IRC logs of #shogun for Tuesday, 2013-02-12

« Monday, 2013-02-11 Index Wednesday, 2013-02-13 »
--- Log opened Tue Feb 12 00:00:45 2013
-!- cavalier- [~cavalier-@70.118.110.73] has quit [Ping timeout: 260 seconds]01:42
-!- cavalier- [~cavalier-@70.118.110.73] has joined #shogun01:45
-!- blackburn [~blackburn@188.168.14.94] has quit [Ping timeout: 244 seconds]03:07
-!- FSCV [~FSCV@204.45.110.163] has quit [Quit: Leaving]03:08
-!- blackburn [~blackburn@188.168.3.25] has joined #shogun03:24
-!- FSCV [~FSCV@189.139.150.187] has joined #shogun04:49
-!- FSCV [~FSCV@189.139.150.187] has quit [Ping timeout: 240 seconds]04:56
-!- FSCV [~FSCV@173.254.212.46] has joined #shogun04:56
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun05:54
n4nd0good news gsoc announcement05:56
n4nd0sonney2k, sonne|work, blackburn : so I read it is important to have a release before gsoc, how can I help with that?05:56
blackburnn4nd0: hey05:57
n4nd0good morning :)05:57
blackburnn4nd0: well we need to close all known bugs05:58
n4nd0blackburn: in github issues?05:59
blackburnn4nd0: yes and in general05:59
-!- blackburn [~blackburn@188.168.3.25] has quit [Quit: Leaving.]06:10
-!- FSCV [~FSCV@173.254.212.46] has quit [Quit: Leaving]06:13
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun06:45
n4nd0blackburn: is tests/integration something still in use or deprecated?07:14
blackburnn4nd0: we still can use it07:22
blackburnbut it could fail a lot I expect07:22
n4nd0blackburn: ok07:23
blackburnn4nd0: could you please run make check-valgrind and paste the result somewhere?07:24
blackburnin libshogun07:24
n4nd0blackburn: ok, but I need a bit of time, I was trying to make all the examples in libshogun07:32
n4nd0but it seems that my Eigen is not detected by configure07:32
n4nd0so converter examples are not properly built07:32
blackburnhmm true07:34
blackburnit is not properly handled I am afraid07:35
n4nd0blackburn: in configure?07:36
blackburnn4nd0: I mean examples are not guarded07:37
blackburnn4nd0: but why it doesn't detect your eigen? any ideas?07:37
n4nd0blackburn: my Eigen is in /usr/local/include07:38
n4nd0and it seems that the configure script doesn't look in there07:38
n4nd0I have used the flag --incdir=/usr/loccal/include/eigen307:39
n4nd0but I am afraid the incdir variable is not used by check eigen3 in configure07:39
blackburnit uses pkg-config as I can see07:42
n4nd0I don't have that program installed here :D07:43
n4nd0blackburn: the thing I am wondering is07:43
n4nd0I can just re-install Eigen and I guess it should be fine07:43
n4nd0but maybe configure should find it if it is in /usr/local/include07:44
n4nd0nvm, pkg-config is installed07:47
blackburnn4nd0: yes it should but it currently uses *only* pkg-config07:47
n4nd0yep07:49
n4nd0that's probably fine07:49
n4nd0it's a local problem of mine, I need to make my pkg-config detect eigen307:49
n4nd0Eigen3 support .. yes :)07:58
blackburnnice08:09
blackburnn4nd0: so finally works?08:09
n4nd0yes08:10
n4nd0making check-valgrind on the way08:11
blackburnthanks!08:15
blackburnI have to go, be back in a few hours08:16
n4nd0ok08:17
n4nd0i will paste the results and take a look08:18
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds]08:38
-!- sumit_ [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun08:58
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has joined #shogun09:06
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]09:38
n4nd0blackburn: valgrind results10:18
n4nd0http://pastebin.com/TYjzrCzJ10:18
n4nd0everything looks fine except from libary_serialization10:18
n4nd0I am guessing that we are OK with suppressed errors10:19
-!- wiking_ is now known as wiking10:32
-!- wiking [~wiking@info2k1.hu] has quit [Changing host]10:33
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun10:33
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun10:36
n4nd0blackburn: hello hello10:36
blackburnhey10:36
n4nd0the results of make valgrind-check10:36
n4nd0http://pastebin.com/TYjzrCzJ10:37
blackburnn4nd0: thanks10:37
blackburnwhoops10:37
n4nd0suppressed errors are fine I guess?10:37
blackburnyes10:37
blackburnnice to know my code is a little b0rken10:38
n4nd0oh fuck10:38
n4nd0converter_kernellocallylinearembedding10:38
n4nd0and some others10:39
blackburnhah my bad10:39
n4nd0do you know why can it be?10:39
blackburnno - I'd have to investigate10:39
blackburnn4nd0: or you can if you want :)10:39
blackburnactually it didn't produce so many errors before - quite strange for me10:40
n4nd0blackburn: maybe something is wrong in my system10:40
n4nd0when I run check.sh I get error everywhere10:41
blackburnno as it should work on any system :)10:41
n4nd0that cannot be good10:41
blackburnyes check.sh doesn't work in libshogun examples10:41
n4nd0why not?10:41
blackburnn4nd0: could you try to run converter_kernellocallylinearembedding?10:41
blackburnand valgrind it too10:41
n4nd0blackburn: yep10:42
n4nd0what is wrong with check.sh then?10:42
blackburnn4nd0: I don't know - I use make check-examples10:43
n4nd0blackburn: http://dl.dropbox.com/u/11020840/shogun/valgrind.log10:46
n4nd0kernellocallylinearembedding10:46
blackburnthanks10:47
n4nd0blackburn: I can investigate a bit now10:47
n4nd0if you are not already on it?10:47
blackburnn4nd0: no I am now under windows at job :(10:48
blackburnn4nd0: looks surprising to me10:48
n4nd0blackburn: I am not sure if I ever valgrind-ed tapkee10:49
n4nd0did you?10:49
blackburnn4nd0: that matrix is not properly initialized strange10:49
blackburnyes I did10:49
blackburnthere were no errors :)10:49
blackburnwell the fact is LLE works ok but KLLE works bad10:49
blackburnas they are the same in tapkee I find that quite strange10:49
blackburnn4nd0: I expect I forgot to initialize something or so10:51
n4nd0blackburn: do you happen to know if resize in Eigen initializes the matrix or vector elements?11:43
n4nd0from the doc it seems to me that it doesn't11:43
blackburnn4nd0: no I don't know11:57
n4nd0blackburn: take a look12:14
n4nd0http://eigen.tuxfamily.org/dox/SimplicialCholesky_8h_source.html12:14
n4nd0valgrind complains about line 819, so d must be unintiliazed12:15
n4nd0and I think this is caused because m_diag is not initialized12:15
n4nd0it is resized in 76412:16
n4nd0but the values are not set12:16
n4nd0and they are used in 803 in an intermediate computation of d12:16
blackburnn4nd0: scary!12:21
n4nd0yeah...12:21
n4nd0but I think it is not due to that probably12:22
n4nd0it must be another thing12:22
n4nd0blackburn: a think that confuses me12:22
n4nd0DefaultSparseSolver in tapkee might be SimplicialCholesky or SimplicialLDLT12:23
n4nd0SimplicialCholesky is used at the moment12:23
n4nd0but in the method where valgrind complains, DoLDLT is set to true12:23
n4nd0it seems a bit contradictory, doesn't it?12:23
blackburnyeah12:24
n4nd0maybe that method should be called with DoLDLT=false?12:25
n4nd0just guessing no real idea :D12:25
blackburnI don't know actually12:28
-!- sumit_ [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Ping timeout: 245 seconds]12:28
blackburnthat all sounds strange12:31
blackburnagain, LLE and KLLE should both fail12:31
blackburnor not12:31
blackburnanyway I'll take a look tonight12:32
n4nd0that's true12:32
n4nd0I am going to check if I discover why it fails in one but not the other12:35
blackburnmakes sense, thanks12:45
n4nd0blackburn: look at this, it confuses me but maybe it tells you something :)12:48
n4nd0to simply to std::cout << matrix << std::endl12:48
n4nd0of the matrix we are giving to the solver12:48
n4nd0rises the conditional jump error in klle and not in lle12:49
n4nd0it must be something related to how this matrix is created12:49
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Ping timeout: 240 seconds]12:50
-!- shogun-buildbot_ [~shogun-bu@7nn.de] has joined #shogun12:50
blackburnyes12:50
blackburnn4nd0: https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/converter/LocallyLinearEmbedding.cpp#L8712:52
blackburnabsent in https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/converter/KernelLocallyLinearEmbedding.cpp#L6512:52
n4nd0blackburn: you are a professional :)12:53
n4nd0let me try it12:53
blackburnn4nd0: probably that's the reason - this should fire some warning though12:53
blackburnn4nd0: added the issue to not forget about setting that12:56
n4nd0blackburn: that solved valgrind's complains indeed12:57
blackburnI expect others are caused by something similar..12:57
n4nd0probably12:57
n4nd0blackburn: about the warning thing12:58
n4nd0you mean that we should enable that in tapkee right?12:58
n4nd0if some parameters are not set, say it12:59
blackburnyes sure n4nd013:02
n4nd0I apply the change if so13:03
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has quit [Ping timeout: 256 seconds]13:14
-!- Hausas [~admin@185.6.154.4] has joined #shogun13:19
-!- Hausas [~admin@185.6.154.4] has quit [Changing host]13:19
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun13:19
-!- Hausas [~admin@unaffiliated/hausas] has left #shogun []13:30
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has joined #shogun13:30
wikingmmm14:01
wikingblackburn: one gsoc idea: recommendation system14:01
blackburnwiking: mentor?14:05
wikingblackburn: happy to be14:17
wikingmoreover i can get one person i think14:17
wikingfrom a company that is working on recommendation systems14:18
wikingi'll have a chat with him today14:19
wikingi can ask him what he would have in his mind14:19
wikingfor sure he would be ok to comentor with me a project like that14:19
wiking(co-mentor as he is rather busy)14:19
-!- FSCV [~FSCV@204.45.110.163] has joined #shogun14:32
-!- Guest84830 [73fe7104@gateway/web/freenode/ip.115.254.113.4] has joined #shogun15:52
Guest84830Hi everyone, I am Parthasarathy G, a computer science under-grad student from IIT Madras, I am interested in contributing for Shogun, can anyone please guide me regarding this?15:53
n4nd0Guest84830: hey15:59
n4nd0the roadmap normally starts forking the repo, fetching a local copy and building it16:00
n4nd0then you could work on something you are interested in or we could suggest you some kind of contribution16:01
Guest84830Is there an existing git repo i could pull from?16:01
n4nd0Guest84830: yeah sure16:01
n4nd0the repository is in github16:01
n4nd0https://github.com/shogun-toolbox/shogun16:01
Guest84830I am currently cloning the repo, thanks for the link.16:06
n4nd0you are welcome16:06
Guest84830n4nd0: I have finished building the code, but not exactly found where the binaries are16:23
n4nd0Guest84830: libshogun.so should be in src/shogun16:25
n4nd0it is a good idea to make install too so you can use it from any part on your system16:26
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:27
Guest84830ah yes, i have finished that.16:29
n4nd0I will be back in afew minutes16:30
-!- Guest84830 [73fe7104@gateway/web/freenode/ip.115.254.113.4] has quit [Ping timeout: 245 seconds]16:33
n4nd0re16:40
-!- blackburn [~blackburn@83.234.54.3] has joined #shogun17:36
-!- FSCV_ [~FSCV@173.254.212.46] has joined #shogun17:38
-!- FSCV [~FSCV@204.45.110.163] has quit [Ping timeout: 252 seconds]17:39
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has quit [Quit: leaving]17:49
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun17:51
-!- naruto [73fe7104@gateway/web/freenode/ip.115.254.113.4] has joined #shogun18:11
-!- naruto is now known as Guest6104518:11
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Quit: Page closed]18:11
Guest61045Hi sonney2k , I was working on building from source some time ago, i have finished the building part, but what about launching the gui?18:16
-!- blackburn [~blackburn@83.234.54.3] has quit [Quit: Leaving.]18:39
-!- Guest61045 [73fe7104@gateway/web/freenode/ip.115.254.113.4] has quit [Quit: Page closed]18:48
-!- SCV__ [~FSCV@173.254.212.46] has joined #shogun20:19
-!- Netsplit *.net <-> *.split quits: FSCV_20:25
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun21:31
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has joined #shogun23:01
KMcQuistenHello all.  Hope things are going well23:02
KMcQuistenI have a quick question that I hope I can get answered23:03
wikingshoot23:05
KMcQuistenI use LibSVR quite a bit, and have generally used the epsilon-SVR flavor.  The parameter optimization framework only allows one to optimize nu, the maximum fraction of data points that are support vectors.  I know that there's a relationship between nu and the tube epsilon.23:05
KMcQuistenI've done the optimization on nu, but when I try to apply those parameters manually, changin nu doesn't seem to be changing the results.23:06
KMcQuistenAre the default values for tube_epsilon stepping on my manual setting of nu before training?23:07
KMcQuistenAnd if so how do I fix it?23:08
KMcQuistenHow does LibSVR determine if it's doing eps-SVR or nu-SVR?23:08
wikingmmm23:09
wiking<pass>23:09
wikingi don't know who would be an expert in libsvr here23:09
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has quit [Ping timeout: 245 seconds]23:30
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun23:39
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Ping timeout: 245 seconds]23:41
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has joined #shogun23:41
KMcQuistenWell damn.23:42
KMcQuistenWell, does anyone know why nu is available for optimization in LibSVR, but the tube epsilon isn't?  It would certainly be very handy to be able to use the built-in model selection fromework to find the best tube epsilon23:43
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Quit: Page closed]23:50
--- Log closed Wed Feb 13 00:00:45 2013
« Monday, 2013-02-11 Index Wednesday, 2013-02-13 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!