--- Log opened Sat Apr 07 00:00:15 2012 | ||
--- Day changed Sat Apr 07 2012 | ||
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 00:00 | |
blackburn | hey there | 00:02 |
---|---|---|
blackburn | anybody out there? :) | 00:02 |
PhilTillet | hey | 00:03 |
PhilTillet | yes | 00:03 |
-!- harshit_ [~harshit@182.68.94.221] has joined #shogun | 00:03 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has joined #shogun | 00:07 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has quit [Changing host] | 00:07 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 00:07 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has quit [Remote host closed the connection] | 00:14 | |
CIA-64 | shogun: iglesias master * rfbc2f8f / (src/shogun/classifier/KNN.cpp src/shogun/classifier/KNN.h): * convention and naming fixes - http://git.io/M-rhjQ | 00:26 |
CIA-64 | shogun: Soeren Sonnenburg master * r8d1fd5d / (4 files in 3 dirs): | 00:26 |
CIA-64 | shogun: Merge pull request #426 from iglesias/knn-covertree | 00:26 |
CIA-64 | shogun: CoverTree integration for KNN - http://git.io/Vf-hNg | 00:26 |
@sonney2k | blackburn, look https://github.com/shogun-toolbox/shogun/pull/428 | 00:27 |
@sonney2k | heiko did it again :) | 00:27 |
blackburn | karlnapf: you rock | 00:27 |
@sonney2k | karlnapf, why does x-val get slower | 00:27 |
karlnapf | sonney2k, blackburn, hi, thanks :) | 00:28 |
karlnapf | sonney2k, because a subset is pushed in every fold, and then the active subset has to be recomputed | 00:28 |
karlnapf | that creates some overhead | 00:28 |
karlnapf | esp for large number of folds, large number of repetitions compared to small amount of data (as in examples) | 00:29 |
blackburn | what is overhead? | 00:29 |
@sonney2k | karlnapf, I would prefer 'add_subset' / 'remove_subset' just because we use the same terms with add_kernel / add_feature ... | 00:29 |
karlnapf | but should not be noticable in large scale real world examples | 00:29 |
karlnapf | sonney2k, ok I can change this | 00:29 |
blackburn | I thought this overhead is rather light | 00:29 |
@sonney2k | karlnapf, but isn't this a NOP when there is just 1 subset? | 00:29 |
@sonney2k | I mean it doesn't need to recompute anything or? | 00:29 |
karlnapf | oh yes, true | 00:30 |
karlnapf | I should add that :) | 00:30 |
karlnapf | currently it does (copy) | 00:30 |
karlnapf | blackburn, I just noticed that for my examples it takes a bit longer | 00:31 |
@sonney2k | karlnapf, nice work! | 00:33 |
@sonney2k | do these 'fixes' and TODO's and merge at your own will :D | 00:33 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 00:33 | |
karlnapf | sonney2k, alright :) | 00:33 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 00:36 | |
blackburn | that connection sucks | 00:36 |
blackburn | have I missed anything? | 00:36 |
blackburn | I am curious whether commits can go to 2 github issues | 00:37 |
blackburn | with number # and pull request | 00:37 |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has joined #shogun | 00:39 | |
karlnapf | blackburn, no i you did not miss anything :) | 00:39 |
karlnapf | Ill go to bed | 00:39 |
karlnapf | blackburn, good night! | 00:39 |
karlnapf | sonney2k, good night! | 00:39 |
blackburn | karlnapf: good night | 00:40 |
-!- karlnapf [~heiko@host86-185-9-173.range86-185.btcentralplus.com] has left #shogun [] | 00:43 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Quit: Page closed] | 00:54 | |
-!- harshit_ [~harshit@182.68.94.221] has quit [Remote host closed the connection] | 01:18 | |
-!- pluskid [~chatzilla@111.120.47.200] has joined #shogun | 02:23 | |
shogun-buildbot | build #452 of java_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/452 blamelist: fernando.iglesiasg@gmail.com | 02:30 |
shogun-buildbot | build #448 of python_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/448 blamelist: fernando.iglesiasg@gmail.com | 02:37 |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has quit [Ping timeout: 252 seconds] | 02:50 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has joined #shogun | 02:52 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has quit [Ping timeout: 260 seconds] | 03:02 | |
shogun-buildbot | build #449 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/449 | 03:25 |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has joined #shogun | 03:36 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has quit [Ping timeout: 272 seconds] | 03:43 | |
-!- pluskid [~chatzilla@111.120.47.200] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120312181643]] | 03:51 | |
-!- PhilTillet [~Philippe@157.159.42.154] has joined #shogun | 03:55 | |
-!- vikram360 [~vikram360@117.192.160.138] has quit [Read error: Connection reset by peer] | 04:28 | |
-!- PhilTillet [~Philippe@157.159.42.154] has quit [Ping timeout: 272 seconds] | 04:44 | |
-!- pluskid [~chatzilla@173.254.214.60] has joined #shogun | 05:19 | |
gsomix | good morning | 06:19 |
gsomix | damn, I was late for classes about complex analysis | 06:20 |
-!- n4nd0 [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 07:18 | |
-!- iglesias [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 07:35 | |
-!- n4nd0 [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Read error: Connection reset by peer] | 07:35 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 07:50 | |
-!- iglesias [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Ping timeout: 260 seconds] | 07:53 | |
-!- pythonroar [d2198538@gateway/web/freenode/ip.210.25.133.56] has joined #shogun | 07:57 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Quit: leaving] | 07:57 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 07:57 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Client Quit] | 07:57 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 07:59 | |
-!- pythonroar [d2198538@gateway/web/freenode/ip.210.25.133.56] has quit [Quit: Page closed] | 08:20 | |
n4nd0 | sonney2k: around? | 08:23 |
-!- gsomix [~gsomix@188.168.5.2] has quit [Quit: ????? ? ?? ??? (xchat 2.4.5 ??? ??????)] | 08:40 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Quit: leaving] | 09:23 | |
-!- harshit_ [~harshit@182.68.94.221] has joined #shogun | 09:39 | |
-!- n4nd0 [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 10:30 | |
-!- n4nd0 [~androirc@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Client Quit] | 10:32 | |
-!- gsomix [~gsomix@85.26.232.64] has joined #shogun | 10:47 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 10:54 | |
-!- gsomix [~gsomix@85.26.232.64] has quit [Quit: ????? ? ?? ??? (xchat 2.4.5 ??? ??????)] | 10:56 | |
-!- gsomix [~gsomix@85.26.232.64] has joined #shogun | 10:56 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has joined #shogun | 11:00 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has quit [Changing host] | 11:00 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 11:00 | |
-!- blackburn [~qdrgsm@83.234.54.186] has joined #shogun | 11:42 | |
-!- harshit_ [~harshit@182.68.94.221] has quit [Ping timeout: 240 seconds] | 11:45 | |
-!- harshit_ [~harshit@182.68.179.160] has joined #shogun | 11:57 | |
wiking | heheh it's kind of quite here since gsoc application ended :P | 12:17 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 12:17 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 12:17 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 12:20 | |
n4nd0 | sonney2k: hey! thank you for your answer, but I didn't get any qda file in data/testsuite/tests | 12:21 |
n4nd0 | maybe there is something wrong in my configuration | 12:21 |
wiking | n4nd0: have you checked out the submodule? | 12:29 |
wiking | git submodule init && git submodule update | 12:29 |
n4nd0 | wiking: the only thing I have done related to the submodule is | 12:34 |
n4nd0 | git submodule update --init | 12:34 |
n4nd0 | I did that in my shogun local repository in order to set up the data directories | 12:34 |
wiking | mmm | 12:34 |
wiking | well that should do it | 12:34 |
-!- PhilTillet [~Philippe@157.159.42.154] has joined #shogun | 12:34 | |
n4nd0 | oh wait | 12:35 |
n4nd0 | I have these files | 12:35 |
wiking | and there's nothing in data/testsuite/tests ? | 12:35 |
n4nd0 | ./tests/classifier_qda_modular1.txt | 12:35 |
n4nd0 | ./tests/classifier_qda_modular0.txt | 12:35 |
wiking | there should be at least 276 files | 12:35 |
n4nd0 | where . is shogun/data/testsuite | 12:36 |
n4nd0 | I understood it wrong, I was looking into my shogun-data repository | 12:36 |
wiking | if not more since i haven't updated my repo for a while... | 12:36 |
n4nd0 | so it looks like shogun/data is another repository right? | 12:37 |
n4nd0 | my git understands it like that at least | 12:37 |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 12:39 | |
n4nd0 | how should I update this repo then? | 12:39 |
n4nd0 | I mean, upstream and origin can be used in the same way? | 12:39 |
n4nd0 | gtg, I will try to find more about this later, thank you! | 12:41 |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Quit: leaving] | 12:41 | |
@sonney2k | uhh record number of students applied http://google-opensource.blogspot.in/ | 12:50 |
PhilTillet | wow | 12:51 |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 12:51 | |
PhilTillet | That's a lot | 12:53 |
blackburn | >>> (48.0/6685)*1100 | 12:56 |
blackburn | 7.898279730740464 | 12:56 |
PhilTillet | sonney2k, I forgot to mention in my pull request : ViennaCL only supports float and double right now (will probably evolve soon :p), that is why there is that viennacl_compatibility.h header :p | 12:58 |
blackburn | sonney2k: first allocation should be 7-8 then :) | 12:59 |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 13:01 | |
-!- harshit_ [~harshit@182.68.179.160] has quit [Remote host closed the connection] | 13:08 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 13:13 | |
-!- wiking_ [~wiking@78-23-191-201.access.telenet.be] has joined #shogun | 13:25 | |
-!- wiking_ [~wiking@78-23-191-201.access.telenet.be] has quit [Changing host] | 13:25 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun | 13:25 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 248 seconds] | 13:28 | |
-!- wiking_ is now known as wiking | 13:28 | |
-!- gsomix [~gsomix@85.26.232.64] has quit [Ping timeout: 246 seconds] | 13:42 | |
-!- PhilTillet [~Philippe@157.159.42.154] has quit [Ping timeout: 240 seconds] | 13:46 | |
wiking | mmm i really wonder if it's a good idea to use viennacl for opencl support :( | 13:54 |
-!- gsomix [~gsomix@85.26.233.247] has joined #shogun | 13:55 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has joined #shogun | 13:58 | |
-!- PhilTillet [~Philippe@npasserelle10.minet.net] has quit [Ping timeout: 252 seconds] | 14:03 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Quit: Leaving] | 14:11 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 14:11 | |
blackburn | wiking: what are your thoughts on that? | 14:20 |
wiking | i mean opencl it self has a standardized api | 14:20 |
wiking | just like opengl | 14:21 |
blackburn | ah yes | 14:21 |
wiking | so currently i really don't see way is it necessary to add an extra abstraction layer to it | 14:21 |
wiking | i mean yes | 14:22 |
wiking | it for sure makes easier the implementation | 14:22 |
wiking | but then it's just another library dependency u will need if u want support for it | 14:22 |
blackburn | yes I agree | 14:23 |
blackburn | wiking: did you submit a proposal in time? ;) | 14:27 |
-!- pluskid [~chatzilla@173.254.214.60] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120314124128]] | 14:57 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 15:04 | |
n4nd0 | blackburn: how is it going? | 15:05 |
blackburn | hey | 15:05 |
blackburn | fine but getting busier | 15:05 |
n4nd0 | university, job? | 15:06 |
blackburn | yeah I have to do some university stuff | 15:06 |
-!- PhilTillet [~android@92.90.20.8] has joined #shogun | 15:06 | |
n4nd0 | blackburn: I see, I was going to ask you about this data submodule stuff, but it can wait so no problem ;) | 15:06 |
blackburn | n4nd0: no, ask :) | 15:06 |
n4nd0 | blackburn: so I've noticed that after doing this git submodule update --init | 15:07 |
n4nd0 | blackburn: I get like another repository in shogun/data | 15:07 |
blackburn | yes | 15:08 |
n4nd0 | blackburn: I mean another repository because when I am in another directory of the project I see the branch where I'm (qda) and when I am under shogun/data I see another branch's name | 15:08 |
blackburn | yes | 15:08 |
blackburn | it is another repo | 15:08 |
n4nd0 | ok | 15:09 |
n4nd0 | so how should I manage it? | 15:09 |
n4nd0 | for example this morning I did with tester.py | 15:09 |
n4nd0 | I generated a new example for qda | 15:09 |
blackburn | probably you'd need to fork it too | 15:09 |
blackburn | sounds crazy I know | 15:09 |
n4nd0 | I see now a couple of new files classifier_qda_modular0.txt | 15:10 |
n4nd0 | and the same with modular1 | 15:10 |
n4nd0 | mmm but actually I forked shogun-data | 15:10 |
n4nd0 | I have it locally | 15:10 |
n4nd0 | but it is in another directory in my machine than the one I was telling you | 15:10 |
n4nd0 | in one placed I hace the forked I did of shogun-data | 15:11 |
n4nd0 | and inside shogun I have this shogun/data/ repository | 15:11 |
n4nd0 | I don't know if that really makes sense | 15:11 |
blackburn | ohh | 15:11 |
blackburn | probably you need to do PR with your shogun-data fork | 15:11 |
n4nd0 | ok | 15:12 |
n4nd0 | but in my shogun-data fork it is not where I can see these new files+ | 15:12 |
n4nd0 | this is getting weird :P | 15:12 |
blackburn | just put your files to your fork | 15:12 |
n4nd0 | ok | 15:12 |
blackburn | once your PR is merged I'll update data hash in main shogun repo | 15:13 |
blackburn | and it would be ok | 15:13 |
n4nd0 | ok | 15:13 |
n4nd0 | I'll move the files and do the PR now, first I'll rebase and that stuff | 15:14 |
n4nd0 | btw I checked the contents of these files | 15:14 |
n4nd0 | really cryptic | 15:14 |
CIA-64 | shogun: Sergey Lisitsyn master * ra553686 / src/shogun/classifier/svm/NewtonSVM.cpp : Proper memory handling for #409 - http://git.io/n0rgBQ | 15:14 |
blackburn | yeah but it would be worse to have data in the main repo | 15:15 |
wiking | blackburn: heheh yes i did :) | 15:16 |
blackburn | that's nice then | 15:16 |
n4nd0 | blackburn: oh sorry, I think that was confusing ... I meant that the contents of these files (classifier_qda_modular[01].txt) I generated for the testsuite are really weird | 15:18 |
n4nd0 | blackburn: I cannot understand them at least | 15:18 |
blackburn | hmmm | 15:18 |
n4nd0 | I suppose it has something to do with serialization | 15:18 |
blackburn | n4nd0: is it similar to others ? | 15:18 |
n4nd0 | let me check | 15:19 |
n4nd0 | blackburn: yeah, it is as weird as classifier_knn_modular0.txt for example | 15:20 |
blackburn | all ok then | 15:20 |
blackburn | ;) | 15:20 |
n4nd0 | blackburn: do you know sth about how the contents of these files work? | 15:20 |
n4nd0 | S?ren told me yesterday that it would be useful if I work with the testsuite, trying to fix the errors | 15:21 |
blackburn | n4nd0: yes it is pickle | 15:21 |
n4nd0 | blackburn: python serialization right? | 15:21 |
blackburn | yes | 15:21 |
n4nd0 | the thing is that with S?ren's suggestion, when some test is wrong, I execute the tester like this | 15:22 |
n4nd0 | ./tester.py -d wrong_test.py | 15:22 |
n4nd0 | that basically shows me the contents of those files | 15:22 |
n4nd0 | I don't know how one is supposed to find the errors from the contents there | 15:22 |
blackburn | hah | 15:23 |
blackburn | no way | 15:23 |
blackburn | did he suggest to compare the outputs? he was pretty tired probably hah | 15:24 |
-!- PhilTillet [~android@92.90.20.8] has quit [Read error: Connection reset by peer] | 15:24 | |
blackburn | n4nd0: ok if seriously you probably would need to read this stuff and compare parameters | 15:24 |
n4nd0 | blackburn: but I think that stuff is not readable at all :) | 15:25 |
n4nd0 | blackburn: check for example these new ones https://github.com/iglesias/shogun-data/commit/e0c8878d2515e94cf7997c3f22df7bfaca3bd00d | 15:25 |
blackburn | no | 15:25 |
blackburn | -d should show some detailed sutff | 15:26 |
n4nd0 | that is what I get when I do -d | 15:26 |
n4nd0 | without -d is just ERROR | 15:27 |
blackburn | yes | 15:27 |
blackburn | something is wrong there then | 15:27 |
blackburn | check compare_dbg function | 15:27 |
blackburn | hmmm | 15:28 |
blackburn | I see | 15:28 |
blackburn | line 57 is kinda stupid | 15:28 |
blackburn | n4nd0: ok I think we need to change examples that return SGObjects | 15:29 |
blackburn | that makes no sense to me to compare that | 15:29 |
n4nd0 | blackburn: it makes no sense to me either | 15:30 |
n4nd0 | blackburn: also I think that something must be missing because S?ren told me sth like | 15:30 |
blackburn | n4nd0: but numpy array compare is ok | 15:30 |
n4nd0 | with -d you can do debu | 15:30 |
n4nd0 | debug sorry | 15:30 |
n4nd0 | with print a you can print object a | 15:30 |
n4nd0 | with print b you can print object b | 15:30 |
n4nd0 | I have not entered in that kind of debug mode though :( | 15:31 |
blackburn | yeah reading pickle is kind of mastery | 15:32 |
-!- PhilTillet [~android@92.90.20.8] has joined #shogun | 15:37 | |
-!- gsomix [~gsomix@85.26.233.247] has quit [Ping timeout: 245 seconds] | 15:41 | |
-!- gsomix [~gsomix@85.26.233.247] has joined #shogun | 15:43 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 15:51 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Quit: leaving] | 15:56 | |
-!- PhilTillet [~android@92.90.20.8] has quit [Ping timeout: 252 seconds] | 15:57 | |
-!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has joined #shogun | 16:19 | |
PhilTillet | blackburn, I think there is still a "float64_t" cast in the buggy pull request i sent a few weeks ago ( CMath::dot(float32_t*,float32_t*,int32_t)) | 16:22 |
blackburn | really? why? | 16:23 |
PhilTillet | I don't know, I rebased the upstream but there was still a float64_t cast | 16:23 |
PhilTillet | even though dsdot was back to sdot | 16:24 |
blackburn | due to performance reasons we changed it back to sdot | 16:27 |
PhilTillet | yes | 16:27 |
blackburn | so what is the problem? :) | 16:27 |
PhilTillet | int he for loop there is still the float64_t cast | 16:27 |
PhilTillet | which halves performances :p | 16:27 |
blackburn | ah | 16:28 |
blackburn | do you think it should be removed? | 16:28 |
PhilTillet | yes, I tested without it and peformances for float doubled :p | 16:29 |
PhilTillet | I think multiplying a float with a double is a bad idea with SSE | 16:29 |
blackburn | yeah | 16:29 |
blackburn | however on my machine blas is used probably | 16:30 |
PhilTillet | hmm, you probably have #define HAVE_LAPACK | 16:30 |
CIA-64 | shogun: Sergey Lisitsyn master * rca1f345 / src/shogun/mathematics/Math.cpp : Performance fix for single precision dot operation - http://git.io/Qd0MXA | 16:30 |
PhilTillet | that was a fail pull request of mine :D | 16:31 |
-!- gsomix [~gsomix@85.26.233.247] has left #shogun ["????? ? ?? ??? (xchat 2.4.5 ??? ??????)"] | 16:46 | |
-!- gsomix [~gsomix@85.26.233.247] has joined #shogun | 16:49 | |
-!- harshit_ [~harshit@182.68.254.100] has joined #shogun | 16:50 | |
-!- puneet [~puneet@115.242.119.39] has joined #shogun | 16:53 | |
-!- blackburn [~qdrgsm@83.234.54.186] has quit [Quit: Leaving.] | 17:14 | |
-!- blackburn [~qdrgsm@83.234.54.186] has joined #shogun | 17:18 | |
-!- puneet [~puneet@115.242.119.39] has quit [Ping timeout: 260 seconds] | 17:22 | |
harshit_ | blackburn: hi ! | 17:32 |
harshit_ | having some problems with CDotFeatures .. | 17:32 |
harshit_ | in implementing lbp | 17:32 |
blackburn | hi | 17:32 |
-!- harshit_ [~harshit@182.68.254.100] has quit [Quit: Leaving] | 17:34 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 18:02 | |
-!- gsomix [~gsomix@85.26.233.247] has quit [Quit: ????? ? ?? ??? (xchat 2.4.5 ??? ??????)] | 18:10 | |
-!- gsomix [~gsomix@85.26.233.247] has joined #shogun | 18:10 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 18:21 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has joined #shogun | 18:29 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has quit [Changing host] | 18:29 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 18:29 | |
-!- blackburn [~qdrgsm@83.234.54.186] has quit [Ping timeout: 252 seconds] | 18:42 | |
-!- harshit_ [~harshit@182.68.254.100] has joined #shogun | 18:59 | |
harshit_ | sonney2k: around ? | 19:03 |
-!- blackburn [~qdrgsm@83.234.54.186] has joined #shogun | 19:45 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 20:06 | |
@sonney2k | blackburn, I just learned that scikits-learn got 3 proposals | 20:06 |
n4nd0 | I am taking a look to the seg fault someone pointed out in the mailing list if you are not with it guys | 20:07 |
@sonney2k | harshit_, now yes | 20:07 |
n4nd0 | sonney2k: they just got three proposals!? | 20:07 |
@sonney2k | yes I guess these 3 are of high quality thouh! | 20:07 |
@sonney2k | though | 20:07 |
@sonney2k | so everyone will be in | 20:07 |
n4nd0 | yeah sure ... but I don't know | 20:08 |
@sonney2k | I would say we have 10-15 HQ students - I am still not through the whole list | 20:08 |
n4nd0 | what does HQ stand for? high quality maybe? | 20:08 |
@sonney2k | yeah | 20:08 |
n4nd0 | ok | 20:08 |
CIA-64 | shogun: Soeren Sonnenburg master * r8b6902c / examples/undocumented/python_modular/classifier_qda_modular.py : | 20:09 |
CIA-64 | shogun: Merge pull request #429 from iglesias/qda | 20:09 |
CIA-64 | shogun: QDA for python_modular's testsuite - http://git.io/R3hi_Q | 20:09 |
@sonney2k | n4nd0, did you understand a bit better how the testsuite is supposed to work? | 20:10 |
n4nd0 | sonney2k: well, just a little bit | 20:10 |
n4nd0 | sonney2k: I still don't really understand how one is supposed to fix the tests that are wrong :S | 20:11 |
harshit_ | sonney2k: Need to know how CDotFeatures operates internally , | 20:11 |
harshit_ | As in I am implementing LBP features, so when user creates ClbpFeatures object by giving images and other argument, Then do I need to compute corresponding LBP feautes for those images there and then only ? | 20:11 |
harshit_ | or create LBP feature representation for one image when user request it . | 20:12 |
@sonney2k | n4nd0, the way I do it is run tester.py -d bad_test.py | 20:13 |
n4nd0 | sonney2k: but the output there is quite unreadable | 20:13 |
@sonney2k | then when things are not the same I have these two lists a and b to investigate | 20:13 |
n4nd0 | sonney2k: do you work with that kind of format? | 20:13 |
@sonney2k | well I can see what is different | 20:13 |
@sonney2k | if things are numpy arrays I can compute the deviation, e.g. numpy.max(numpy.abs(a[1]-b[1])) | 20:14 |
@sonney2k | or so | 20:14 |
@sonney2k | the only thing where it is difficult is SGObjects | 20:14 |
n4nd0 | sonney2k: oh ok | 20:14 |
@sonney2k | for these I do pickle.dumps(a[1]) etc | 20:14 |
@sonney2k | and then have a general llok | 20:14 |
@sonney2k | look | 20:14 |
@sonney2k | for SGObjects one can see though which arrays should be the same | 20:15 |
n4nd0 | sonney2k: let me show you just in case, http://snipt.org/uhoj1 | 20:16 |
@sonney2k | so for example if a[0] and b[0] are two objects derived from CSVM I can check if they are the same by computing the objective | 20:16 |
n4nd0 | sonney2k: that is the kind of output one should wait for? | 20:16 |
@sonney2k | well not the same actually - they might deviate slightly due to different numerical libs etc | 20:16 |
@sonney2k | so I mark this as ok and run generator.py badscript.py | 20:16 |
n4nd0 | I see | 20:17 |
@sonney2k | wait how do you get that output | 20:17 |
n4nd0 | ./tester.py -d classifier_larank_modular.py | 20:18 |
@sonney2k | (but answer is yes if you do p a[0] and a[0] is an sgobject) | 20:18 |
n4nd0 | I choose that file because tester.py detects it as ERROR | 20:18 |
@sonney2k | dammed some import pdb / pdb.set_trace() is missing | 20:19 |
n4nd0 | aham | 20:19 |
n4nd0 | that is probably why I didn't see this interactive menu you told me about | 20:20 |
n4nd0 | to do print a, print b and so on | 20:20 |
@sonney2k | looks like when I added these -m etc flags I did trash the pdb stuff | 20:20 |
harshit_ | sonney2k: In matlab example in libsvmocas for lbpfeatures, what is the format of image matrix, row major or column major | 20:21 |
@sonney2k | harshit_, things are already halve in shogun btw LBPPyrDotFeatures.h | 20:22 |
n4nd0 | harshit_: AFAIK images coordinates are normally (0,0) for the top left corner and (M,N) for the bottom right | 20:23 |
@sonney2k | harshit_, I am not sure - better ask vojtech about it (he is the author) | 20:23 |
@sonney2k | n4nd0, yeah but you never know ... | 20:23 |
n4nd0 | harshit_: do you know if the images represent something you can see? | 20:24 |
harshit_ | sonney2k: okay i'll ask vojtech about format of it .. but can you tell me how exactly you want lbppyrDotFeatures need to be rebuilt | 20:25 |
n4nd0 | harshit_: if so, you could try to represent them with imagesc | 20:25 |
harshit_ | n4nd0: they are in a mat file | 20:25 |
n4nd0 | harshit_: that could give you a clue about the coordinates | 20:25 |
harshit_ | 7400*500 | 20:25 |
n4nd0 | harshit_: I meant if they are images of, idk, of horses for example | 20:25 |
harshit_ | 500 is the no of features | 20:25 |
n4nd0 | harshit_: ok, so we are talking about feature vectors no raw images | 20:26 |
harshit_ | I dont know whether they are real images are not, have to ask vojtech | 20:26 |
CIA-64 | shogun: Soeren Sonnenburg master * r418a6b5 / testsuite/python_modular/tester.py : enter debug mode for tester when -d is specified - http://git.io/sKn36g | 20:26 |
@sonney2k | n4nd0, now have a look again | 20:26 |
@sonney2k | n4nd0, actually this particular svm you are testing is probabilistic... | 20:27 |
@sonney2k | so it is no wonder this test fails... | 20:27 |
n4nd0 | sonney2k: aham! that would be make it more difficult to get the same output ... | 20:28 |
n4nd0 | :) | 20:28 |
n4nd0 | yeah, now I get the pdb promt, cool | 20:28 |
@sonney2k | n4nd0, so now you could do p a[0].get_labels() | 20:28 |
@sonney2k | and see the values | 20:28 |
PhilTillet | Hello everybody :) | 20:29 |
n4nd0 | sonney2k: thank you .9 | 20:30 |
n4nd0 | ups | 20:30 |
n4nd0 | :) | 20:30 |
PhilTillet | .9 is a pretty neat smiley :p | 20:30 |
-!- gsomix [~gsomix@85.26.233.247] has quit [Quit: ????? ? ?? ??? (xchat 2.4.5 ??? ??????)] | 20:38 | |
-!- gsomix [~gsomix@85.26.233.247] has joined #shogun | 20:38 | |
blackburn | sonney2k: yeah I have seen that in twitter of someone from scikits team | 20:39 |
-!- blackburn [~qdrgsm@83.234.54.186] has quit [Quit: Leaving.] | 20:41 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 20:44 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Read error: Connection reset by peer] | 20:47 | |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 20:48 | |
@sonney2k | if we get only 3 slots we are in serious trouble | 20:49 |
@sonney2k | anyways we will know on april 11 | 20:49 |
blackburn | no way | 20:49 |
blackburn | april 11? | 20:49 |
blackburn | sonney2k: why they would give us less than last year? | 20:50 |
@sonney2k | april 11 end of workday / pacific time | 20:50 |
blackburn | wow they are pretty fast | 20:50 |
n4nd0 | sonney2k: will you let us know how many we get? | 20:50 |
@sonney2k | sure | 20:50 |
n4nd0 | :) | 20:50 |
blackburn | however there is a waiting list | 20:51 |
@sonney2k | it might be apr 12 though | 20:51 |
blackburn | that would become nervous if he didn't say that | 20:51 |
n4nd0 | I hope at least for 5 | 20:51 |
n4nd0 | you got 5 last year and everyone succeeded | 20:51 |
n4nd0 | I think it's fair | 20:51 |
@sonney2k | I can actually enter how many amazing proposals we have (the pure minimum of #slots we want) | 20:51 |
@sonney2k | and how many we would want if there were inf slots available | 20:52 |
blackburn | sonney2k: enter to? | 20:52 |
@sonney2k | problem is that I am not through all the proposals yet so it is hard to say | 20:52 |
@sonney2k | enter to melange | 20:52 |
blackburn | are there special fields for that? | 20:52 |
@sonney2k | yes | 20:52 |
gsomix | sonney2k, http://pastebin.com/3Ni2yX8p and http://pastebin.com/Wiumz3qv | 20:53 |
blackburn | sonney2k: and what do you think is a minimum? | 20:53 |
@sonney2k | independent of that each mentor can score students and there is some (hidden) discussion going on among mentors about students | 20:53 |
gsomix | hi all | 20:53 |
@sonney2k | blackburn, I don't know yet I would guess 7-9 | 20:54 |
blackburn | sonney2k: heh and maximum is 9? | 20:55 |
@sonney2k | lets be realistic | 20:55 |
blackburn | sonney2k: your opinion on that is changed ;) | 20:55 |
@sonney2k | we only have 10 mentors | 20:55 |
@sonney2k | blackburn, what do you mean? | 20:55 |
blackburn | sonney2k: yesterday you told me you think chances are we get 7-8 slots or so | 20:56 |
blackburn | and now you think about 3 even :) | 20:56 |
@sonney2k | and? | 20:56 |
@sonney2k | no I don't think about 3 | 20:56 |
blackburn | okay ;) | 20:57 |
@sonney2k | I am just saying 3 would be catastrophic | 20:57 |
blackburn | yes | 20:57 |
blackburn | 4,5 as well | 20:57 |
@sonney2k | then it is like rolling a dice to chose one of the very good students | 20:57 |
@sonney2k | anyway since they will announce on apr 11 - I am under high pressure to actually get mentors to do sth over easter vacations | 20:58 |
@sonney2k | not easy - I was hoping we have the tuesday too | 20:58 |
blackburn | do you really need these rates? I mean you would have to choose ideas rather than students applying for the same idea | 20:59 |
n4nd0 | blackburn: btw did you see mailing list? pluskid suggested about using smart pointers as we talked about | 21:01 |
blackburn | n4nd0: yes | 21:01 |
n4nd0 | sonney2k: what do you think about it? | 21:01 |
blackburn | I have no opinion/expertise on that | 21:01 |
n4nd0 | blackburn: it's ok to be gay and use garbage collection :) | 21:02 |
blackburn | n4nd0: too gay | 21:02 |
blackburn | :D | 21:02 |
n4nd0 | blackburn: I'll never forget that man | 21:02 |
n4nd0 | lol | 21:02 |
blackburn | n4nd0: that video? | 21:02 |
blackburn | have you seen azis btw? | 21:02 |
n4nd0 | blackburn: both the video and your answer :D azis? | 21:02 |
blackburn | oh you haven't seen that | 21:03 |
blackburn | ok a min | 21:03 |
blackburn | http://www.youtube.com/watch?v=F4hQ4J4BFOM&feature=player_detailpage | 21:03 |
blackburn | that's probably the most gay video in the universe | 21:03 |
gsomix | azis *__* | 21:03 |
n4nd0 | talking about see, since we are in easter I think we all deserve a rest and watch star wars | 21:03 |
n4nd0 | telnet towel.blinkenlights.nl | 21:03 |
blackburn | ah yeah I remember that | 21:04 |
n4nd0 | blackburn: that video looks pretty hardcore, not a good moment to see it in the living room with other people .P | 21:04 |
blackburn | n4nd0: hahahah lol | 21:04 |
n4nd0 | wtf is wrong with my keyboard today | 21:04 |
n4nd0 | :P | 21:04 |
n4nd0 | that's better | 21:05 |
blackburn | probably not the best video to show gf's family | 21:05 |
blackburn | oh lolz | 21:05 |
blackburn | I imagine how could they react | 21:05 |
gsomix | blackburn, have you seen the clip `Hop'? :3 | 21:06 |
blackburn | is it as gay as azis is? | 21:08 |
blackburn | gsomix: no, I can't remember that | 21:08 |
blackburn | I prefer hardcore gay video | 21:08 |
PhilTillet | blackburn, like justin bieber? | 21:09 |
blackburn | PhilTillet: exactly I am a belieber lol | 21:09 |
PhilTillet | lol | 21:09 |
PhilTillet | omg Azis. | 21:10 |
PhilTillet | I did not know about him. | 21:10 |
blackburn | gsomix: while moving upstairs 15 minutes before I was thinking how to overload you a little more | 21:10 |
gsomix | blackburn, and..? | 21:11 |
blackburn | gsomix: do you think it would be interesting for you to get shogun well-known in the .ru zone? | 21:12 |
blackburn | PhilTillet: lucky | 21:12 |
PhilTillet | would be great if one day one gay commit had a sha1 starting with F4hQ4J4BFOM | 21:13 |
blackburn | probably it is the task for me but it would be useful for you | 21:13 |
gsomix | blackburn, what do you mean? site? advertising? | 21:13 |
blackburn | gsomix: e.g. there is no russian page on Shogun in wiki | 21:14 |
blackburn | however python3 stuff and other are in prior | 21:14 |
blackburn | gsomix: habrahabr would work as well :) | 21:15 |
gsomix | blackburn, hmm. I can translate article from English. | 21:16 |
@sonney2k | google translate can too :D | 21:16 |
blackburn | yeah | 21:16 |
blackburn | ok let me describe my intention | 21:16 |
blackburn | but I am already used to some writing | 21:16 |
CIA-64 | shogun: Soeren Sonnenburg master * rc835442 / (2 files): | 21:17 |
CIA-64 | shogun: convert the float32 based dot function to return float32 - no need for | 21:17 |
CIA-64 | shogun: float64_t inflasion - http://git.io/NVPVzA | 21:17 |
blackburn | I can do that for sure | 21:17 |
blackburn | but you are not yet so it would be useful | 21:17 |
blackburn | sonney2k: that function has been popular last days | 21:17 |
* PhilTillet feels embarassed | 21:18 | |
blackburn | gsomix: so what do you think? | 21:18 |
gsomix | sonney2k, Hulk sad. However, I translate better than I write. :) | 21:19 |
blackburn | gsomix: the more you do things the better you do it | 21:19 |
blackburn | year ago sonney2k usually had a headache while understanding my english | 21:20 |
gsomix | blackburn, can we discuss details in private? | 21:21 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Quit: Page closed] | 21:21 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 21:22 | |
gsomix | hmm | 21:22 |
blackburn | bad connection | 21:22 |
blackburn | sonney2k: I think we need to establish that # referring in github commits | 21:26 |
blackburn | however risk is that we could forgot to refer issues too often | 21:27 |
blackburn | one way to avoid this is to restrict commits w/o # | 21:28 |
blackburn | i.e. any commit should go to some issue | 21:28 |
blackburn | n4nd0: what do you think? | 21:28 |
n4nd0 | blackburn: do you mean to use symbols for the commits? | 21:29 |
blackburn | n4nd0: no, e.g. all doc things should go to #408 | 21:30 |
blackburn | once you add #408 to the commit msg | 21:30 |
blackburn | https://github.com/shogun-toolbox/shogun/issues/418 | 21:30 |
blackburn | #418 sorry | 21:30 |
blackburn | it goes to the issue | 21:30 |
n4nd0 | blackburn: ok, but later the commits can be added as usual, using PRs? | 21:31 |
blackburn | hmm | 21:31 |
blackburn | however.. | 21:32 |
blackburn | n4nd0: could you please fix some doc and commit with #418? | 21:32 |
blackburn | probably it is only for me, Soeren and heiko commits | 21:32 |
n4nd0 | I should go something like | 21:32 |
blackburn | I am curious whether it will appear in both PR and issue | 21:32 |
n4nd0 | fix doc #418 right? | 21:32 |
blackburn | yeah something blablba #418 | 21:33 |
n4nd0 | blackburn: done, in my branch qda | 21:36 |
n4nd0 | blackburn: yeah, it looks like it is included in Documentation improvements | 21:38 |
n4nd0 | I think that is cool | 21:38 |
blackburn | that's nice | 21:39 |
n4nd0 | but anyway | 21:39 |
blackburn | sonney2k: https://github.com/shogun-toolbox/shogun/issues/418 | 21:39 |
n4nd0 | how does this work to merge later? | 21:39 |
blackburn | it does nothing with merge/etc | 21:39 |
blackburn | just groups commits | 21:39 |
n4nd0 | ok | 21:39 |
n4nd0 | jo the method is the same | 21:39 |
n4nd0 | do pull requests apart | 21:40 |
blackburn | yes | 21:40 |
n4nd0 | I like it | 21:40 |
blackburn | me too | 21:41 |
blackburn | I will suggest it using the ML | 21:41 |
-!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has left #shogun ["Leaving"] | 21:42 | |
-!- PhilTillet [~Philippe@vir78-1-82-232-38-145.fbx.proxad.net] has joined #shogun | 21:42 | |
shogun-buildbot | build #446 of ruby_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/446 blamelist: fernando.iglesiasg@gmail.com | 21:43 |
blackburn | gsomix: already? nice work | 21:44 |
gsomix | blackburn, http://pastebin.com/Wiumz3qv | 21:45 |
blackburn | gsomix: is it a proof that it works? | 21:47 |
gsomix | blackburn, yep. | 21:47 |
blackburn | I believe it works ;) | 21:51 |
blackburn | gsomix: so you need next task? | 21:53 |
gsomix | blackburn, yep. | 21:53 |
blackburn | hmm | 21:54 |
blackburn | gsomix: you prefer hardcore tasks, right? ;) | 21:56 |
gsomix | blackburn, yep. for example, a researching of this: http://xkcd.com/323/ | 21:57 |
gsomix | :] | 21:57 |
blackburn | I've never heard of your alco addiction ;) | 21:57 |
gsomix | blackburn, hehe. it's a joke. | 21:58 |
blackburn | mine addict is not ;( | 21:58 |
blackburn | gsomix: I know that custom kernel is wrong | 22:01 |
blackburn | exact problem is in triangular from full | 22:02 |
blackburn | or vice versa | 22:02 |
PhilTillet | blackburn, I'm looking for some tasks too. Would it be a good idea to implement more Kernels (like most of the dot kernels) in my opencl pull request? | 22:11 |
blackburn | PhilTillet: I think not, do you mind doing something not opencl related? | 22:14 |
PhilTillet | I don't mind | 22:14 |
PhilTillet | I'd rather do something C++ related though | 22:14 |
blackburn | yeah sure | 22:14 |
harshit_ | blackburn: Where is RealFeatures used in octave implemented in libshogun ? | 22:15 |
harshit_ | cant find any thing similar in libshogun | 22:15 |
blackburn | harshit_: it is SimpleFeatures<float64_t> | 22:15 |
harshit_ | so if i have to refer SimpleFeatures<int32_t> then what do i need to use ? | 22:16 |
blackburn | IntFeatures | 22:16 |
harshit_ | Is there any file where all these conversions are listed ? | 22:17 |
blackburn | yes | 22:17 |
blackburn | interfaces/modular/Features.i | 22:17 |
blackburn | line 276 | 22:17 |
harshit_ | okay thanks :) | 22:17 |
harshit_ | dam how can you remember line no too !!! | 22:17 |
harshit_ | you really are into shogun, must say :) | 22:18 |
blackburn | I took a look | 22:18 |
harshit_ | oh | 22:18 |
blackburn | no, I can't remember that stuff too :) | 22:18 |
harshit_ | but still you know alot about it :) | 22:19 |
blackburn | I have to heh | 22:20 |
PhilTillet | so, blackburn, any idea? :p | 22:21 |
blackburn | PhilTillet: yeah thinking | 22:23 |
PhilTillet | i couldn't find any size function for EFeatureType | 22:23 |
blackburn | any preference? | 22:23 |
PhilTillet | hmmm | 22:23 |
PhilTillet | not really | 22:23 |
PhilTillet | your question was kind of broad :p | 22:24 |
PhilTillet | I can work on design issues, safety or some algorithm | 22:24 |
PhilTillet | or implement garbage collection everywhere (who said troll? :p) | 22:25 |
blackburn | knn regression? | 22:25 |
PhilTillet | why not, i was more used to knn classification but it's ok I guess | 22:26 |
blackburn | nearest centroid | 22:26 |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has quit [Quit: leaving] | 22:26 | |
PhilTillet | hmm | 22:27 |
PhilTillet | both seem fine | 22:27 |
shogun-buildbot | build #202 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/202 | 22:29 |
harshit_ | blackburn: I can not use UIntFeatures, octave gives an error: `UIntFeatures' undefined near line 8 column 13 | 22:30 |
harshit_ | but IntFeatures works fine | 22:30 |
PhilTillet | blackburn, so let's say nearest centroid? | 22:32 |
blackburn | PhilTillet: sure | 22:33 |
@sonney2k | blackburn, I know you can trivially fix this issue http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/202/steps/compile/logs/stdio | 22:33 |
@sonney2k | so please do :) | 22:33 |
blackburn | sonney2k: git is currently unavailable for me | 22:33 |
PhilTillet | guess I got to find inspiration on knn :) | 22:33 |
blackburn | public access point and 22 is blocked | 22:33 |
@sonney2k | blackburn, what happened? | 22:33 |
@sonney2k | blackburn, then tell me what I have to do to fix this! | 22:34 |
blackburn | hah | 22:34 |
blackburn | just add KNN_COVERTREE_POINT to ignores | 22:34 |
PhilTillet | haha | 22:35 |
PhilTillet | metaprogramming errors | 22:35 |
PhilTillet | instantiated from ... | 22:35 |
-!- harshit_ [~harshit@182.68.254.100] has quit [Quit: Leaving] | 22:36 | |
@sonney2k | blackburn, why does this error not occur w/ eg ISOMAP_COVERTREE_POINT | 22:37 |
blackburn | sonney2k: isn't it ignored? | 22:37 |
blackburn | sonney2k: the only difference is DOXYGEN_SHOULD_SKIP_THIS :D | 22:38 |
@sonney2k | gsomix, why did you mark your patch work in progress? | 22:39 |
@sonney2k | blackburn, that cannot be it | 22:39 |
blackburn | ye | 22:39 |
blackburn | s | 22:39 |
gsomix | sonney2k, a habit. | 22:40 |
blackburn | sonney2k: how does swig select classes to include? | 22:40 |
@sonney2k | blackburn, ahh I know - you define the class in the .cpp file, n4nd0 in .h | 22:40 |
blackburn | yes | 22:41 |
shogun-buildbot | build #201 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/201 | 22:41 |
blackburn | I just came over it as well | 22:41 |
@sonney2k | gsomix, did you do a test? I mean set the data and get it and it matched? | 22:41 |
blackburn | sonney2k: there are two ways usually - SERIALIZABLE_DUMMY and ignoe | 22:41 |
gsomix | sonney2k, http://pastebin.com/3Ni2yX8p and http://pastebin.com/Wiumz3qv | 22:41 |
@sonney2k | blackburn, so which one will work ? | 22:42 |
blackburn | sonney2k: any but no need to have covertree point in the interface so just ignore | 22:42 |
@sonney2k | gsomix, ok then great work - I don't understand why you mark this WIP then - it will not become any better than that. | 22:43 |
@sonney2k | blackburn, can a covertree be serialized? | 22:44 |
blackburn | sonney2k: currently no | 22:44 |
blackburn | sonney2k: it is full of std stuff | 22:45 |
gsomix | sonney2k, I removed the WIP status. :) | 22:45 |
@sonney2k | argh then this won't work SG_ADD((CSGObject**) &m_covertree, "m_covertree", "Member cover tree", MS_NOT_AVAILABLE); | 22:45 |
blackburn | yes | 22:45 |
@sonney2k | too bad | 22:46 |
blackburn | sonney2k: mainly because covertree is not a descendant of CSGObject :D | 22:46 |
@sonney2k | not only | 22:46 |
@sonney2k | alright we have to bug n4ndo about this | 22:47 |
blackburn | sonney2k: I am not really happy with that covertree | 22:47 |
@sonney2k | what's up with JL's covertree? | 22:47 |
blackburn | sonney2k: it is hard to integrate there | 22:47 |
blackburn | both I and gsomix tried | 22:47 |
blackburn | and have died with a lot of leaks/bugs while integrating | 22:48 |
@sonney2k | why? | 22:48 |
blackburn | sonney2k: hard question | 22:49 |
gsomix | not hard. just somewhere I'm doing wrong. I need to be careful. | 22:49 |
@sonney2k | hmmhh - I expect JLs code to be fast at least | 22:49 |
blackburn | yeah | 22:49 |
blackburn | yes it is faster | 22:49 |
blackburn | so we would need to make another attempt | 22:49 |
@sonney2k | so it would be really great to get it | 22:49 |
@sonney2k | how long have you tried? | 22:49 |
blackburn | sonney2k: yes I agree | 22:49 |
@sonney2k | and serialization should be simpler too | 22:49 |
blackburn | I spent a week or so on that | 22:49 |
@sonney2k | because JL hates std:: crap | 22:49 |
blackburn | gsomix have spent a month or two | 22:50 |
blackburn | :D | 22:50 |
blackburn | but not as active as me probably | 22:50 |
gsomix | :( | 22:50 |
gsomix | sonney2k, blackburn: I plan to return to work on covertree again. | 22:50 |
@sonney2k | blackburn, gsomix has written the octave sparse matrix typemap - that would have taken you a year :D | 22:50 |
blackburn | sonney2k: I do not really understand why we both failed | 22:51 |
blackburn | sonney2k: I heard it was easy ;) | 22:51 |
@sonney2k | actually :P | 22:51 |
@sonney2k | everything is easy if you know how | 22:51 |
blackburn | in the arpack.cpp | 22:52 |
blackburn | I guess everybody is afraid to enter to that file | 22:52 |
blackburn | sonney2k: I already did some sparse stuff you know | 22:52 |
blackburn | sonney2k: I still do not understand why you hate std :) | 22:53 |
blackburn | shogun/lib/ShogunException.cpp:29 is a funny line | 22:55 |
@sonney2k | I saw this even! | 22:57 |
blackburn | sonney2k: really? did you run shogun on some controller? | 22:57 |
@sonney2k | no a 92G machine | 22:58 |
@sonney2k | you know big data ;D | 22:58 |
blackburn | sonney2k: 92G and no place to allocate a string? lol | 22:59 |
CIA-64 | shogun: Soeren Sonnenburg master * r908c468 / (2 files in 2 dirs): | 23:00 |
CIA-64 | shogun: fix compilation of java_modular | 23:00 |
CIA-64 | shogun: - disable covertree serialization in KNN | 23:00 |
CIA-64 | shogun: - put KNN POINT type into modshogun ignores - http://git.io/yLNpkA | 23:00 |
CIA-64 | shogun: Soeren Sonnenburg master * r46353c3 / (2 files in 2 dirs): | 23:00 |
CIA-64 | shogun: Merge pull request #430 from gsomix/sparse_octave | 23:00 |
CIA-64 | shogun: Sparse matrix typemap for octave modular. - http://git.io/qRzsSQ | 23:00 |
-!- n4nd0 [~nando@h121n3-ksg-a11.ias.bredband.telia.com] has joined #shogun | 23:00 | |
n4nd0 | sonney2k, blackburn I checked the logs | 23:06 |
blackburn | yes | 23:06 |
n4nd0 | sonney2k: I had to define covertree_point in .h because of the covertree type member | 23:06 |
@sonney2k | n4nd0, yeah I know | 23:07 |
@sonney2k | (read your code) | 23:07 |
n4nd0 | ok :) | 23:07 |
n4nd0 | has that turned out to be an issue? | 23:07 |
n4nd0 | sorry for the non-required explanation :S | 23:07 |
@sonney2k | well it didn't compile and would crash when being serialized | 23:08 |
n4nd0 | not good | 23:10 |
@sonney2k | n4nd0, yeah but not so easy to see | 23:10 |
@sonney2k | n4nd0, btw how much faster is it on some realistic data set? | 23:10 |
shogun-buildbot | build #447 of ruby_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/447 | 23:11 |
blackburn | sonney2k: n4nd0: I'll let you know of speedup on GTSRB dataset (39029 vectors of 900d) | 23:11 |
n4nd0 | sonney2k: I proposed it yesterday to make a test on iris dataset to someone who was looking for a patch | 23:12 |
n4nd0 | blackburn: ok, great | 23:12 |
blackburn | iris? | 23:12 |
blackburn | hah | 23:12 |
n4nd0 | yeah, why? | 23:12 |
blackburn | I am afraid to be wrong but isn't it <1K of vectors? | 23:13 |
n4nd0 | 150 instances is said on the webpage | 23:13 |
n4nd0 | mmm yeah, maybe that is too little | 23:14 |
blackburn | yes it will be even slower may be | 23:14 |
n4nd0 | in octave example, super small, it was faster though | 23:14 |
blackburn | slightly overheaded with tree :) | 23:14 |
blackburn | hmm | 23:14 |
blackburn | ok then | 23:14 |
n4nd0 | blackburn: do you have your own example to try it on GTSRB? somewhere in shogun? | 23:15 |
@sonney2k | gsomix, do you want to continue to work on sth? | 23:15 |
blackburn | n4nd0: no | 23:16 |
gsomix | sonney2k, of course. | 23:17 |
blackburn | n4nd0: I can share feature matrices and labels if you want | 23:22 |
n4nd0 | blackburn: yeah, sure | 23:23 |
n4nd0 | blackburn: how should we do it? | 23:23 |
blackburn | http://dl.dropbox.com/u/10139213/s/test_labels_prewitt.npy | 23:23 |
blackburn | http://dl.dropbox.com/u/10139213/s/test_matrix_prewitt.npy | 23:23 |
blackburn | http://dl.dropbox.com/u/10139213/s/train_labels_prewitt.npy | 23:24 |
blackburn | http://dl.dropbox.com/u/10139213/s/train_matrix_prewitt.npy | 23:24 |
blackburn | n4nd0: just load it from numpy and use | 23:24 |
n4nd0 | blackburn: ok, thank you | 23:25 |
n4nd0 | wow, 174MB for one of them | 23:25 |
blackburn | n4nd0: I don't remember dimensionality exactly but there are 39K of train vectors | 23:25 |
blackburn | and 12K of test ones | 23:25 |
n4nd0 | blackburn: have you ever trained sth with this? | 23:26 |
blackburn | n4nd0: yes multiclass svm | 23:26 |
n4nd0 | blackburn: how long did it take? | 23:27 |
blackburn | kernel svm takes 3h or so | 23:27 |
blackburn | GMNP essentially | 23:27 |
blackburn | liblinear took a little | 23:27 |
blackburn | 5-10 minutes | 23:27 |
@sonney2k | So you could do either some ASSERT()'s replacement with meaningful SG_ERROR() messages or start some big task - SGVector | 23:28 |
blackburn | gsomix: it is for you | 23:28 |
blackburn | ^ | 23:28 |
blackburn | :) | 23:28 |
@sonney2k | for SGVector you could either implement refcounting business and/or replace Array.h with SGVector | 23:29 |
@sonney2k | the latter would boil down to adding the functions from Array to SGVector and then replace | 23:30 |
@sonney2k | same fro Array2 -> SGMatrix, etc | 23:30 |
blackburn | n4nd0: if you want to dig into details: features have been constructed using HOG w/o block normalization and prewitt gradient | 23:31 |
blackburn | some median filtering as well however it is really too detailed then | 23:32 |
blackburn | :) | 23:32 |
n4nd0 | blackburn: are the same you use for the paper you showed me? | 23:34 |
blackburn | n4nd0: bingo | 23:34 |
gsomix | sonney2k, do I have the time to think? :) | 23:38 |
blackburn | gsomix: no you need to implement all the things *tonight* | 23:39 |
blackburn | no kidding | 23:39 |
blackburn | ;) | 23:39 |
gsomix | blackburn, ?????, ???. :] | 23:39 |
gsomix | sorry for my Russian | 23:40 |
@sonney2k | gsomix, well you can do other things to - just propose sth | 23:40 |
gsomix | sonney2k, ok. now I'm working on custom_kernel. | 23:46 |
blackburn | sonney2k: btw one more weka extension was published on mloss | 23:50 |
@sonney2k | gsomix, what do you do there? | 23:51 |
blackburn | a kind of | 23:51 |
blackburn | sonney2k: there is a bug still ;) | 23:51 |
@sonney2k | blackburn, what are you talking about? | 23:52 |
blackburn | sonney2k: some set_triangular_.... is wrong | 23:52 |
blackburn | gsomix should know details now | 23:52 |
blackburn | sonney2k: old tester fails on custom kernel | 23:52 |
blackburn | because of that bug | 23:53 |
@sonney2k | ok | 23:53 |
--- Log closed Sun Apr 08 00:00:19 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!