--- Log opened Thu Nov 29 00:00:15 2012 | ||
wiking | this is funny | 00:02 |
---|---|---|
wiking | n4nd0: i've just found 3 different papers with 3 different author sets on 3 different publication years talking about essentially the same | 00:02 |
n4nd0 | nice! | 00:03 |
n4nd0 | what about? | 00:03 |
wiking | n4nd0: apparently i've found the solution for this approx | 00:17 |
wiking | indeed | 00:18 |
n4nd0 | cool | 00:18 |
n4nd0 | what is it? | 00:18 |
wiking | 1/\beta \log \sum exp(\beta * f(x,y,w)) | 00:18 |
wiking | lim \beta -> \inf | 00:18 |
wiking | will result in | 00:18 |
wiking | max (f(x,y,w)) | 00:19 |
wiking | i mean | 00:19 |
wiking | i've missed \delta(y,y_i) | 00:19 |
wiking | but you know what i mena | 00:19 |
wiking | mean | 00:19 |
n4nd0 | yeah I kind of get the idea | 00:19 |
n4nd0 | I don't know right now if that is the true answer for the limit | 00:19 |
n4nd0 | but I believe you :) | 00:19 |
wiking | well in the paper they even have a proof for it | 00:20 |
wiking | so yeah i trust them as well | 00:20 |
wiking | actually i've read it | 00:21 |
wiking | it makes sense | 00:21 |
wiking | so anyhow | 00:21 |
wiking | we could add a softmax thing for so learning | 00:21 |
n4nd0 | for the so-svm? | 00:22 |
wiking | yes | 00:22 |
n4nd0 | ok | 00:22 |
n4nd0 | do you get something similar to a CRF then? | 00:22 |
wiking | yes | 00:23 |
wiking | \beta = 0 CRF | 00:23 |
wiking | \beta = inf SSVM | 00:23 |
n4nd0 | interesting | 00:23 |
wiking | and it's super easy to implement this with BMRM | 00:24 |
wiking | n4nd0: Entropy and Margin Maximization for Structured Output Learning | 00:25 |
wiking | check this paper | 00:25 |
wiking | mmm | 00:36 |
wiking | this paper is cool | 00:36 |
wiking | i've just realised that we can do MIL with latent framework :) | 00:36 |
n4nd0 | what's MIL? | 00:37 |
wiking | multiple instance learning | 00:37 |
wiking | so you get n set of (x_1,...,x_n) | 00:38 |
wiking | and u know that each set has a label y | 00:38 |
n4nd0 | haha it seems that new types of learning don't stop popping out | 00:38 |
wiking | but you don't know the exact label for each x_i | 00:38 |
wiking | and of course it could be that you get a set of x-es | 00:39 |
wiking | with label y_i | 00:39 |
wiking | but that some of those x-es are not really y_i | 00:39 |
wiking | We construct a one-dimensional synthetic dataset which illustrates the de- ficiencies of the MI-SVM. A positive bag consists of p positive instances and 50 - p negative (0 < p <= 50), a negative bag contains 50 negative instances. The individual instances are hard to classify: the positive instances are Gaussian dis- tributed with mean 0.6 whereas the negative instances are Gaussian distributed with mean 0, the variance for both classes is 1. | 00:40 |
wiking | this is what they do | 00:40 |
n4nd0 | aham it sounds cool | 00:41 |
n4nd0 | a way to get rid of data outliers maybe | 00:41 |
wiking | let see if i can do this | 00:42 |
wiking | should be easy to generate this dataset :) | 00:42 |
n4nd0 | yeah | 00:42 |
n4nd0 | time to sleep | 00:53 |
n4nd0 | good night! | 00:53 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 00:55 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 01:28 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has quit [Quit: Ex-Chat] | 02:41 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has joined #shogun | 02:41 | |
shogun-buildbot | build #158 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/158 | 03:11 |
shogun-buildbot | build #189 of nightly_default is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/189 | 03:23 |
-!- blackburn [~blackburn@188.168.3.0] has quit [Ping timeout: 246 seconds] | 04:57 | |
-!- blackburn [~blackburn@37.61.179.230] has joined #shogun | 05:22 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has quit [Read error: Connection reset by peer] | 06:31 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has joined #shogun | 06:32 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has quit [Ping timeout: 250 seconds] | 07:24 | |
-!- ptizoom [~christian@79-71-89-182.dynamic.dsl.as9105.com] has joined #shogun | 07:25 | |
-!- blackburn [~blackburn@37.61.179.230] has quit [Quit: Leaving.] | 08:12 | |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/228f6ab20ea16dd35d4c0f4056f93ef9507ac047 | 08:26 | |
-shogungit:#shogun- shogun/master 228f6ab Soeren Sonnenburg: drop -fPIC / pthreads on cygwin | 08:26 | |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/9417651b7b96268510c64454e720cee08a3a4328 | 08:30 | |
-shogungit:#shogun- shogun/master 9417651 Soeren Sonnenburg: fix compile error on cygwin | 08:30 | |
shogun-buildbot | build #458 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/458 blamelist: Soeren Sonnenburg <sonne@debian.org> | 08:35 |
shogun-buildbot | build #459 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/459 blamelist: Soeren Sonnenburg <sonne@debian.org> | 08:36 |
shogun-buildbot | build #694 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/694 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:04 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/d517033406b1d90e4fb6400224762a951b843ab9 | 09:11 | |
-shogungit:#shogun- shogun/master d517033 Soeren Sonnenburg: make pthreads optional | 09:11 | |
shogun-buildbot | build #460 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/460 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:26 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/c333f44e165a880497253af95d049831c852cb52 | 09:29 | |
-shogungit:#shogun- shogun/master c333f44 Soeren Sonnenburg: fix error: cannot convert 'shogun::SGMatrix<double>' to 'const float64_t*' in assignment... | 09:29 | |
shogun-buildbot | build #695 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/695 | 09:32 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 09:37 | |
shogun-buildbot | build #461 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/461 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:39 |
blackburn | someone has stolen a compiler :( | 09:41 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/fab840bb26e039e7ae8a703ab4fd643ba31e3c36 | 09:49 | |
-shogungit:#shogun- shogun/master fab840b Soeren Sonnenburg: remove leftin function call | 09:49 | |
sonne|work | blackburn: cleaned up your desk? | 09:49 |
blackburn | sonne|work: haha I am cleaning up my tasks | 09:49 |
shogun-buildbot | build #696 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/696 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:50 |
sonne|work | it is important that they shine once you leave :) | 09:50 |
blackburn | well I still have 3 tickets in bugtracking | 09:50 |
blackburn | sonne|work: and hey as I don't use anything but computer I have nothing on my desk | 09:51 |
sonne|work | hmmhh buildslave #4 seems to have some issues | 09:53 |
blackburn | good to have slaves in the modern world we live | 09:53 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 09:55 | |
blackburn | okay I described one problem through phase of moon | 09:55 |
blackburn | n4nd0: aha! catch you | 09:56 |
n4nd0 | blackburn: hello hello :) | 09:56 |
shogun-buildbot | build #462 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/462 | 09:59 |
shogun-buildbot | build #463 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/463 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:59 |
-!- cavalier- [~cavalier-@70.118.110.73] has quit [Read error: Connection reset by peer] | 10:02 | |
-!- cavalier- [cavalier-@70.118.110.73] has joined #shogun | 10:02 | |
wiking | sonne|work: i'm getting this warning with the HEAD of the master branch: ../shogun/lib/SGSparseVector.h:26:27: note: previous use is here | 10:05 |
wiking | template <class T> struct SGSparseVectorEntry ^ | 10:05 |
wiking | ../shogun/io/File.h:23:21: note: did you mean struct here? | 10:05 |
wiking | template <class ST> class SGSparseVectorEntry; | 10:05 |
shogun-buildbot | build #698 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/698 blamelist: Soeren Sonnenburg <sonne@debian.org> | 10:23 |
shogun-buildbot | build #697 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/697 | 10:28 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/a85550094ec860e020803d5a7b5540fd285e3079 | 10:31 | |
-shogungit:#shogun- shogun/master a855500 Soeren Sonnenburg: fix compile error when pthreads are not available | 10:31 | |
sonne|work | previous? | 10:33 |
sonne|work | what? | 10:33 |
wiking | yes | 10:34 |
wiking | sonne|work: i mean in SGSparseVector you declare SGSparseVectorEntry as struct | 10:34 |
wiking | but then for example in File.h you refere to it as a class | 10:35 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/8a5274f3ea1a32182a5e6c6af08fc012a48df8c9 | 10:35 | |
-shogungit:#shogun- shogun/master 8a5274f Soeren Sonnenburg: SGSparseVectorEntry is in fact a struct | 10:35 | |
wiking | voila | 10:35 |
sonne|work | wiking: does it build now? | 10:45 |
wiking | it was building before | 10:46 |
wiking | just flooding with warnings about this | 10:46 |
wiking | yeah better thnx | 10:47 |
sonne|work | k | 10:48 |
shogun-buildbot | build #464 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/464 blamelist: Soeren Sonnenburg <sonne@debian.org> | 10:50 |
wiking | ok i thnk i have an awesome example coming up for shogun | 10:52 |
wiking | it's even graphical | 10:52 |
wiking | so you can post it to the webpage | 10:52 |
blackburn | does it contain tits? | 10:53 |
wiking | blackburn: :D | 10:54 |
wiking | sonne|work: mmm i think there's a way to connect buildbots as well to the pull request hook in github | 10:54 |
sonne|work | wiking: ? | 10:55 |
wiking | sonne|work: well i've just seen this | 10:55 |
sonne|work | it is | 10:55 |
wiking | sonne|work: https://github.com/Itseez/opencv/pull/181 | 10:55 |
sonne|work | ahh | 10:55 |
sonne|work | ok so you mean even before commit | 10:55 |
wiking | sonne|work: and that link there for good to merge | 10:55 |
wiking | is not coming from travis | 10:55 |
wiking | it's from a buildbot | 10:55 |
wiking | sonne|work: indeed, so it's like travis | 10:56 |
wiking | sonne|work: ah ok i think i know how it works | 10:58 |
wiking | sonne|work: you need to define in service hooks | 10:58 |
wiking | a webhook url | 10:58 |
wiking | as it says | 10:58 |
wiking | "We'll hit these URLs with POST requests when you push to us, passing along information about the push. More information can be found in the Post-Receive Guide." | 10:58 |
wiking | sonne|work: and it seems there should be a module for this to be handled in buildbot | 10:59 |
wiking | http://trac.buildbot.net/ticket/2186 | 11:00 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 264 seconds] | 11:00 | |
wiking | althoguh i couldnt' find any docs about this | 11:01 |
wiking | sonne|work: http://buildbot.net/buildbot/docs/0.8.3/reference/buildbot.status.web.hooks.github-module.html | 11:01 |
shogun-buildbot | build #465 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/465 blamelist: Soeren Sonnenburg <sonne@debian.org> | 11:04 |
sonne|work | wiking: not so clear to me how to set this up - I mean we don't want to mix commits with committed stuff | 11:06 |
sonne|work | blackburn: cyg1 is still failing with your last commit :/ | 11:07 |
blackburn | ja | 11:08 |
shogun-buildbot | build #699 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/699 | 11:15 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 11:15 | |
shogun-buildbot | build #700 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/700 blamelist: Soeren Sonnenburg <sonne@debian.org> | 11:19 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 12:37 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 15:55 | |
--- Log closed Thu Nov 29 18:10:38 2012 | ||
--- Log opened Thu Nov 29 18:10:47 2012 | ||
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun | 18:10 | |
-!- Irssi: #shogun: Total of 9 nicks [1 ops, 0 halfops, 0 voices, 8 normal] | 18:10 | |
-!- Irssi: Join to #shogun was synced in 6 secs | 18:10 | |
wiking | mmm it's weird | 19:47 |
wiking | can't i make a 1 dim DenseFeatures via numpy | 19:48 |
wiking | ? | 19:48 |
wiking | i mean features dim = 1 | 19:48 |
wiking | it seems so :( | 19:52 |
wiking | sonne|work: ping? | 19:52 |
@sonney2k | wiking, wassup? | 20:24 |
@sonney2k | wiking, [[1,2,3]] | 20:25 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds] | 20:29 | |
-!- blackburn [~blackburn@188.168.3.33] has joined #shogun | 20:57 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 21:21 | |
wiking | mmm | 21:21 |
wiking | i wonder what was happening here | 21:21 |
wiking | :) | 21:21 |
wiking | sonney2k: yeah in the meanwhile i've found out that i needed to transpose that vector | 21:22 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 22:01 | |
-!- blackburn [~blackburn@188.168.3.33] has quit [Quit: Leaving.] | 22:02 | |
@sonney2k | wiking, I finally understand why the buildbot is failing on cygwin | 22:04 |
@sonney2k | we use a reserved word | 22:04 |
@sonney2k | __from! nothing should be __'t | 22:05 |
wiking | ?? | 22:07 |
wiking | lolface | 22:07 |
wiking | yeah that's uricamic's coe | 22:07 |
wiking | code | 22:07 |
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/1f7a6cca7f7bb58616eb0f2b9f884617c49b19f2 | 22:08 | |
-shogungit:#shogun- shogun/master 1f7a6cc Soeren Sonnenburg: fixup cygwin build -> never use __reserved __keywords! | 22:08 | |
wiking | hahah i bet blackburn woulld love this nameing: MILFeature ;) | 22:08 |
@sonney2k | wiking, no blackburn introduced some convenience constructor + TMultipleCPinfo(uint32_t __from, uint32_t _N) : _from(__from), N(_N) { } | 22:09 |
@sonney2k | and that kills it | 22:09 |
@sonney2k | took me ages to find out :/ | 22:09 |
wiking | :( | 22:10 |
wiking | yeah i would have never guessed that __* is reserved in cygwin's gcc | 22:10 |
@sonney2k | wiking, btw I wrote some example for dynarray and likely found the issue | 22:10 |
wiking | sonney2k: great! | 22:10 |
@sonney2k | we memset(0) the newly allocated memory | 22:10 |
wiking | sonney2k: yeah i've seen that | 22:10 |
@sonney2k | that doesn't work with objects of course | 22:10 |
wiking | but it's only memsetting the new area no? | 22:10 |
@sonney2k | kaboom | 22:10 |
@sonney2k | sure but that would be and SGVector sitting on that new area | 22:11 |
@sonney2k | memset it | 22:11 |
@sonney2k | and then at some point it gets destructed | 22:11 |
@sonney2k | I can easily fix this - only question is if sth relies on the memset 0 behavior | 22:11 |
wiking | who's using DynArray? | 22:12 |
@sonney2k | lots of stuff | 22:12 |
wiking | ah well then yeah good luck with debugging it out | 22:12 |
wiking | :( | 22:12 |
wiking | btw something like this would be fun as well | 22:13 |
wiking | http://lxr.linux.no/+trees | 22:13 |
@sonney2k | good thing with the new malloc templates | 22:13 |
@sonney2k | is though that one can fix it | 22:13 |
@sonney2k | wiking, I actually think that this might not be a problem at all - let me commit and see what fails | 22:15 |
wiking | heheh glhf | 22:15 |
wiking | ;) | 22:15 |
wiking | wtf: TypeError: not a numpy vector of appropriate type | 22:29 |
shogun-buildbot | build #466 of cyg1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/466 | 22:30 |
wiking | sonney2k: seems u fixed it! \o/ | 22:37 |
shogun-buildbot | build #701 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/701 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:42 |
--- Log closed Fri Nov 30 00:00:15 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!