--- Log opened Tue Aug 16 00:00:38 2011 | ||
--- Day changed Tue Aug 16 2011 | ||
@sonney2k | it is too bad that I cannot really schedule any meeting because my schedule is totally unclear/open | 00:00 |
---|---|---|
@sonney2k | *sighs* | 00:00 |
@sonney2k | anyway, we are good | 00:00 |
@sonney2k | I hope heiko manages to fix the migration stuff and then we can rock the test suite :) | 00:01 |
blackburn | killed >50 warnings | 00:01 |
blackburn | will commit in a min | 00:01 |
@sonney2k | from then on the buildbot will ensure that things never diverge again :) | 00:02 |
@sonney2k | (famous last words) | 00:02 |
blackburn | why famous? | 00:02 |
@sonney2k | blackburn, don't you know the famous last words? | 00:03 |
@sonney2k | nothing can go wrong | 00:03 |
@sonney2k | i have everthing under control | 00:03 |
blackburn | I don't really understand what is you are talking about :D | 00:03 |
@sonney2k | doesn't matter :) | 00:03 |
@sonney2k | I was joking | 00:04 |
blackburn | okay.. | 00:04 |
@sonney2k | as always | 00:04 |
CIA-87 | shogun: Sergey Lisitsyn master * r1e71db1 / (7 files in 3 dirs): More warning hunt - http://bit.ly/qUecom | 00:04 |
blackburn | sonney2k: what a soul-stirring mail from you ;) | 00:08 |
blackburn | arhh can't wait till it compiles | 00:10 |
blackburn | power off now, sonney2k see you | 00:10 |
-!- blackburn [~blackburn@188.122.238.50] has quit [Quit: Leaving.] | 00:12 | |
serialhex | i *just* got home... and blackburn *just6* got off :( | 00:15 |
--- Log closed Tue Aug 16 02:53:22 2011 | ||
--- Log opened Tue Aug 16 02:53:28 2011 | ||
-!- shogun-irclog [~shogun@7nn.de] has joined #shogun | 02:53 | |
-!- Irssi: #shogun: Total of 7 nicks [1 ops, 0 halfops, 0 voices, 6 normal] | 02:53 | |
-!- Irssi: Join to #shogun was synced in 5 secs | 02:53 | |
-!- f-x [~user@117.192.199.77] has quit [Ping timeout: 260 seconds] | 04:25 | |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 05:29 | |
-!- in3xes [~in3xes@27.107.147.125] has quit [Ping timeout: 250 seconds] | 07:16 | |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 07:28 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Read error: Connection reset by peer] | 07:28 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 07:28 | |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 08:05 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 258 seconds] | 08:08 | |
-!- in3xes_ is now known as in3xes | 08:18 | |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 08:54 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 08:57 | |
-!- in3xes_ [~in3xes@180.149.49.227] has quit [Remote host closed the connection] | 09:11 | |
-!- vojta [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 09:25 | |
-!- vojta is now known as VojtechFranc | 09:25 | |
CIA-87 | shogun: Shashwat Lal Das master * r6b4e7a2 / (2 files in 2 dirs): Added VW example and fixed a write-mode argument in IOBuffer. - http://bit.ly/qxmX9T | 09:28 |
CIA-87 | shogun: Soeren Sonnenburg master * r41e3393 / (2 files in 2 dirs): | 09:28 |
CIA-87 | shogun: Merge pull request #298 from frx/streaming_vw | 09:28 |
CIA-87 | shogun: Added VW example and fixed a write-mode argument in IOBuffer. - http://bit.ly/ohH3Rf | 09:28 |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 10:03 | |
-!- in3xes [~in3xes@27.107.147.125] has quit [Ping timeout: 252 seconds] | 10:29 | |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 10:30 | |
-!- in3xes [~in3xes@27.107.147.125] has quit [Ping timeout: 276 seconds] | 10:42 | |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 10:42 | |
-!- bb__ [53f4b424@gateway/web/freenode/ip.83.244.180.36] has joined #shogun | 10:46 | |
bb__ | hi ? | 10:46 |
@sonney2k | bb__, hi | 10:47 |
@sonney2k | alright | 10:47 |
@sonney2k | bb__, there seems to be sth wrong with your matlab detection | 10:47 |
bb__ | apparently so. i'm using a recent matlab install and configure doesn't seem to pick it up | 10:48 |
@sonney2k | bb__, so lets run a few commands and check | 10:48 |
-!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 10:48 | |
@sonney2k | what does | 10:48 |
@sonney2k | uname -s | 10:48 |
@sonney2k | say? | 10:48 |
bb__ | Linux | 10:48 |
@sonney2k | so far so good | 10:49 |
@sonney2k | matlab -nojvm -nodisplay -r 'version;quit' >/dev/null 2>&1 && echo yes | 10:49 |
@sonney2k | bb__, and? | 10:50 |
bb__ | it says 'yes' | 10:50 |
@sonney2k | great | 10:50 |
@sonney2k | now | 10:50 |
@sonney2k | echo "unix('env');quit" | matlab -nojvm -nodisplay | grep '^MATLAB=' | cut -c 8- | 10:50 |
bb__ | it doesn't produce anything (sorry for delay - i have to type into another screen) | 10:52 |
@sonney2k | ok that is the reason | 10:52 |
@sonney2k | then lets see | 10:52 |
@sonney2k | echo "unix('env');quit" | matlab -nojvm -nodisplay | grep MATLAB | 10:52 |
bb__ | I can't paste back the output , but there's no "MATLAB" env variable set. Closest thing is BASEMATLABPATH | 10:53 |
@sonney2k | bb__, why can't you past? | 10:54 |
@sonney2k | paste | 10:54 |
bb__ | I'm using a machine that's off the network. | 10:55 |
@sonney2k | bb__, when you do unix('env') does the path where matlab is installed somehow appear in any of the variables? | 10:55 |
@sonney2k | bb__, problem really is to detect the path where matlab is installed | 10:56 |
bb__ | well, matlab is on my path. is that what you mean ? | 10:56 |
@sonney2k | bb__, it doesn't help that it is in your $PATH - we need a way to figure out where it is installed | 10:57 |
@sonney2k | like /opt/software/matlab2011a/ whatever | 10:57 |
bb__ | Well, i could set an environement variable to where its installed. But i don't see its install location in the output of the matlab -nojvm ... command | 10:58 |
@sonney2k | bb__, so did unix('env') in matlab show you anything that contains that? | 10:58 |
bb__ | No . | 10:59 |
@sonney2k | bb__, and if you do which matlab from cmdline ? | 11:00 |
bb__ | sure, it finds it because its on my path. but apart from the path variable, i don't see it in the environment | 11:00 |
@sonney2k | bb__, or even better from matlab type matlabroot | 11:00 |
bb__ | that finds it. | 11:00 |
@sonney2k | bb__, alright then I will use that | 11:02 |
bb__ | how so ? will you have to modify the configure script ? | 11:04 |
@sonney2k | bb__, I will do a patch soon - stay tuned. got to go now but will submit a patch to git in the next few hrs. | 11:04 |
@sonney2k | bb__, cu | 11:04 |
bb__ | thanks !! | 11:04 |
-!- bb__ [53f4b424@gateway/web/freenode/ip.83.244.180.36] has quit [Quit: Page closed] | 11:05 | |
-!- in3xes [~in3xes@27.107.147.125] has quit [Ping timeout: 240 seconds] | 11:13 | |
CIA-87 | shogun: Soeren Sonnenburg master * r3f2cb7e / src/configure : | 11:15 |
CIA-87 | shogun: use matlabroot to determine matlab root path | 11:15 |
CIA-87 | shogun: hopefully fixes matlab 2011a compile issues - http://bit.ly/orUFXd | 11:15 |
-!- blackburn [~blackburn@188.122.238.50] has joined #shogun | 11:22 | |
-!- heiko [~heiko@134.91.52.150] has joined #shogun | 12:01 | |
-!- heiko [~heiko@134.91.52.150] has quit [Client Quit] | 12:03 | |
-!- heiko [~heiko@134.91.52.150] has joined #shogun | 12:03 | |
CIA-87 | shogun: Sergey Lisitsyn master * redc7424 / (14 files in 6 dirs): Doc safari - http://bit.ly/p1iy4i | 12:12 |
CIA-87 | shogun: Sergey Lisitsyn master * r703659c / (10 files in 4 dirs): More warning hunt - http://bit.ly/pXopcQ | 12:36 |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 12:49 | |
-!- in3xes [~in3xes@27.107.147.125] has quit [Ping timeout: 258 seconds] | 14:05 | |
@sonney2k | heiko, did you make any progress on the bug? | 14:12 |
heiko | hey sonney2k | 14:12 |
@sonney2k | heiko, we are trying to get the examples to run again (w/o crashes...) | 14:12 |
heiko | I did not really work on it yet because i am getting nuts with the parameter migration stuff | 14:12 |
@sonney2k | such that we can have them in the build tests | 14:12 |
heiko | should i do the bug first? | 14:12 |
@sonney2k | yes please | 14:12 |
heiko | ok then | 14:12 |
@sonney2k | bugs are always more important | 14:12 |
heiko | ok i will approach it then | 14:13 |
blackburn | sonney2k: 571 | 14:13 |
@sonney2k | blackburn, hey hey | 14:13 |
@sonney2k | great stuff | 14:13 |
@sonney2k | like down to 50% :) | 14:13 |
blackburn | I'd like to get it to zero this week | 14:13 |
@sonney2k | blackburn, then we can start the test suite | 14:13 |
blackburn | yeah | 14:13 |
@sonney2k | and then we will see more :) | 14:13 |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 14:27 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 250 seconds] | 14:32 | |
-!- petiera [~ohyvarin@kosh.org.aalto.fi] has left #shogun [] | 14:48 | |
blackburn | should be <500 now | 14:50 |
CIA-87 | shogun: Sergey Lisitsyn master * r747d65d / (17 files in 8 dirs): More warning hunt - http://bit.ly/mQkiLW | 14:51 |
@sonney2k | heiko, any idea about the bug? | 14:57 |
heiko | sonney2k, just wrote a test which shows normal behaviour | 14:57 |
heiko | trying to reproduce | 14:58 |
heiko | but still extending the test | 14:58 |
@sonney2k | heiko, you can reproduce it by running any of the *serial* examples in undocumented/python_modular | 14:58 |
heiko | oh | 14:58 |
heiko | sonney2k, oh, I thought you were talking about the cross.-validation bug | 14:59 |
heiko | that was reported to the tracker | 14:59 |
heiko | sorry, misunderstood you | 14:59 |
heiko | Well i will check the other bug afterwards, the x-val-splitting bug should be done soon | 15:00 |
-!- blackburn [~blackburn@188.122.238.50] has quit [Ping timeout: 250 seconds] | 15:01 | |
-!- blackburn [~blackburn@188.122.238.50] has joined #shogun | 15:30 | |
blackburn | sonney2k: 464! | 15:30 |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 15:33 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 258 seconds] | 15:42 | |
-!- in3xes [~in3xes@27.107.147.125] has joined #shogun | 15:43 | |
-!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 16:02 | |
@sonney2k | heiko, the serialization bug is the critical one - as this is currently what fails in the test suite | 16:15 |
-!- blackburn [~blackburn@188.122.238.50] has quit [Read error: Operation timed out] | 16:15 | |
@sonney2k | heiko, it is a blocker - before that is one is not fixed we cannot enable make check-examples in the tests | 16:16 |
@sonney2k | err buildbot | 16:16 |
heiko | sonney2k, ok, i will start very soon, the x-val-splitting thing is nearly done. will send a patch in the next minutes | 16:17 |
@sonney2k | ok | 16:19 |
heiko | sonney2k, have you seen the xval bugreport? | 16:38 |
@sonney2k | heiko, yes but didn't pay attention to it | 16:38 |
heiko | I think there is no error, I added some warnings to prevent false usage | 16:39 |
@sonney2k | heiko, you can talk to the submitter of the bug | 16:39 |
@sonney2k | or leave a comment etc | 16:39 |
heiko | (if there are not enough elements of each class to put one in every subset, its clear that there may be a problem | 16:39 |
heiko | how to login into the tracker? | 16:39 |
@sonney2k | heiko, http://shogun-toolbox.org/trac ? | 16:40 |
heiko | i need login | 16:43 |
@sonney2k | heiko, you got mail | 16:43 |
@sonney2k | heiko, did it work? | 16:47 |
heiko | testing | 16:50 |
heiko | did only get a passwort, no login | 16:50 |
@sonney2k | heiko, guess 3 times what your login is :D | 16:51 |
heiko | hehe, welllll | 16:51 |
heiko | working :) | 16:57 |
heiko | sonney2k, logged in but still cannot see email adress or comment | 17:02 |
@sonney2k | heiko, now? | 17:08 |
heiko | yes, now its working | 17:09 |
@sonney2k | ok | 17:11 |
heiko | sonney2k, here comes the pull request | 17:17 |
CIA-87 | shogun: Soeren Sonnenburg master * r14d5e34 / (9 files in 4 dirs): | 17:19 |
CIA-87 | shogun: Merge pull request #299 from karlnapf/master | 17:19 |
CIA-87 | shogun: added example/test for stratified cross-validation splitting (+6 more commits...) - http://bit.ly/pyKAH5 | 17:19 |
@sonney2k | heiko, ok now please the 'real' bug and I am happy | 17:19 |
heiko | sonney2k, yes | 17:19 |
-!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 17:20 | |
shogun-buildbot | build #44 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/libshogun/builds/44 | 17:24 |
shogun-buildbot | build #40 of cmdline_static is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org:5347/builders/cmdline_static/builds/40 blamelist: heiko.strathmann@gmail.com | 17:28 |
@sonney2k | heiko, http://www.shogun-toolbox.org:5347/builders/cmdline_static/builds/40/steps/compile/logs/stdio ? | 17:28 |
heiko | yes, just saw it | 17:29 |
heiko | fixing .... | 17:29 |
@sonney2k | the buildbot was really worth the work :D | 17:31 |
heiko | true :) | 17:31 |
heiko | mmh | 17:31 |
heiko | i do not understand that error | 17:31 |
heiko | and here libshogun compiles | 17:32 |
heiko | ah ok | 17:32 |
heiko | cmdline static | 17:32 |
heiko | sonney2k, that error message is strange | 17:32 |
heiko | sonney2k, the matrix serial example works | 17:37 |
heiko | the complex example fails due to an illegal import | 17:37 |
heiko | and the svm serial example segfaults | 17:38 |
@sonney2k | heiko, I sent to the valgrind output of that crash | 17:38 |
heiko | of which example was it? | 17:38 |
@sonney2k | these examples used to work | 17:39 |
@sonney2k | svmlight | 17:39 |
heiko | ok | 17:39 |
@sonney2k | dammed crash of build slave | 17:39 |
@sonney2k | running a memtest now | 17:43 |
@sonney2k | heiko, file serialization_svmlight_modular.py - seems to be some double free() | 17:44 |
@sonney2k | no idea why though | 17:44 |
heiko | sonney2k, I removed the stuff i added to SGObject and the example still crashes | 17:44 |
heiko | mmmhm | 17:44 |
@sonney2k | ehh I know that it worked reliably - if everything else fails we need to git bisect | 17:45 |
@sonney2k | heiko, binary search in git revisions for the 'bad' commit | 17:46 |
-!- in3xes [~in3xes@27.107.147.125] has quit [Quit: Leaving] | 18:00 | |
heiko | sonney2k, i have got to go now, sorry, will do this tommorrow at first place | 18:05 |
heiko | checked my commits, think this was caused some time ago, not by recent work | 18:06 |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 18:09 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Remote host closed the connection] | 18:09 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 18:10 | |
-!- heiko [~heiko@134.91.52.150] has quit [Ping timeout: 258 seconds] | 18:11 | |
-!- blackburn [~blackburn@188.122.238.50] has joined #shogun | 19:48 | |
@sonney2k | blackburn, shogun-toolbox.org is now twice as powerful :) | 20:00 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 258 seconds] | 20:01 | |
-!- f-x [~user@117.192.215.151] has joined #shogun | 20:23 | |
-!- f-x [~user@117.192.215.151] has left #shogun [] | 20:24 | |
-!- f-x [~user@117.192.215.151] has joined #shogun | 20:24 | |
blackburn | sonney2k: what happened? | 20:28 |
@sonney2k | blackburn, I paid a few euros to upgrade it | 20:28 |
blackburn | nice | 20:28 |
@sonney2k | blackburn, 3G memory now instead of 768M ... however we still have load 5 on that server | 20:30 |
@sonney2k | lots of connections to apache it seems | 20:30 |
@sonney2k | and git | 20:30 |
blackburn | I hope you will receive google money and able to pay greater server | 20:31 |
blackburn | btw we can host some site on github | 20:32 |
@sonney2k | blackburn, it certainly is big enough now for just hosting websites/git/trac/buildbot | 20:32 |
@sonney2k | blackburn, <1G of memory in use currently | 20:33 |
blackburn | okay | 20:33 |
@sonney2k | (under high load) | 20:33 |
@sonney2k | I think money would be better invested in some buildbot | 20:34 |
@sonney2k | err build-slave | 20:34 |
@sonney2k | e.g. an osx machine running linux/cygwin in virtual machines too | 20:34 |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 20:56 | |
blackburn | sonney2k: got a plan for next ~3 months | 21:00 |
@sonney2k | world domination? | 21:00 |
blackburn | not yet | 21:00 |
@sonney2k | yeah true - that takes 6 months | 21:00 |
blackburn | spectral clustering, t-sne, sammon mapping, and some more | 21:00 |
@sonney2k | t-sne ? | 21:00 |
@sonney2k | sammon mapping? | 21:01 |
blackburn | yeah dim reduction techniques | 21:01 |
blackburn | more applicable to real data than isomap/lle stuff | 21:01 |
* sonney2k googles | 21:01 | |
blackburn | ah yes, LLTSA too | 21:01 |
blackburn | we have LTSA but linear LTSA - LLTSA should be better for real data | 21:02 |
@sonney2k | I guess I will be stuck in admin stuff / fixing structural errors all my life long | 21:02 |
blackburn | it is quite frustrating but all of methods are fit for toy data rather than real | 21:03 |
blackburn | of developed already | 21:03 |
* sonney2k runs git bisect to detect the commit that killed serialization | 21:03 | |
blackburn | why do you think so? | 21:03 |
@sonney2k | blackburn, welcome to the wonderful world of ML research | 21:04 |
blackburn | structural stuff should be fixed soon I hope | 21:04 |
@sonney2k | blackburn, I currently see so many things that I would love to polish | 21:04 |
@sonney2k | blackburn, and even just parallelizing code is what I consider fixing bugs | 21:04 |
@sonney2k | but you never know :D | 21:04 |
* sonney2k the serialization stuff broke between july 26't and aug 16 | 21:05 | |
blackburn | well we should fix as much as possible | 21:05 |
blackburn | ohoh | 21:05 |
blackburn | nice | 21:05 |
blackburn | very important info :D | 21:05 |
shogun-buildbot | build #4 of nightly_none is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/nightly_none/builds/4 | 21:05 |
blackburn | nightly built?? | 21:05 |
@sonney2k | blackburn, git bisect just rocks | 21:06 |
blackburn | never used it | 21:06 |
@sonney2k | blackburn, nightly_all failed again but nightly_none works now :) | 21:06 |
@sonney2k | blackburn, it does binary search in commits :) | 21:07 |
blackburn | where are code warnings | 21:07 |
blackburn | yes I know what it does | 21:07 |
blackburn | but never used | 21:07 |
blackburn | there are* | 21:07 |
@sonney2k | I know you never make any mistakes :) | 21:07 |
blackburn | http://www.shogun-toolbox.org:5347/builders/nightly_none/builds/4/steps/compile/logs/warnings%20%2865%29 | 21:07 |
blackburn | I do a lot | 21:07 |
blackburn | hmm | 21:07 |
blackburn | I should fix it | 21:08 |
@sonney2k | blackburn, ignore these for now | 21:08 |
blackburn | why? | 21:08 |
@sonney2k | it is better to work on the real thing | 21:08 |
@sonney2k | this is with all options disabled | 21:08 |
blackburn | ah forgot | 21:08 |
blackburn | I want to make some thing to compute intrinsic dimension of data | 21:08 |
@sonney2k | blackburn, regarding your plan - please fire up trac and create a new milestone for shogun 1.2 and add your algo's as bugs in there | 21:09 |
blackburn | why 1.2? | 21:09 |
@sonney2k | blackburn, 1.1 ? | 21:10 |
blackburn | when do you plan to release 1.1? | 21:10 |
@sonney2k | true - next release should be around x-mas | 21:10 |
@sonney2k | though we might have to have bug fix releases inbetween | 21:10 |
blackburn | yeah | 21:11 |
blackburn | sonney2k: how soon will it be in debian? | 21:11 |
@sonney2k | blackburn, as soon as I manage to package it | 21:11 |
blackburn | unstable, right? | 21:11 |
@sonney2k | not so easy because we made so many structural changes | 21:11 |
@sonney2k | y | 21:11 |
@sonney2k | but when that is done I can include building a debian package in the nightly test | 21:12 |
blackburn | sonney2k: should we update ui? | 21:13 |
@sonney2k | blackburn, update in which respect? | 21:13 |
blackburn | e.g. my preprocs are not available here | 21:14 |
blackburn | new kernels too | 21:14 |
@sonney2k | blackburn, I wouldn't do it right now | 21:14 |
blackburn | not for release? | 21:14 |
@sonney2k | I think time is more well spend on developing a working swig-R / swig-matlab than trying to get this to work | 21:15 |
@sonney2k | blackburn, no | 21:15 |
@sonney2k | modular interfaces is what people should use to get all the features | 21:15 |
blackburn | do you plan to kill it some day? | 21:15 |
@sonney2k | blackburn, if things would work for matlab and R yes | 21:15 |
@sonney2k | until then no | 21:16 |
blackburn | I see | 21:16 |
blackburn | sonney2k: I think about simple task for evgeniy=gsomix | 21:17 |
blackburn | fixing warnings is not interesting for him (as expected) :D | 21:18 |
@sonney2k | blackburn, how about JL's covertree? | 21:20 |
blackburn | it seems he finds it too difficult | 21:20 |
blackburn | may be logistic regression? | 21:20 |
blackburn | will ask | 21:20 |
@sonney2k | blackburn, we have logistic regression in there (in liblinear) | 21:21 |
blackburn | okay klr | 21:21 |
blackburn | we have no klr | 21:21 |
blackburn | mean kerner logistic regression | 21:21 |
blackburn | kernel | 21:21 |
@sonney2k | ok | 21:21 |
blackburn | ehhr | 21:21 |
blackburn | but I would say it is more difficult hehe | 21:21 |
blackburn | he is first-year undergraduate in physics - not really familiar with probabilities stuff | 21:22 |
blackburn | but we will see | 21:22 |
@sonney2k | blackburn, alternative is simple file i/o stuff - like support nd-array i/o | 21:22 |
@sonney2k | maybe not that interesting | 21:23 |
blackburn | aha | 21:23 |
@sonney2k | or developing some tutorial - I mean examples for tutorial? | 21:23 |
@sonney2k | and then cool figures for everything? | 21:24 |
blackburn | yeah | 21:24 |
blackburn | will ask too | 21:24 |
@sonney2k | 6 more rebuilds of shogun and we will know which commit caused the problem :) | 21:24 |
@sonney2k | 2 more rebuilds left | 21:36 |
@sonney2k | between aug 1 and 16 now | 21:36 |
blackburn | interesting | 21:38 |
blackburn | sonney2k: I just noticed no bettyboo and mlsec here hehe | 21:39 |
blackburn | okay time to warning safari | 21:40 |
@sonney2k | blackburn, servers at tu were moved ... | 21:40 |
@sonney2k | blackburn, I found the reason for the serialization crash | 22:13 |
blackburn | sonney2k: and it is ..? | 22:13 |
@sonney2k | how to explain... | 22:13 |
@sonney2k | no idea how to explain | 22:14 |
@sonney2k | it is caused by removal of new[]() | 22:14 |
blackburn | oh | 22:14 |
@sonney2k | reason is that new Bla[123](); calls default constructor for Bla | 22:15 |
@sonney2k | but that is not all | 22:15 |
@sonney2k | I created SGVector object | 22:15 |
@sonney2k | that contains a virtual function | 22:15 |
@sonney2k | as soon as the virtual function is in there -> kaboom | 22:16 |
blackburn | uh | 22:16 |
@sonney2k | that is sth I would never have found w/o git bisect | 22:16 |
blackburn | yeah | 22:16 |
@sonney2k | thank $DEITY we switched to git | 22:17 |
blackburn | hmm.. warning hunt becomes harder | 22:18 |
blackburn | there was easy 100-200 warning killers | 22:18 |
@sonney2k | great one can even automatically run git bisect calling test scripts :) | 22:19 |
blackburn | sonney2k: one more thing worrying me: there are a lot of svm solvers / etc but ow can we manage to update it if origin updated | 22:19 |
blackburn | I mean it definitely have bugs and we don't watch origins | 22:20 |
@sonney2k | blackburn, only libsvm / liblinear gets updated | 22:20 |
@sonney2k | I watch from time to time | 22:20 |
blackburn | other things died? :D | 22:20 |
@sonney2k | and do vimdiff style merges | 22:20 |
@sonney2k | no | 22:20 |
@sonney2k | they just work as they did years ago | 22:20 |
blackburn | no bugs still? | 22:21 |
@sonney2k | blackburn, well I merged e.g. svmlight version 6.xx so there is hope svmlight would work by then | 22:21 |
blackburn | why there are a lot of svm solvers? | 22:22 |
@sonney2k | and libsvm usually does some enhancements not really helpful to us (x-validation ...) | 22:22 |
@sonney2k | everyone has its strenght and it is great if you can compare speed stability... | 22:22 |
@sonney2k | s/everyone/each | 22:22 |
blackburn | why there are no pretty good for every task? | 22:23 |
blackburn | thanks %DEITY I work on exact methods haha | 22:24 |
blackburn | sonney2k: btw the best thing I did in project is arpack wrapper - really useful but still only for me | 22:25 |
@sonney2k | welcome to numerics | 22:27 |
@sonney2k | all solutions are just approximations | 22:27 |
@sonney2k | some approximations are worse than others under certain condtions (parameters) for certain properties of the data ... | 22:28 |
blackburn | crazy | 22:28 |
@sonney2k | blackburn, even a+b+c != c+b+a in floating point | 22:29 |
blackburn | yeah sure but I don't grasp why svm opt task's solution ways are so different | 22:30 |
@sonney2k | blackburn, for non-floating point numbers they would all (after infinite) iterations converge to the same solution | 22:31 |
@sonney2k | but in real-life one wants to stop more early :D | 22:31 |
blackburn | sonney2k: have you seen my picture with parkinsons data dimreduction? | 22:34 |
@sonney2k | no | 22:35 |
blackburn | sonney2k: http://dl.dropbox.com/u/10139213/shogun/parkinsons.pdf | 22:35 |
blackburn | useless embedding though | 22:35 |
blackburn | voice freq data 18d->2d | 22:36 |
@sonney2k | blackburn, did you %ignore SGIO? | 22:36 |
blackburn | yeah | 22:36 |
blackburn | is it needed? :D | 22:36 |
blackburn | SHIT | 22:37 |
blackburn | crossvalidationresult too | 22:37 |
@sonney2k | and version too | 22:37 |
@sonney2k | I am unignoring them :) | 22:37 |
blackburn | java modular will fail | 22:37 |
blackburn | okay | 22:37 |
blackburn | I will add dummy for them | 22:38 |
blackburn | I really like heiko's code | 22:39 |
blackburn | I definitely should marry him | 22:40 |
@sonney2k | blackburn, tell him tomorrow | 22:40 |
blackburn | yeah sure | 22:40 |
@sonney2k | we need a big hug'ing session :D | 22:40 |
blackburn | heh | 22:40 |
blackburn | sonney2k: will you commit unignoring now? | 22:42 |
@sonney2k | done | 22:43 |
CIA-87 | shogun: Soeren Sonnenburg master * r6446418 / src/interfaces/modular/modshogun_ignores.i : unignore a few more classes like SGIO, version, ... - http://bit.ly/oOCsvW | 22:43 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 22:43 | |
blackburn | sonney2k: heiko's python example will work again | 22:44 |
@sonney2k | yeah | 22:44 |
blackburn | I hopefully fixed java | 22:44 |
blackburn | will check | 22:44 |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 22:44 | |
@sonney2k | blackburn, uhoh the bug is more involved then I thouhgt | 22:47 |
blackburn | is serialization still fucked up? | 22:47 |
@sonney2k | it was only working because we were lucky | 22:47 |
@sonney2k | we need to treat SGVector/SGMatrix/... differently in serialization | 22:48 |
@sonney2k | heiko made them special cases of the double* x, int len stuff | 22:48 |
@sonney2k | but that won't work | 22:48 |
blackburn | ehh | 22:48 |
@sonney2k | because on loading the ptr is cast as e.g. doulbe ptr | 22:48 |
@sonney2k | not SGVector | 22:49 |
@sonney2k | so kaboom | 22:49 |
@sonney2k | but only when using virtual because the vtable is at the beginning of the SGVector object | 22:49 |
@sonney2k | if we don't have virtual in there the SGVector object starts with a T* vector | 22:49 |
@sonney2k | so just because we are lucky it was starting with the object ptr!!! | 22:49 |
@sonney2k | pure horror | 22:50 |
blackburn | hehe | 22:50 |
blackburn | a lot of sorrow with serialization | 22:50 |
@sonney2k | I hope heiko doesn't get totally demotivated when attempting to fix these bugs | 22:51 |
@sonney2k | not really his code / task but sooo useful / important | 22:51 |
CIA-87 | shogun: Soeren Sonnenburg master * r49ff851 / src/shogun/lib/DataType.h : temporary workaround for crasher in serialization - http://bit.ly/ov830I | 23:00 |
@sonney2k | blackburn, ok python modular tests run except for structure_dynprog_modular.py | 23:00 |
blackburn | sonney2k: why? | 23:00 |
@sonney2k | why what? | 23:00 |
blackburn | dynprog | 23:00 |
blackburn | fails | 23:00 |
@sonney2k | SGVector transition | 23:00 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Read error: Connection reset by peer] | 23:01 | |
blackburn | ah | 23:01 |
CIA-87 | shogun: Soeren Sonnenburg master * r54d0eba / src/shogun/mathematics/Math.h : fix compile error when logsumarray is enabled - http://bit.ly/mW7F22 | 23:06 |
@sonney2k | another bugfix for tomorrows nightly_all | 23:06 |
blackburn | can't expand SGIO for some reason | 23:11 |
@sonney2k | extend you mean? | 23:17 |
blackburn | yeah | 23:18 |
blackburn | same for dummkopf like me hah | 23:18 |
blackburn | sonney2k: SGIO have no public constructor in swig! | 23:51 |
@sonney2k | ? | 23:54 |
@sonney2k | why not? | 23:54 |
blackburn | фр | 23:54 |
blackburn | ah | 23:54 |
blackburn | it shouldn't have I think | 23:54 |
blackburn | but it doesnt' work | 23:54 |
CIA-87 | shogun: Shashwat Lal Das master * r094d8ec / (4 files in 4 dirs): Corrections to expand_if_required for StreamingVwFeatures.Added StreamingVwFeatures example. - http://bit.ly/rn9r6h | 23:55 |
CIA-87 | shogun: Soeren Sonnenburg master * r4b721c8 / (4 files in 4 dirs): | 23:55 |
CIA-87 | shogun: Merge pull request #300 from frx/streaming_vw | 23:55 |
CIA-87 | shogun: SGD using CStreamingVwFeatures - http://bit.ly/oyrog3 | 23:55 |
CIA-87 | shogun: Sergey Lisitsyn master * ra419f9b / (4 files in 4 dirs): Merge branch 'master' of github.com:shogun-toolbox/shogun - http://bit.ly/nnNLiU | 23:57 |
CIA-87 | shogun: Sergey Lisitsyn master * r8f5d697 / (3 files): Java externalization fixes - http://bit.ly/qujIpt | 23:57 |
blackburn | no idea how to get SGIO serializable | 23:57 |
@sonney2k | blackburn, what is the difference to other classes? | 23:57 |
@sonney2k | (I don't see any) | 23:58 |
blackburn | didn't find any | 23:58 |
blackburn | it says no save_serializable | 23:58 |
blackburn | but it is here! | 23:58 |
@sonney2k | that doesn't really make sense | 23:59 |
--- Log closed Wed Aug 17 00:00:59 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!