrot.png | 12:44 |
n4nd0 | that is with inner_degree=5, let me change | 12:44 |
@sonney2k | this certainly needs some normalization too | 12:46 |
n4nd0 | https://dl.dropbox.com/u/11020840/2013-03-30-124805_1600x900_scrot.png | 12:47 |
n4nd0 | with inner degree 1, ^ | 12:47 |
shogun-buildbot | build #1031 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1031 | 12:49 |
@sonney2k | n4nd0, strange | 12:49 |
-!- travis-ci [~travis-ci@ec2-107-21-142-105.compute-1.amazonaws.com] has joined #shogun | 12:50 | |
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/5919085 | 12:50 |
-!- travis-ci [~travis-ci@ec2-107-21-142-105.compute-1.amazonaws.com] has left #shogun [] | 12:50 | |
@sonney2k | n4nd0, please email me the code I will have to investigate | 12:51 |
n4nd0 | sonney2k: ok | 12:51 |
n4nd0 | sent | 12:54 |
shogun-buildbot | build #848 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/848 blamelist: Soeren Sonnenburg <sonne@debian.org> | 12:59 |
shogun-buildbot | build #975 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/975 blamelist: Soeren Sonnenburg <sonne@debian.org> | 13:18 |
n4nd0 | ok, I've got a plane to take :) see you on Tuesday again | 13:23 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 13:23 | |
wiking | sonney2k: can u plz change fbsd bot to clang | 13:48 |
wiking | it should be fine now... | 13:48 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 15:43 | |
-!- lambday [0e8b614d@gateway/web/freenode/ip.14.139.97.77] has quit [] | 16:39 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 16:42 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 19:41 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 19:42 | |
@sonney2k | shogun-buildbot, force build "bsd1 - libshogun" | 19:42 |
shogun-buildbot | build forced [ETA 5m38s] | 19:42 |
shogun-buildbot | I'll give a shout when the build finishes | 19:42 |
shogun-buildbot | Hey! build bsd1 - libshogun #843 is complete: Success [build successful] | 19:45 |
shogun-buildbot | Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/843 | 19:45 |
@sonney2k | wiking ^ all good | 20:05 |
blackburn | sonney2k: did you notice anything wrong with these kernels? | 20:09 |
@sonney2k | blackburn, LocalAlignment kernel looked ok | 20:10 |
@sonney2k | I still have to check the others | 20:10 |
blackburn | so what to do with it then? | 20:10 |
@sonney2k | I've changed the default kernel normalizer to something I consider more reasonable - just generate new data for the la-kernel and good | 20:11 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 20:17 | |
shogun-notifier- | shogun: Soeren Sonnenburg :master * 640ab7f / src/shogun/kernel/string/SimpleLocalityImprovedStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/640ab7f4a1bf93d0c073e662ae59babac77d5c2d | 20:17 |
shogun-notifier- | shogun: normalize diagonal to one for LIK by default | 20:17 |
-!- travis-ci [~travis-ci@ec2-54-234-19-227.compute-1.amazonaws.com] has joined #shogun | 20:20 | |
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/5925728 | 20:20 |
-!- travis-ci [~travis-ci@ec2-54-234-19-227.compute-1.amazonaws.com] has left #shogun [] | 20:20 | |
shogun-buildbot | build #1032 of deb1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1032 blamelist: Soeren Sonnenburg <sonne@debian.org> | 20:28 |
shogun-notifier- | shogun: Soeren Sonnenburg :master * 39c115e / src/shogun/kernel/string/SimpleLocalityImprovedStringKernel.cpp: https://github.com/shogun-toolbox/shogun/commit/39c115eaf37ce691968d8e4a6808ef15d30acfdd | 20:28 |
shogun-notifier- | shogun: include normalizer after kernel header file to fix compile error | 20:28 |
-!- travis-ci [~travis-ci@ec2-23-20-101-65.compute-1.amazonaws.com] has joined #shogun | 20:35 | |
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/5925926 | 20:35 |
-!- travis-ci [~travis-ci@ec2-23-20-101-65.compute-1.amazonaws.com] has left #shogun [] | 20:35 | |
shogun-buildbot | build #1033 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1033 | 20:40 |
@sonney2k | blackburn, I've checked the LIK kernel | 20:43 |
@sonney2k | it also works as I expect it to work | 20:43 |
blackburn | sonney2k: quite strange it fails.. | 20:43 |
@sonney2k | it is just that parameter choice is crucial | 20:43 |
shogun-buildbot | build #849 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/849 blamelist: Soeren Sonnenburg <sonne@debian.org> | 20:43 |
@sonney2k | blackburn, Oligokernel is weird. I get only 1's in the km | 20:53 |
@sonney2k | blackburn, do you know how to see the git log when a file was moved? | 20:53 |
blackburn | sonney2k: no not really | 20:53 |
@sonney2k | blackburn, git log --follow shogun/kernel/string/OligoStringKernel.cpp | 20:58 |
blackburn | well that's just history? | 20:59 |
@sonney2k | blackburn, exactly what I want | 20:59 |
blackburn | dirk messages in the bottom | 21:00 |
blackburn | sonney2k: did you fire him? | 21:00 |
blackburn | :D | 21:00 |
@sonney2k | blackburn, no university kicked him out because he failed to enroll in time | 21:01 |
blackburn | oh he writes poetry in commits | 21:01 |
@sonney2k | but we helped him to get his state back (our secretary went with him to the right people and all solved) | 21:02 |
@sonney2k | blackburn, I think the bug is in oligokernel for 5 years or so | 21:03 |
blackburn | ha which means | 21:03 |
blackburn | nobody is using it | 21:03 |
@sonney2k | yes implementation in shogun is too slow anyway and the WD kernels are just much much more fast | 21:03 |
@sonney2k | in 2009 I 'slightly sped up the oligokernel' haha | 21:04 |
@sonney2k | maybe it is broken since then | 21:04 |
shogun-buildbot | build #976 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/976 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:19 |
blackburn | argh another day of dental thing is setting me to sleep | 21:20 |
blackburn | sonney2k: I am now spending each saturday at dentist :D | 21:21 |
blackburn | sonney2k: I'll take a look at sgd again tomorrow | 21:22 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 23:28 | |
--- Log closed Sun Mar 31 00:00:18 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!