--- Log opened Mon Mar 26 00:00:19 2012 | ||
makokal | hey are they any mentors online? | 00:06 |
---|---|---|
makokal | esp for GPs | 00:07 |
blackburn | I am afraid GPs mentor could be reached only by email | 00:07 |
blackburn | have never seen him there | 00:07 |
makokal | kk, mail on the way | 00:08 |
blackburn | makokal: I can't be considered as mentor but I am responsible for any development | 00:09 |
makokal | oh, I just want to ask about clarification on imeplementing GPs for gsoc this year | 00:09 |
blackburn | yeah ask | 00:10 |
makokal | on the desc, on regression is mentioned, I was wondering if using GPs for classification will be part of the project | 00:11 |
makokal | for dev it should not mean much, but just a clarification | 00:12 |
makokal | as I am working on finalizing my document for applying for this | 00:12 |
blackburn | it is up to you generally | 00:13 |
blackburn | I mean if you want to do both regression and classification - it would be better for sure | 00:13 |
makokal | ok, I will just add it on the final bits(so depending on the progress, add it or not) | 00:14 |
blackburn | but this should be clarified with Oliver as well | 00:14 |
makokal | ya am mailing him at the moment | 00:14 |
blackburn | and consider to come with small patch! it is a requirement | 00:15 |
makokal | sure | 00:16 |
makokal | thanks | 00:18 |
-!- makokal [~makokal@brmn-4d0acdfe.pool.mediaWays.net] has left #shogun [] | 00:49 | |
-!- blackburn [~qdrgsm@188.168.5.14] has quit [Quit: Leaving.] | 01:09 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has quit [Quit: wiking_] | 01:40 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds] | 01:48 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 02:27 | |
-!- PhilTillet [~Philippe@38.124.160.48] has quit [Remote host closed the connection] | 02:39 | |
-!- vikram360 [~vikram360@117.192.165.141] has quit [Ping timeout: 252 seconds] | 04:24 | |
-!- gsomix [~gsomix@188.168.2.197] has joined #shogun | 05:23 | |
-!- gsomix [~gsomix@188.168.2.197] has quit [Read error: Operation timed out] | 06:11 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 07:12 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has joined #shogun | 07:32 | |
-!- wiking [~wiking@78-23-191-201.access.telenet.be] has quit [Changing host] | 07:32 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 07:32 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 08:00 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:01 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 264 seconds] | 08:05 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:11 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 08:18 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 08:34 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:39 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit] | 08:40 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 08:50 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 09:03 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 09:13 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 09:26 | |
blackburn | monday is a hard day lol | 09:31 |
n4nd0 | blackburn: feeling sleepy? :P | 09:36 |
blackburn | n4nd0: strange but not at all - funny thing is only 9 here are idling | 09:37 |
blackburn | but before it was ~20 :) | 09:37 |
n4nd0 | that's actually right, funny | 09:37 |
blackburn | n4nd0: what time is it there you are? | 09:38 |
blackburn | I am curious whether you practice to sleep sometimes :D | 09:38 |
n4nd0 | blackburn: haha I do, I do | 09:39 |
n4nd0 | but not that much actually, I've always been ok not sleeping that much | 09:39 |
n4nd0 | it is 9:40 | 09:39 |
blackburn | for me it is ok as it is 11.39 here | 09:39 |
blackburn | but it is 8.39 in Sweden? | 09:40 |
n4nd0 | mmm no | 09:40 |
blackburn | 9? :) | 09:40 |
n4nd0 | it is 9:40 in Sweden | 09:40 |
n4nd0 | yesterday we changed the time in here, summer time! :) | 09:40 |
blackburn | ah yeah | 09:40 |
n4nd0 | did you change it too there? | 09:41 |
blackburn | thing our gov't decided to abandon | 09:41 |
blackburn | no, we have static const time now | 09:41 |
blackburn | :D | 09:41 |
-!- harshit_ [~harshit@120.56.229.25] has joined #shogun | 09:41 | |
harshit_ | hey blackburn | 09:41 |
blackburn | hi | 09:41 |
harshit_ | did you hav a look at my pull request ? | 09:41 |
harshit_ | NewtonSVM | 09:42 |
blackburn | not really | 09:42 |
blackburn | I think Soeren should take a look as well | 09:42 |
harshit_ | okay .. So what do you think how long will it take to merge into shogun | 09:43 |
harshit_ | Also wanted to discuss Deep learning idea ! | 09:44 |
blackburn | I will check the code again a little bit later | 09:46 |
harshit_ | okay, with whom do you think i should discuss my deep learning idea ! | 09:47 |
harshit_ | soren ? | 09:47 |
blackburn | yeah probably | 09:48 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 10:09 | |
-!- harshit_ [~harshit@120.56.229.25] has quit [Ping timeout: 276 seconds] | 10:15 | |
-!- harshit_ [~harshit@120.56.229.25] has joined #shogun | 10:16 | |
-!- harshit_ [~harshit@120.56.229.25] has quit [Ping timeout: 244 seconds] | 10:23 | |
CIA-64 | shogun: Soeren Sonnenburg master * r001c17c / (8 files in 5 dirs): | 10:30 |
CIA-64 | shogun: Merge pull request #384 from gsomix/python3_interface | 10:30 |
CIA-64 | shogun: Partial support for python3 interface - http://git.io/Unk3vw | 10:30 |
blackburn | sonney2k: but he reported a problem on strings? | 10:31 |
CIA-64 | shogun: Soeren Sonnenburg master * r97e8e21 / (13 files in 7 dirs): | 10:31 |
CIA-64 | shogun: Merge pull request #386 from iglesias/qda | 10:31 |
CIA-64 | shogun: QDA - http://git.io/zwE1FQ | 10:31 |
blackburn | merge machine | 10:32 |
n4nd0 | oh QDA :) | 10:34 |
@sonney2k | blackburn, btw I tried this director stuff | 10:47 |
@sonney2k | got only crashes though | 10:47 |
blackburn | sonney2k: was it working? I am very curious with it | 10:47 |
blackburn | bad bad | 10:47 |
@sonney2k | I got some infinite recursions | 10:47 |
blackburn | well lets extend SWIG then | 10:47 |
blackburn | :D | 10:47 |
@sonney2k | i will add this class nevertheless | 10:47 |
@sonney2k | but ifdef'd | 10:47 |
blackburn | I see | 10:48 |
-!- pluskid [~chatzilla@60.191.2.238] has joined #shogun | 10:51 | |
shogun-buildbot | build #549 of python_static is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/549 blamelist: gsomix@gmail.com | 10:56 |
shogun-buildbot | build #620 of libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/620 blamelist: sonne@debian.org, fernando.iglesiasg@gmail.com | 10:58 |
n4nd0 | blackburn: why does shogun-buildbot usually fail to compile the first time? | 10:58 |
blackburn | n4nd0: he doesn't actually | 10:58 |
blackburn | this time he failed because of erroneous code :) | 10:58 |
n4nd0 | blackburn: really? our fault then? | 10:59 |
blackburn | n4nd0: it seems you forgot to import cblas | 10:59 |
blackburn | http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/620/steps/compile/logs/stdio | 10:59 |
n4nd0 | blackburn: mmm why does it get to compile without problems in my machine then? | 11:00 |
@sonney2k | well it build *every* commit | 11:00 |
@sonney2k | so if someone commits something crappy | 11:00 |
blackburn | n4nd0: different configuration | 11:00 |
blackburn | that is usual ^ | 11:00 |
blackburn | :D | 11:00 |
@sonney2k | and fixes it afterwards it will first fail and then succeed | 11:00 |
blackburn | sonney2k: it seems that python stuff should be elaborated more | 11:00 |
@sonney2k | if people were using git amend to merge fixes together this wouldn't happen - but hey I don't complain | 11:01 |
blackburn | n4nd0: have you secured QDA with HAVE_LAPACK? | 11:01 |
@sonney2k | yes | 11:01 |
n4nd0 | yes | 11:01 |
blackburn | n4nd0: no include for lapack.h | 11:02 |
blackburn | probably that is the reason | 11:02 |
n4nd0 | I see | 11:02 |
CIA-64 | shogun: Soeren Sonnenburg master * rdfbb2fd / (9 files in 3 dirs): add support for director kernel - http://git.io/GicbVQ | 11:03 |
blackburn | sonney2k: I think if we manage to add these directors it should be an option | 11:06 |
blackburn | --enable-directors or so | 11:07 |
pluskid | Do you have procedure to evaluate the performance of clustering result in shogun? | 11:07 |
blackburn | sonney2k: ah yes I did implement a proof-of-concept for libedrt | 11:07 |
blackburn | works even faster | 11:07 |
blackburn | and totally flexible | 11:07 |
blackburn | pluskid: no you are welcome to add any to shogun/evaluation ;) | 11:08 |
@sonney2k | blackburn, yeah - it is currently on but I certainly will add this ifdef stuff | 11:08 |
pluskid | blackburn: OK, I'll do that for my GSoC sample patch | 11:08 |
pluskid | BTW: I get a much much more simple patch | 11:09 |
pluskid | in your example code | 11:09 |
pluskid | for the matlab and octave module | 11:09 |
pluskid | the content in clustering_hierarchical.m is for kmeans clustering | 11:10 |
pluskid | while the content for clustering_kmeans.m is for hierarchical clustering | 11:10 |
pluskid | It should be a simple renaming | 11:10 |
pluskid | but I think this should be fixed | 11:11 |
pluskid | :p | 11:11 |
blackburn | feel free to do that using pull requests | 11:11 |
pluskid | sure :) | 11:11 |
blackburn | haha we have one task that would scary any student from participating in GSoC | 11:14 |
blackburn | https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/distributions/HMM.cpp there is a race condition somewhere LOL | 11:14 |
n4nd0 | blackburn: oh, any idea to reproduce it? | 11:15 |
blackburn | n4nd0: oh better do not do that | 11:16 |
shogun-buildbot | build #550 of python_static is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/550 | 11:16 |
blackburn | :D | 11:16 |
blackburn | hooray | 11:16 |
n4nd0 | blackburn: haha already tried? | 11:16 |
blackburn | hmm | 11:16 |
blackburn | n4nd0: well easy to reproduce but 5K source lines there | 11:17 |
blackburn | we figured it out and nothing mor | 11:17 |
blackburn | results with 2 threads are totally different | 11:17 |
blackburn | than with 1 | 11:17 |
n4nd0 | it could be useful anyway to take a look to that code for SO | 11:19 |
shogun-buildbot | build #390 of java_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/390 blamelist: gsomix@gmail.com | 11:23 |
blackburn | didn't I fix it?? | 11:23 |
-!- romi_ [~mizobe@187.101.141.174] has quit [Ping timeout: 246 seconds] | 11:25 | |
-!- pluskid [~chatzilla@60.191.2.238] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120312181643]] | 11:29 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 11:36 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 11:37 | |
-!- romi_ [~mizobe@187.57.4.159] has joined #shogun | 11:37 | |
-!- gsomix [~gsomix@188.168.128.179] has joined #shogun | 11:48 | |
blackburn | gsomix: notwithstanding that you've got some troubles with strings your patch is in | 11:51 |
blackburn | gsomix: what is the exact problem with strings here? | 11:51 |
n4nd0 | blackburn: should I fix something in QDA then? this include lapack? | 12:04 |
-!- romi_ [~mizobe@187.57.4.159] has quit [Ping timeout: 246 seconds] | 12:04 | |
-!- romi_ [~mizobe@187.57.4.159] has joined #shogun | 12:04 | |
shogun-buildbot | build #391 of java_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/391 | 12:13 |
blackburn | n4nd0: yes | 12:17 |
n4nd0 | n4nd0: any to configure locally like in buildbot so I'll be sure what compiles here will do it there? | 12:18 |
n4nd0 | blackburn: or any way to test it at least | 12:18 |
blackburn | should be a way, but I don't know exactly | 12:19 |
n4nd0 | mmm ok | 12:19 |
n4nd0 | so include lapack.h in QDA.h should get rid of that | 12:19 |
blackburn | if my guess is right | 12:20 |
gsomix | hi | 12:31 |
gsomix | blackburn, http://pastebin.com/dBb5bb8F | 12:31 |
blackburn | gsomix: I see.. | 12:32 |
gsomix | blackburn, I think problem is in the swig_typemaps file. I didn't realized something. | 12:34 |
n4nd0 | blackburn: I'll open now a pull request with it, can you try to merge it? apart from that I've added do_free member to SGNDArray similar to the one in SGMatrix and SGVector | 12:35 |
blackburn | n4nd0: I'll be able to merge a little bit later | 12:36 |
n4nd0 | blackburn: perfect ;) | 12:37 |
blackburn | actually when I get to my notebook :) | 12:37 |
gsomix | blackburn, O_o | 12:50 |
blackburn | gsomix: what's up? | 12:52 |
gsomix | blackburn, it's work, but... | 12:52 |
gsomix | however, for some reason installing libraries is not working properly. | 12:52 |
blackburn | what kind of libraries? | 12:53 |
gsomix | blackburn, _modshogun. | 12:53 |
blackburn | what is the error then? | 12:53 |
blackburn | it is pretty clear that typemaps get broken with python3 | 12:54 |
blackburn | while string list is not being recognized | 12:56 |
gsomix | blackburn, it does not matter. I just processed incorrectly new unicode strings. | 12:56 |
blackburn | I can suggest you to take a look at static int is_pystring_list(PyObject* obj, int typecode) | 12:56 |
blackburn | swig_typemaps.i:192 | 12:56 |
gsomix | blackburn, even last night, as it turned out. | 12:56 |
gsomix | blackburn, nevermind | 12:56 |
blackburn | I am pretty sure problem is related to such detection | 12:57 |
gsomix | I just could not check, because libraries are not installed. | 12:57 |
gsomix | but now everything is working | 12:57 |
blackburn | I don't get it - is it working now? | 12:58 |
gsomix | blackburn, there are not problems with strings. | 12:59 |
blackburn | is there any problem then? | 12:59 |
gsomix | a little | 13:00 |
-!- romi_ [~mizobe@187.57.4.159] has quit [Ping timeout: 246 seconds] | 13:00 | |
-!- romi_ [~mizobe@187.57.4.159] has joined #shogun | 13:00 | |
gsomix | blackburn, now running over 90 percent of the tests. | 13:02 |
-!- pluskid [~chatzilla@60.191.2.238] has joined #shogun | 13:06 | |
-!- pluskid_ [~chatzilla@li400-235.members.linode.com] has joined #shogun | 13:24 | |
-!- pluskid__ [~chatzilla@60.191.2.238] has joined #shogun | 13:25 | |
-!- pluskid [~chatzilla@60.191.2.238] has quit [Ping timeout: 252 seconds] | 13:25 | |
-!- pluskid__ is now known as pluskid | 13:25 | |
-!- pluskid_ [~chatzilla@li400-235.members.linode.com] has quit [Ping timeout: 264 seconds] | 13:29 | |
gsomix | sonney2k, hey | 13:30 |
CIA-64 | shogun: iglesias master * r388d734 / src/shogun/lib/DataType.h : + dofree member in SGNDArray - http://git.io/JJNxqA | 13:48 |
CIA-64 | shogun: iglesias master * r8d16f45 / src/shogun/classifier/QDA.cpp : + QDA include to avoid failure in buildbot and ~ ndarray destroy -> free - http://git.io/fj1iLw | 13:48 |
CIA-64 | shogun: Soeren Sonnenburg master * rdefc6b5 / (src/shogun/classifier/QDA.cpp src/shogun/lib/DataType.h): | 13:48 |
CIA-64 | shogun: Merge pull request #393 from iglesias/qda | 13:48 |
CIA-64 | shogun: QDA - http://git.io/0xTFDg | 13:48 |
CIA-64 | shogun: Soeren Sonnenburg master * r16be749 / (2 files): | 13:49 |
CIA-64 | shogun: Merge pull request #392 from pluskid/master | 13:49 |
CIA-64 | shogun: The filenames for two example code are wrong - http://git.io/LwwRqw | 13:49 |
shogun-buildbot | build #622 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/622 | 13:55 |
blackburn | sonney2k: feel like being a merge machine? :D | 13:59 |
pluskid | why java modular failed to build? | 14:07 |
blackburn | it was built succesfully later | 14:08 |
pluskid | but I failed on my local box | 14:09 |
blackburn | hmm | 14:09 |
blackburn | what is the error? | 14:09 |
pluskid | blackburn: https://gist.github.com/2204698 | 14:11 |
pluskid | something related to SWIG? | 14:11 |
blackburn | well yes but could you please try to clean current build and try to build it again? | 14:12 |
pluskid | OK | 14:12 |
n4nd0 | blackburn: I think that maybe we have a bug in MulticlassMachine | 14:19 |
blackburn | n4nd0: could be | 14:19 |
n4nd0 | blackburn: if apply is passed without arguments, it doesn't use the features that were used for training | 14:19 |
n4nd0 | blackburn: if perfectly ok if apply is called again with apply(feats_train), the same ones used for training for example | 14:20 |
blackburn | ehm.. why? | 14:20 |
n4nd0 | blackburn: I don't know yet ... I think I was messing up with one_vs_one but the problem is the same in one_vs_rest | 14:20 |
n4nd0 | blackburn: I have to check whether the MulticlassMachine loses reference to CFeatures somewhere or it happens in m_machines | 14:22 |
blackburn | wait | 14:22 |
blackburn | so | 14:22 |
blackburn | when you train() | 14:22 |
blackburn | and apply() it | 14:22 |
blackburn | what is going on? | 14:22 |
n4nd0 | seg fault | 14:23 |
blackburn | oh fine | 14:23 |
blackburn | :D | 14:23 |
n4nd0 | features is null when we do m_machines[i]->apply() | 14:24 |
n4nd0 | so the outputs[i] are all null | 14:24 |
blackburn | I see | 14:24 |
n4nd0 | and when we want to count the votes ... boom! :O | 14:25 |
n4nd0 | I can work on it | 14:26 |
blackburn | yeaah please do | 14:27 |
blackburn | n4nd0: probably once it trained for some features it should init_machines_for_apply | 14:28 |
n4nd0 | blackburn: I will take a look to it later today, have to do some other stuff now | 14:30 |
blackburn | ok I'll add this once I get home | 14:31 |
n4nd0 | mmm | 14:32 |
n4nd0 | maybe I should push then my changes in my branch and you modify it there? | 14:32 |
n4nd0 | or are you pushing to master and later I can rebase? | 14:32 |
blackburn | yes I can push to master | 14:33 |
blackburn | so just rebase, do not fix it in your branch | 14:33 |
-!- pluskid [~chatzilla@60.191.2.238] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120312181643]] | 14:34 | |
-!- pluskid [~chatzilla@60.191.2.238] has joined #shogun | 14:40 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 14:40 | |
shogun-buildbot | build #401 of csharp_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/csharp_modular/builds/401 blamelist: fernando.iglesiasg@gmail.com | 14:49 |
n4nd0 | shogun-buildbot: what have I done now? :( | 14:56 |
n4nd0 | shogun-buildbot: work! | 14:56 |
shogun-buildbot | What you say! | 14:56 |
-!- vikram360 [~vikram360@117.192.180.68] has joined #shogun | 15:04 | |
shogun-buildbot | build #389 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/389 | 15:13 |
-!- PhilTillet [~Philippe@38.124.160.48] has joined #shogun | 15:40 | |
PhilTillet | Heyy | 15:40 |
-!- PhilTillet is now known as RagnarX | 15:41 | |
-!- RagnarX [~Philippe@38.124.160.48] has quit [Client Quit] | 15:41 | |
-!- romi_ [~mizobe@187.57.4.159] has quit [Ping timeout: 246 seconds] | 15:43 | |
shogun-buildbot | build #393 of java_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/java_modular/builds/393 blamelist: pluskid@gmail.com, fernando.iglesiasg@gmail.com, sonne@debian.org | 15:52 |
-!- pluskid_ [~chatzilla@60.191.2.230] has joined #shogun | 15:53 | |
-!- pluskid [~chatzilla@60.191.2.238] has quit [Ping timeout: 240 seconds] | 15:54 | |
-!- pluskid_ is now known as pluskid | 15:54 | |
-!- romi_ [~mizobe@189-18-40-233.dsl.telesp.net.br] has joined #shogun | 15:56 | |
-!- pluskid_ [~chatzilla@li400-235.members.linode.com] has joined #shogun | 15:58 | |
-!- pluskid_ [~chatzilla@li400-235.members.linode.com] has quit [Client Quit] | 16:00 | |
-!- pluskid [~chatzilla@60.191.2.230] has quit [Ping timeout: 260 seconds] | 16:02 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 16:16 | |
shogun-buildbot | build #390 of python_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/390 blamelist: pluskid@gmail.com, fernando.iglesiasg@gmail.com, sonne@debian.org | 16:19 |
-!- pluskid [~chatzilla@li400-235.members.linode.com] has joined #shogun | 16:24 | |
pluskid | is there a recent API change to rename classify() to apply() in SVM? | 16:24 |
pluskid | I find I could not run the generate_testdata.py script | 16:25 |
pluskid | unless I change its calling of svm.classify() to svm.apply() | 16:25 |
shogun-buildbot | build #386 of ruby_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/386 blamelist: pluskid@gmail.com, fernando.iglesiasg@gmail.com, sonne@debian.org | 16:31 |
n4nd0 | sonney2k: any idea what's going on? all the builds are failing to compile :( | 16:32 |
-!- pluskid [~chatzilla@li400-235.members.linode.com] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120312181643]] | 17:07 | |
-!- blackburn [~qdrgsm@188.168.128.91] has joined #shogun | 17:30 | |
-!- PhilTillet [~Philippe@38.124.160.48] has joined #shogun | 17:32 | |
blackburn | hah all is failing lol | 17:32 |
n4nd0 | blackburn: what did I do man? | 17:34 |
blackburn | nothing | 17:34 |
blackburn | director kernel fails | 17:34 |
n4nd0 | shogun-buildbot blames me all the time | 17:34 |
n4nd0 | :( | 17:34 |
blackburn | nevermind, rather check logs | 17:34 |
n4nd0 | how do we fix that anyway? | 17:38 |
blackburn | no idea I'd rather ifdef that kernel for now | 17:41 |
blackburn | haha 614 warnings | 17:42 |
n4nd0 | did I hear record? | 17:44 |
blackburn | n4nd0: we had >2.5k once | 17:44 |
blackburn | when Soeren turned on doc warnings :D | 17:45 |
n4nd0 | blackburn: shit, we have to make more efforts then :P | 17:45 |
blackburn | well it is easy - just insert a couple of useless variables somewhere in template class | 17:46 |
-!- harshit_ [~harshit@182.68.160.94] has joined #shogun | 17:47 | |
harshit_ | hey n4nd0 | 17:49 |
n4nd0 | harshit_: hey man | 17:49 |
harshit_ | whats going on .. are you coding any thing for shogun for now ? | 17:50 |
harshit_ | Is there any thing you knw i can work on | 17:50 |
n4nd0 | harshit_: so blackburn had an idea about some scripting work | 17:51 |
n4nd0 | harshit_: to integrate tests with some datasets | 17:52 |
harshit_ | in which language ? | 17:52 |
n4nd0 | harshit_: idk if you would like that or you prefer we look for another thing more for C++ | 17:52 |
n4nd0 | that would probably depend more on how you want to do it but I'd that the most appropriate is to use a shell script probably | 17:53 |
blackburn | nice plots with iris dataset would be a nice example | 17:53 |
harshit_ | not much experience with shell script .. but i was thinking to incorporate any nice math library other than cblas and lapack | 17:54 |
harshit_ | will that example work, if i code it in C++ | 17:54 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 17:54 | |
n4nd0 | what do you think blackburn? | 17:55 |
n4nd0 | do we need other math libraries? | 17:55 |
blackburn | n4nd0: no idea what can be useful | 17:55 |
harshit_ | while porting code from MATLAB i found it really difficult to find functions in shogun | 17:56 |
blackburn | which? | 17:56 |
harshit_ | so i think that would be useful | 17:56 |
n4nd0 | harshit_: but those functions existed right? | 17:56 |
harshit_ | not really | 17:57 |
harshit_ | like for finding solutions for Ax=B | 17:57 |
n4nd0 | mmm | 17:57 |
harshit_ | i had to use pinv(A)*B | 17:57 |
harshit_ | and i am working on the which part, blackburn | 17:57 |
n4nd0 | I think that maybe cblas and lapack are a bit difficult to use | 17:58 |
n4nd0 | but probably the stuff is there | 17:58 |
blackburn | no, lapack is wonderful balanced for me | 18:00 |
harshit_ | yeah, cblas and lapack are the most powerful ones but still variety is always good | 18:00 |
n4nd0 | yeah might be | 18:01 |
n4nd0 | harshit_: wiking and I were talking some days ago about QP solvers | 18:02 |
n4nd0 | now we depend on private solvers (mosek or CPLEX) | 18:02 |
n4nd0 | would be nice to have something of that | 18:02 |
n4nd0 | but I don't know how difficult could that be | 18:02 |
blackburn | to integrate shouldn't be difficult, but to use.. | 18:02 |
harshit_ | havn't had much experience with optimization, but still nice idea .i'll think about it | 18:04 |
-!- PhilTillet [~Philippe@38.124.160.48] has quit [Ping timeout: 248 seconds] | 18:05 | |
CIA-64 | shogun: Sergey Lisitsyn master * r904696a / src/shogun/lib/DataType.h : Warnings removal - http://git.io/ZHYpNg | 18:06 |
CIA-64 | shogun: Sergey Lisitsyn master * rc2b5c34 / src/shogun/machine/MulticlassMachine.cpp : Fixed mc machine training - http://git.io/vaUG2g | 18:06 |
n4nd0 | wiking: do you remember the name of the QP library we talked about? I think that one was open source | 18:07 |
blackburn | libqp? | 18:08 |
n4nd0 | yeah that one | 18:09 |
blackburn | vojtech really rocks | 18:10 |
blackburn | :D | 18:10 |
n4nd0 | why? | 18:11 |
blackburn | he is an author of libqp, libocas and a lot of other things | 18:12 |
harshit_ | i think there is something related to libqp already in shogun .. | 18:12 |
blackburn | yes libqp itself | 18:12 |
harshit_ | oh | 18:13 |
n4nd0 | but can it be used? | 18:14 |
blackburn | why not | 18:14 |
n4nd0 | or is it left to do wrappers and stuff? | 18:14 |
n4nd0 | mmm I don't understand then why some parts need CPLEX | 18:14 |
blackburn | well it is a general library :) | 18:14 |
n4nd0 | maybe CPLEX does more than that | 18:15 |
blackburn | I have no expertise with it :( | 18:16 |
blackburn | n4nd0: actually I'm moving towards to the way libqp and libocas are implemented | 18:19 |
n4nd0 | blackburn: what do you mean with moving towards their way? | 18:19 |
blackburn | callbacks and other such stuff | 18:19 |
n4nd0 | callbacks? | 18:20 |
n4nd0 | is it a kind of event-driven programming? | 18:20 |
blackburn | yeah I'll show you in a min | 18:20 |
blackburn | n4nd0: https://github.com/shogun-toolbox/shogun/pull/394/files | 18:28 |
blackburn | ok have to go | 18:28 |
blackburn | see you | 18:28 |
n4nd0 | any special interesting part with callbacks? | 18:29 |
n4nd0 | ok nvm | 18:29 |
n4nd0 | will check it ;) | 18:29 |
n4nd0 | bye | 18:29 |
harshit_ | n4nd0 , started with gsoc application ? | 18:32 |
n4nd0 | not yet | 18:33 |
n4nd0 | you? | 18:33 |
harshit_ | was thinking to start today | 18:33 |
harshit_ | there are 6-8 days | 18:33 |
-!- blackburn [~qdrgsm@188.168.128.91] has quit [Ping timeout: 276 seconds] | 18:33 | |
harshit_ | but i heard somewhere that its good to submit early | 18:33 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 18:34 | |
n4nd0 | I think they will wait to check the applications | 18:35 |
n4nd0 | at least that makes sense for me | 18:35 |
n4nd0 | but who knows | 18:35 |
harshit_ | feeling depressed :( | 18:36 |
harshit_ | didnt find anyone who could help me with my deep learning idea | 18:36 |
harshit_ | :( | 18:36 |
harshit_ | i think its too young too be incorporated into shogun | 18:37 |
harshit_ | to* | 18:37 |
n4nd0 | maybe it could be interesting to do it after GSoC or sth | 18:38 |
harshit_ | yeah, but i thought my research and gsoc could go along with each other.. | 18:39 |
harshit_ | never mind | 18:39 |
n4nd0 | look at this way | 18:40 |
n4nd0 | you have the oportunity of learning other stuff | 18:41 |
harshit_ | yeah, right .. | 18:42 |
harshit_ | n btw do you ppl speak spanish in sweden | 18:43 |
harshit_ | i remember you saying that you use spanish keyboard | 18:44 |
n4nd0 | my laptop is Spanish | 18:44 |
n4nd0 | I'm Spanish :) | 18:44 |
n4nd0 | I speak Spanish with my Spanish friends in Sweden :P | 18:44 |
n4nd0 | but people speaks Swedish here normally | 18:45 |
harshit_ | okay actually i was also learning spanish : | 18:45 |
harshit_ | its a really nice language | 18:45 |
harshit_ | must say | 18:45 |
harshit_ | not better than C++ though | 18:45 |
harshit_ | :) | 18:45 |
n4nd0 | hahaha | 18:46 |
n4nd0 | I also prefer C++ | 18:47 |
n4nd0 | I don't like that much to speak Spanish actually | 18:47 |
n4nd0 | just sometimes | 18:47 |
-!- harshit_ [~harshit@182.68.160.94] has quit [Quit: Leaving] | 18:56 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 19:40 | |
CIA-64 | shogun: Soeren Sonnenburg master * r0f26327 / src/shogun/kernel/DirectorKernel.h : fix directorkernel init warning - http://git.io/WyC_mA | 20:51 |
shogun-buildbot | build #190 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/190 | 21:13 |
shogun-buildbot | build #189 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/189 | 21:24 |
-!- vikram360 [~vikram360@117.192.180.68] has quit [Ping timeout: 252 seconds] | 21:26 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 21:28 | |
blackburn | sonney2k: re to flexibility of libedrt | 21:36 |
blackburn | check compute_kernel and LocallyLinearEmbedding | 21:36 |
blackburn | in* | 21:36 |
blackburn | it can be easily changed with custom kernel or some other kernel | 21:37 |
CIA-64 | shogun: Soeren Sonnenburg master * re6d427d / src/shogun/kernel/DirectorKernel.h : copy some virtual function calls to director kernel to fix some errors - http://git.io/-h85rg | 21:38 |
@sonney2k | blackburn, not now - I have to look at newtonsvm | 21:39 |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 21:40 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 21:41 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 21:41 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 21:44 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 21:49 | |
n4nd0 | blackburn: I have just seen you pushed a change on mc :) | 21:58 |
blackburn | n4nd0: yes | 21:58 |
blackburn | and changed my name on github lol | 21:58 |
n4nd0 | blackburn: I see, surname style | 21:59 |
blackburn | yeah | 21:59 |
blackburn | n4nd0: I wonder do you know how to spell it? ;) | 21:59 |
n4nd0 | let me try | 22:00 |
n4nd0 | blackburn: shit, I think I am going to do it wrong | 22:00 |
blackburn | try ;) | 22:01 |
n4nd0 | blackburn: so I rather not to do it :P | 22:01 |
@sonney2k | blackburn, did anyone here talk to harshit (newton svm ...) | 22:02 |
blackburn | sonney2k: yes | 22:02 |
@sonney2k | his code is not yet perfect but not too bad for first attempt | 22:02 |
blackburn | sonney2k: why do you ask? | 22:02 |
@sonney2k | it is very inefficient at the moment (computing the whole feature matrix) | 22:02 |
@sonney2k | but that can be fixed... | 22:03 |
@sonney2k | he seems to know what he is doing | 22:03 |
blackburn | n4nd0: let me try to describe - lee-see-ts-yh-n | 22:03 |
@sonney2k | blackburn, wasnt' he the one wanting to integrate NNs? | 22:03 |
n4nd0 | blackburn: lisitsyn | 22:03 |
blackburn | sonney2k: yes deep learning | 22:03 |
blackburn | sonney2k: he wants still | 22:04 |
n4nd0 | sonney2k: he was asking about it today | 22:04 |
n4nd0 | sonney2k: do you think it could be a possible project? | 22:04 |
blackburn | I have doubts | 22:05 |
blackburn | nobody asked me yet but I have doubts :D | 22:05 |
n4nd0 | about what? | 22:05 |
blackburn | deep learning | 22:05 |
n4nd0 | about the project you mean in deep learning? | 22:05 |
blackburn | yeah | 22:05 |
n4nd0 | what doubts? | 22:05 |
@sonney2k | we have no mentor for that ... | 22:06 |
blackburn | yes | 22:06 |
blackburn | exactly | 22:06 |
n4nd0 | aham I see | 22:06 |
blackburn | I am not an expert in anything | 22:06 |
blackburn | and sonney2k is probably not an expert in deep learning | 22:06 |
blackburn | so hard to decide | 22:06 |
blackburn | :D | 22:06 |
n4nd0 | I remember that someone suggested in the mailing list that it was not required to have a mentor to carry out the project | 22:07 |
n4nd0 | a mentor specialized in the topic | 22:07 |
n4nd0 | sorry | 22:07 |
@sonney2k | n4nd0, yeah gunnar did | 22:07 |
@sonney2k | (the shoGUN gunnar ) | 22:07 |
blackburn | sonney2k: do you agree? | 22:07 |
@sonney2k | problem is that this is only true if we know a high quality C/C++ impl. of deep learning stuff | 22:08 |
@sonney2k | which would merely need porting/shogun integration | 22:08 |
@sonney2k | I know that real experts are the people doing lush and torch5 | 22:09 |
blackburn | I can hardly imagine yann lecun mentoring for us | 22:09 |
@sonney2k | but lush is some lisp dialect stuff | 22:09 |
blackburn | :D | 22:09 |
@sonney2k | and ronan collobert (torch main author) comes up with torch6 soon I guess :) | 22:10 |
blackburn | sonney2k: btw we have no lisp typemaps! ;) | 22:10 |
@sonney2k | luckily :D | 22:10 |
blackburn | gsomix: let you implement it ;) | 22:11 |
blackburn | to get sonney2k feeling not so lucky | 22:11 |
gsomix | blackburn, L??P? | 22:11 |
@sonney2k | if we want deep learning in shogun I think I should talk to some big shots and find someone who wants to help us | 22:11 |
blackburn | gsomix: yeah! | 22:11 |
@sonney2k | LUCP :) | 22:11 |
blackburn | sonney2k: you got it right! | 22:12 |
blackburn | sonney2k: try if you have time for that | 22:12 |
@sonney2k | I cannot write the pppp - i-i | 22:12 |
gsomix | sonney2k, check mail, please | 22:13 |
@sonney2k | shogun seriously lacks boosting algortihsm, decision trees and deep learning (even when all projects a done after this gsoc) | 22:13 |
blackburn | sonney2k: decision trees is painful thing | 22:14 |
n4nd0 | sonney2k: what about SO, isn't that important? | 22:14 |
@sonney2k | n4nd0, yeah but you said you will do it this gsoc - right ? | 22:15 |
@sonney2k | blackburn, yeah but people *love* trees | 22:15 |
blackburn | haha | 22:15 |
n4nd0 | sonney2k: :) that's what I wish | 22:15 |
blackburn | sonney2k: most painful is to design it properly | 22:15 |
blackburn | I would rather jump out of the window :D | 22:16 |
@sonney2k | blackburn, anyway what I am saying is that harshit could do a good job - not sure if he wants to do sth else (like trees :D | 22:16 |
blackburn | sonney2k: lets see it is unclear for me yet | 22:16 |
@sonney2k | blackburn, yeah for me it is also not clear - lets see how he reacts to my (many!) comments | 22:17 |
@sonney2k | gsomix, no worries about incomplete patches (as long as they don't break python2 support) | 22:17 |
blackburn | he asked to review his code | 22:17 |
blackburn | many times actually | 22:17 |
n4nd0 | sonney2k: this afternoon he asked for something to work on so I think it could be ok another topic | 22:18 |
gsomix | sonney2k, ok | 22:18 |
blackburn | I just did not want to do it hahah | 22:18 |
n4nd0 | we better suggest him next time in any case | 22:18 |
@sonney2k | blackburn, I only saw his pull request 2 days ago? anyway I did now more or less | 22:18 |
blackburn | sonney2k: well you are an svm expert so let you review such things | 22:19 |
blackburn | ;) | 22:19 |
@sonney2k | I have too many comments so I need him to answer first before digging deeper | 22:19 |
@sonney2k | blackburn, well he said that the matlab code and his impl. of newton svm give the same result | 22:19 |
blackburn | for 5 feature vectors | 22:19 |
blackburn | that can be a problem.. | 22:19 |
@sonney2k | good enough for me :) | 22:20 |
blackburn | thanks google for grouping emails | 22:20 |
@sonney2k | anyway the C++ code is too memory demanding / slow currently | 22:20 |
blackburn | I just received a bunch of your comments | 22:20 |
@sonney2k | n4nd0, ok - I have no idea what timezone he is in - lets see if we find sth we can mentor... | 22:21 |
blackburn | gsomix: check directorkernel, that's the thing I have described you earlier | 22:21 |
blackburn | sonney2k: I am pretty sure he +4-5 from you | 22:21 |
n4nd0 | n4nd0: he is Indian, I normally speak with him in the mornings ... I quite not remember meeting him at this time | 22:22 |
blackburn | n4nd0: you always highlight yourself ;) | 22:23 |
blackburn | how can it be? | 22:23 |
n4nd0 | fuck!! I have a problem with that | 22:23 |
n4nd0 | I am an egocentric :P | 22:23 |
blackburn | lol | 22:23 |
n4nd0 | sonney2k: he is Indian, I normally speak with him in the mornings ... I quite not remember meeting him at this time | 22:23 |
@sonney2k | n4nd0, we all are | 22:23 |
n4nd0 | ok, you got it :) | 22:24 |
n4nd0 | sonney2k: you think so? I shouldn't worry about that then | 22:24 |
* sonney2k cares most (after shogun code quality) about his hairstyle, taint, muscles :D | 22:25 | |
blackburn | sonney2k: btw are you well again? | 22:25 |
@sonney2k | I would wish | 22:25 |
blackburn | but not? | 22:25 |
@sonney2k | must be the code quality I was seeing :D | 22:25 |
@sonney2k | ...and am producing | 22:26 |
@sonney2k | blackburn, we really need the tests we have in python also in other languages | 22:26 |
@sonney2k | this legacy test suite is a pain in the .... | 22:27 |
blackburn | sonney2k: yeah but it turned to be useful | 22:27 |
@sonney2k | blackburn, yeah but the example based tests have higher coverage... but are not available in all languages yet | 22:28 |
@sonney2k | only python_{static,modular} | 22:28 |
@sonney2k | so much easier to maintain | 22:28 |
blackburn | sonney2k: I have doubts it has high coverage.. | 22:29 |
@sonney2k | n4nd0, btw not sure if you have seem it - the comments you wrote in your example - I moved them to descriptions (examples have no documentation in the file - these are all separate) | 22:30 |
@sonney2k | n4nd0, the rationale is: if someone converts your example to e.g. octave_modular then the same description will be prepended | 22:30 |
@sonney2k | blackburn, 170 tests much higher then we have in the old at least | 22:31 |
@sonney2k | (for python modular only...) | 22:31 |
blackburn | sonney2k: oh March, 27 is the day I talked to you for the first time :D | 22:31 |
gsomix | blackburn, do you mean director classes? | 22:31 |
blackburn | gsomix: yeaahh I hope they would work someday | 22:32 |
n4nd0 | sonney2k: I understand, ok | 22:32 |
@sonney2k | gsomix, yeah | 22:32 |
n4nd0 | I should do the doc for QDA by the way, I think I still have a TODO there :O | 22:32 |
gsomix | blackburn, ok | 22:32 |
blackburn | sonney2k: the problem is they should be updated and cleaned | 22:33 |
@sonney2k | n4nd0, yes please | 22:33 |
@sonney2k | blackburn, the problem is we don't run them by default | 22:33 |
blackburn | in means of quantity it covers more things for sure | 22:33 |
@sonney2k | if we would we would see any breakage when it appears and they would be kept up to date | 22:33 |
@sonney2k | and it is sooo easy to update them | 22:33 |
@sonney2k | just run generator.py <example_name> | 22:33 |
blackburn | we will see a lot of breakages next months | 22:34 |
@sonney2k | and commti | 22:34 |
@sonney2k | blackburn, do you think so? I am not aware of changes like we had with SGVector etc | 22:34 |
@sonney2k | for this gsoc at least | 22:35 |
blackburn | sonney2k: yeah I am pretty sure, with labels subclassing at least | 22:35 |
@sonney2k | only additions... | 22:35 |
@sonney2k | ahh yes | 22:35 |
@sonney2k | and actually SGVector refcounting :) | 22:35 |
blackburn | features subclassing | 22:35 |
@sonney2k | features subclassign? | 22:35 |
blackburn | probably.. | 22:35 |
blackburn | sonney2k: ah I had one idea before | 22:36 |
blackburn | do you think it can be useful to add list of strings into the features | 22:36 |
blackburn | to store feature vector names | 22:36 |
blackburn | sonney2k: new features will come with latent models | 22:36 |
@sonney2k | blackburn, it is probably useful but the overhead... | 22:37 |
blackburn | if it is needed | 22:37 |
blackburn | sonney2k: what is the overhead? | 22:37 |
@sonney2k | we used to have 0 byte overhead | 22:37 |
blackburn | it can be optional | 22:37 |
@sonney2k | I mean SGVector used to be ptr + len | 22:37 |
blackburn | I can't agree with you | 22:37 |
blackburn | serialization and migrations stuff | 22:37 |
blackburn | have a lot of overhead | 22:37 |
n4nd0 | btw what about this guy who asked about reference counting for SGVector? | 22:37 |
@sonney2k | serailization /migration are not in sgvector/matrix | 22:38 |
blackburn | sonney2k: but in features | 22:38 |
blackburn | and other classes | 22:38 |
@sonney2k | blackburn, yeah but that is just once for whole feature matrix | 22:38 |
@sonney2k | not per feature / feature vector | 22:38 |
@sonney2k | n4nd0, I am just browsing the list... | 22:39 |
blackburn | I can't see extra overhead there | 22:39 |
n4nd0 | sonney2k: cool, sorry I rushed | 22:39 |
blackburn | if list of names is provided - it stores this list | 22:39 |
blackburn | if not - no overhead at all | 22:39 |
@sonney2k | blackburn, if you have an SGVector of bytes and each dim is described there? | 22:39 |
blackburn | one pointer | 22:39 |
@sonney2k | huge overhead | 22:40 |
blackburn | noo | 22:40 |
blackburn | I did not suggest this crazy way of describing features | 22:40 |
blackburn | just optional list of names in CFeatures* | 22:40 |
blackburn | "1.png" "2.png" "3.png" "4.png" | 22:41 |
blackburn | sonney2k: did you suggest to Evan to implement it? | 22:43 |
@sonney2k | blackburn, yeah sure no problem with that (even for dims then - but not sure if this is so useful...) | 22:43 |
@sonney2k | dammed | 22:43 |
@sonney2k | gtg | 22:43 |
@sonney2k | cu | 22:43 |
blackburn | see you | 22:43 |
blackburn | ;) | 22:43 |
n4nd0 | bye | 22:44 |
blackburn | n4nd0: is o-v-o stuff ready? | 22:45 |
n4nd0 | blackburn: I have not tested it yet with your last commit | 22:46 |
blackburn | ok! | 22:46 |
n4nd0 | blackburn: but classification is ready, yes | 22:46 |
n4nd0 | blackburn: the set_subset part for kernel machines is still missing though | 22:47 |
blackburn | ah ok I'll implement it | 22:47 |
n4nd0 | as you wish, I will test the classify o-v-o and commit it soon | 22:48 |
blackburn | n4nd0: you already have contibuted with 1.7k LoC | 22:51 |
blackburn | and gsomix with 1.9k LoC | 22:52 |
n4nd0 | blackburn: oh, is that good? | 22:52 |
n4nd0 | I use lot of spacing :P | 22:52 |
blackburn | n4nd0: no idea I am just clicking buttons on github | 22:52 |
@sonney2k | n4nd0, comments are even better :D | 22:53 |
n4nd0 | sonney2k: yeah, I realized I have not commented the code that much :( | 22:53 |
n4nd0 | I feel bad about that | 22:53 |
-!- prm [~prm@115.248.130.148] has joined #shogun | 22:54 | |
blackburn | sonney2k: are you back? ;) | 22:54 |
@sonney2k | yes | 22:54 |
@sonney2k | n4nd0, well good code doesn't need comments... | 22:55 |
blackburn | sonney2k: have you seen list of scikits gsoc topics? | 22:55 |
blackburn | https://github.com/scikit-learn/scikit-learn/wiki/A-list-of-topics-for-a-google-summer-of-code-%28gsoc%29-2012 | 22:55 |
@sonney2k | blackburn, the situation relaxed | 22:55 |
n4nd0 | sonney2k: I don't dare to say that mine's such a good code | 22:55 |
blackburn | sonney2k: that's fine ;) | 22:56 |
blackburn | sonney2k: most interesting thing is they keep 'possible candidate' there | 22:57 |
@sonney2k | blackburn, I didn't see these before - not a lot though? | 22:57 |
@sonney2k | well every highly interested student knows that n4nd0 is possible candidate for SO, gsomix for accessibility, wiking for latent svm | 22:58 |
@sonney2k | no idea about others | 22:58 |
n4nd0 | blackburn for multitask! :) | 22:58 |
blackburn | hah | 22:59 |
@sonney2k | ahh yes ;) | 22:59 |
* sonney2k goes back answering that email | 22:59 | |
@sonney2k | n4nd0, btw regarding students - it also depends on how many slots we get. last year we had 5 but I would not expect we get much more than that if at all...) | 23:03 |
blackburn | sonney2k: I hope more.. | 23:04 |
blackburn | sonney2k: remember we have heiko | 23:04 |
n4nd0 | sonney2k: yeah ... and lot of people are starting to be around here ... is going to be difficult to get a place there | 23:04 |
blackburn | I really hope they would give 6-7 | 23:05 |
@sonney2k | it is not that difficult ... just contribute a lot before gsoc even starts (potentially even start with `your' project) and you are in. | 23:06 |
@sonney2k | last year heiko and blackburn where students like this (submitting patches like crazy before gsoc did start)... | 23:06 |
@sonney2k | the other 3 submitted convincing patches too - but by far not so many | 23:07 |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 23:07 | |
n4nd0 | we'll have to keep up with the hard work then :) | 23:07 |
-!- blackburn_ [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 23:08 | |
@sonney2k | n4nd0, blackburn certainly always has ideas :) | 23:08 |
n4nd0 | that is very good actually :) | 23:09 |
blackburn_ | had you been talking about me by the moment I came? ;) | 23:09 |
blackburn_ | yes I have a lot of crazy ideas | 23:09 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 23:10 | |
blackburn_ | n4nd0: do you mind constructing a big submarine? | 23:11 |
n4nd0 | blackburn_: eing?!? what do you mean? lol | 23:11 |
gsomix | blackburn_, yellow submarine :] | 23:11 |
blackburn_ | sonney2k: and yes, SGObject -> CCCP transition is still ongoing | 23:11 |
blackburn_ | gsomix: why not, yellow works well | 23:12 |
blackburn_ | sounds pretty beatly | 23:12 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 23:14 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 23:23 | |
@sonney2k | n4nd0, blackburn_ - can I get your opinion on that reference counting thing and SGVector etc business? | 23:23 |
@sonney2k | we basically need reference counts even for the simple storage container SGVector / SGMatrix etc | 23:24 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 23:24 | |
@sonney2k | we realized that we are running into trouble without | 23:24 |
n4nd0 | sonney2k: I think that it would be very interesting to have it | 23:24 |
@sonney2k | <sonney2k> n4nd0, blackburn - can I get your opinion on that reference counting thing and SGVector etc business? | 23:24 |
blackburn | damn are you talking only when I get disconnected? | 23:24 |
@sonney2k | n4nd0, so we have it currently for SGObject | 23:25 |
@sonney2k | one can do SG_REF / SG_UNREF and all good | 23:25 |
-!- blackburn_ [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 23:25 | |
@sonney2k | now with SGVector we were always passing around copies | 23:25 |
@sonney2k | e.g. return SGVector<flaot64_t>(xxx) | 23:25 |
@sonney2k | that is ptr + associated length | 23:25 |
@sonney2k | now if we add more fields overhead will kill performance | 23:26 |
@sonney2k | after all we do this *everywhere* | 23:26 |
@sonney2k | in learning algorithms etc | 23:26 |
n4nd0 | I understand | 23:26 |
@sonney2k | so even a inc_ref() is significant overhead | 23:26 |
@sonney2k | and having a common base class for reference counts will create even more overhead (vtable...) | 23:27 |
@sonney2k | so the question is what we can do about it (or should) | 23:27 |
n4nd0 | but even if this inc_ref() and the reference counts will increase overhead | 23:28 |
n4nd0 | isn't better than working with the copies? | 23:28 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 23:29 | |
-!- blackburn [~qdrgsm@188.168.128.91] has joined #shogun | 23:30 | |
blackburn | ok stable connection finally | 23:30 |
blackburn | changed room :D | 23:30 |
blackburn | sonney2k: n4nd0: still here? | 23:30 |
@sonney2k | n4nd0, ok so you are +1 for using references? | 23:30 |
@sonney2k | so we return SGVector<float64_t>& ? | 23:31 |
n4nd0 | I think so | 23:31 |
n4nd0 | from my point of view it is also more intuitive | 23:31 |
@sonney2k | and then add whatever we want, blackburns dimnames, derivative class etc | 23:31 |
blackburn | I do not think we should count references of vectors.. | 23:31 |
n4nd0 | but that is something more personal I guess | 23:32 |
@sonney2k | blackburn, we have to | 23:32 |
n4nd0 | blackburn: why? | 23:32 |
blackburn | what is the motivation? | 23:32 |
@sonney2k | blackburn, it is messy with this do_free business - one never knows if this vector is still in use by someone | 23:32 |
blackburn | where in code it becomes a problem? | 23:33 |
@sonney2k | blackburn, several algos fetch vectors and we never know who should free it | 23:33 |
@sonney2k | blackburn, I don't have a concrete example but any place where we don't have a sole owner of the vector | 23:33 |
@sonney2k | (or matrix / string list) | 23:33 |
blackburn | it is pretty intrusive change | 23:34 |
blackburn | we shall to think about its pros and cons.. | 23:34 |
@sonney2k | so I like Evan Shelhamer's suggestion with the base class | 23:34 |
@sonney2k | blackburn, exactly - that is why I ask you both | 23:34 |
blackburn | for me access/free approach is ok | 23:34 |
blackburn | but if there is any example of something not possible with it - yeah we should change it then | 23:35 |
@sonney2k | for me no longer - it is too messy | 23:35 |
@sonney2k | I never know if I can SG_FREE the vector or not (and I wrote that code...) | 23:35 |
n4nd0 | I think that it gets a bit difficult to know that, exactly | 23:36 |
n4nd0 | I also assumed that it was lack of practice though ... | 23:36 |
@sonney2k | blackburn, for example consider CLabels having an SGVector internally and passing it to the outside | 23:36 |
blackburn | actually we can avoid reference counting and just use it as object | 23:36 |
blackburn | and add free stuff to destructor | 23:36 |
@sonney2k | blackburn, but who frees it then? | 23:36 |
blackburn | destructor | 23:37 |
@sonney2k | blackburn, who calls delete then :D | 23:37 |
blackburn | scope | 23:37 |
n4nd0 | wouldn't that increase the calls to delete around the code? | 23:37 |
blackburn | { | 23:38 |
blackburn | ... | 23:38 |
blackburn | SGVector<float64_t> vector(); | 23:38 |
blackburn | ... | 23:38 |
blackburn | } | 23:38 |
n4nd0 | isn't that a limitation? | 23:38 |
blackburn | well it won't be deleted if we return it | 23:38 |
blackburn | or pass a reference to other function | 23:38 |
@sonney2k | blackburn, except that if vecto is passed to the outside it might get deleted because the original object is gone | 23:38 |
n4nd0 | it may be interesting to destroy the objects before the end of the scope | 23:39 |
n4nd0 | yeah, that is even a bigger limitation I guess | 23:39 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 23:39 | |
blackburn | sonney2k: that shouldn't happen, are you sure? | 23:39 |
@sonney2k | blackburn, CLabels | 23:39 |
@sonney2k | someone calls get_labels and gets the vector | 23:39 |
@sonney2k | then the CLabels object is deleted | 23:40 |
blackburn | ah in that mean | 23:40 |
@sonney2k | bumm | 23:40 |
blackburn | I forgot that is shared | 23:40 |
blackburn | ok then | 23:40 |
@sonney2k | question is how expensive this is | 23:41 |
-!- prm [~prm@115.248.130.148] has quit [Quit: Leaving] | 23:41 | |
@sonney2k | not sure if we can do it the SGObject way | 23:41 |
blackburn | I actually wish we could map memory at least for python | 23:41 |
n4nd0 | I don't really see that point, why the reference counting can increase overhead that much? | 23:41 |
blackburn | without copying it in in/out typemaps | 23:41 |
@sonney2k | blackburn, different issue | 23:42 |
blackburn | yes but much more important (for me) | 23:42 |
-!- prm [~prm@115.248.130.148] has joined #shogun | 23:42 | |
blackburn | I mean SGVector stuff works OK for now :) | 23:42 |
@sonney2k | n4nd0, at least currently we check for NULL ptrs etc and deriving from some class with potentially virtual functions will create a vtable for that class | 23:42 |
blackburn | sonney2k: I believe you care about it too much | 23:43 |
n4nd0 | sonney2k: oh I see, I clearly understimated this problem | 23:43 |
blackburn | passing vectors is occured only a few times | 23:44 |
blackburn | there can't be such huge overhead | 23:45 |
@sonney2k | blackburn, well what we could do for 0 overhead cases is pass around ptr / len directly (if needed) | 23:47 |
-!- prm [~prm@115.248.130.148] has quit [Client Quit] | 23:47 | |
@sonney2k | so I am very much in favor of having that ref counting across SGObject/SGVector/SG* | 23:48 |
@sonney2k | with some common base class | 23:48 |
blackburn | I'm ok with it | 23:48 |
@sonney2k | *but* we need a common way of passing SGVector / SGObject around | 23:48 |
@sonney2k | SGObject's are passed around as ptr's | 23:49 |
@sonney2k | so I we should pass SGVectors as pointers too | 23:49 |
-!- makokal [~makokal@212.201.44.246] has joined #shogun | 23:49 | |
blackburn | that would mean a lot of work possibly | 23:49 |
@sonney2k | yeah - the alternative is to have it inconsistent for now and maybe later even change SGObject's to not pass pointers but references? | 23:50 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 23:51 | |
gsomix | sonney2k, blackburn http://pastebin.com/Gb2RMfPH latest build. I continue to work on it. | 23:52 |
blackburn | sonney2k: it seems we would have to try | 23:52 |
@sonney2k | gsomix, that is python3 already? | 23:52 |
gsomix | sonney2k, yep. there are some small problems. | 23:53 |
@sonney2k | blackburn, so with references first? | 23:53 |
blackburn | gsomix: please install pickle and scipy | 23:53 |
blackburn | sonney2k: I like references mostly | 23:53 |
gsomix | blackburn, ok | 23:53 |
blackburn | but it is obvious that any experiments should be done in a separate branch | 23:54 |
@sonney2k | gsomix, line 178 - python3 converts int's to float when x/2 returns a float | 23:55 |
@sonney2k | so you just need to cast to int | 23:55 |
@sonney2k | blackburn, yeah one could do a+b for SGVector objects | 23:55 |
blackburn | sonney2k: have you ever seen accurate derivation of crammer-singer formulation? | 23:55 |
@sonney2k | so that is nice | 23:56 |
blackburn | dual task | 23:56 |
blackburn | sorry | 23:56 |
blackburn | dual task of C-S | 23:56 |
@sonney2k | blackburn, derivation of dual starting from primal? | 23:56 |
blackburn | yes | 23:56 |
@sonney2k | no - but I guess it is easy | 23:56 |
gsomix | sonney2k, thx | 23:56 |
gsomix | sonney2k, also I know how to fix the problem "TypeError" in kernel_weighted_degree_position_string_modular.py | 23:57 |
blackburn | sonney2k: have time to help? ;) | 23:57 |
@sonney2k | gsomix, you don't have lzo installed | 23:57 |
@sonney2k | blackburn, haha | 23:58 |
blackburn | sonney2k: I am for real actually :) | 23:59 |
--- Log closed Tue Mar 27 00:00:19 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!