--- Log opened Thu Feb 09 00:00:19 2012 | ||
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 00:08 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 02:39 | |
-!- dfrx [~f-x@inet-hqmc02-o.oracle.com] has joined #shogun | 05:46 | |
-!- n4nd0 [~androirc@s83-179-44-135.cust.tele2.se] has joined #shogun | 08:01 | |
-!- n4nd0 [~androirc@s83-179-44-135.cust.tele2.se] has quit [Remote host closed the connection] | 08:14 | |
-!- wiking [~wiking@78-22-115-59.access.telenet.be] has joined #shogun | 08:32 | |
-!- wiking [~wiking@78-22-115-59.access.telenet.be] has quit [Changing host] | 08:32 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:32 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 08:44 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:48 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 08:49 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:49 | |
-!- n4nd0 [~androirc@n151-p201.kthopen.kth.se] has joined #shogun | 09:21 | |
-!- n4nd0 [~androirc@n151-p201.kthopen.kth.se] has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )] | 09:54 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 09:59 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 10:30 | |
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Read error: Operation timed out] | 11:39 | |
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun | 11:42 | |
-!- dfrx [~f-x@inet-hqmc02-o.oracle.com] has quit [Quit: Leaving.] | 12:09 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 12:56 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 13:23 | |
-!- n4nd0 [~androirc@n151-p201.kthopen.kth.se] has joined #shogun | 13:25 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 13:45 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 13:46 | |
wiking | mm those are 2011 ideas that have not been implemented? | 14:31 |
---|---|---|
-!- axitkhurana [~akshit@180.149.49.227] has joined #shogun | 14:44 | |
-!- n4nd0 [~androirc@n151-p201.kthopen.kth.se] has quit [Read error: Connection reset by peer] | 14:50 | |
-!- akshit_ [~akshit@210.212.58.111] has joined #shogun | 14:58 | |
-!- axitkhurana [~akshit@180.149.49.227] has quit [Ping timeout: 248 seconds] | 15:02 | |
-!- iglesias [~androirc@n151-p201.kthopen.kth.se] has joined #shogun | 16:50 | |
-!- iglesias [~androirc@n151-p201.kthopen.kth.se] has quit [Ping timeout: 252 seconds] | 16:55 | |
wiking | why is that that when i create a kernel with the constructor with the left and right hand features shogun runs parallel, i.e. all 4 of my cores are used by shogun, while if i create the kernel with the default constructor and after that i call init function with the left and right-hand features, shogun runs only on one cpu | 17:01 |
wiking | talking here about multiclasssvm train/classification | 17:01 |
-!- akshit_ [~akshit@210.212.58.111] has quit [Ping timeout: 240 seconds] | 17:56 | |
-!- blackburn [~qdrgsm@83.234.54.5] has joined #shogun | 18:43 | |
blackburn | sonney2k: we've got Cheng Soon as Action Editor :) | 18:44 |
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has joined #shogun | 19:16 | |
wiking | blackburn: heeey | 19:18 |
wiking | blackburn: got another question... it's more about the internals of shogun... | 19:18 |
wiking | blackburn: why is that that when i create a kernel with the constructor with the left and right hand features shogun runs parallel, i.e. all 4 of my cores are used by shogun, while if i create the kernel with the default constructor and after that i call init function with the left and right-hand features, shogun runs only on one cpu | 19:19 |
n4nd0 | hi there | 19:30 |
n4nd0 | yesterday I installed shogun with support for the matlab static interface | 19:30 |
n4nd0 | but I get some troubles trying out some of the examples | 19:31 |
n4nd0 | do I have to install matlab_modular or any other for them to work or is it another problem? | 19:31 |
n4nd0 | has someone installed before cplex in ubuntu? | 19:56 |
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 252 seconds] | 20:44 | |
blackburn | wiking: no idea, will have to check later | 20:53 |
wiking | mmm | 20:53 |
blackburn | wiking: but you can use kernel.parallel.set_num_threads(4) | 20:53 |
wiking | tried | 20:53 |
wiking | no help | 20:53 |
blackburn | to set it explicitly | 20:53 |
wiking | it seems there's something else | 20:54 |
wiking | because during training time it's always sequential, i.e. never use threads | 20:54 |
wiking | but for some reason in case of histogramintersectionkernel there's a different speed depending on which constructor i'm using | 20:54 |
wiking | although i know it doesn't make much sense, but this is what i get :S | 20:54 |
blackburn | what exactly do you do? | 20:54 |
blackburn | funny I use HIK too now :) | 20:55 |
wiking | hehehe | 20:55 |
wiking | well i'm trying to test my features with different kernels | 20:55 |
wiking | so my usual thing is to create a kernel and then do the init | 20:56 |
wiking | i mean do it later the init, after i've created the right kernel | 20:57 |
wiking | and so with HIK it's weird | 20:57 |
wiking | with poly kernel it's fine for instance | 20:57 |
blackburn | well there is no difference between ^ and ^ | 20:57 |
blackburn | I mean CKernel class is parallelized | 20:58 |
blackburn | not particular ones | 20:58 |
wiking | yeah | 20:58 |
wiking | i've checked the c++ code | 20:58 |
wiking | so it's really weird | 20:58 |
wiking | but now it's not even that | 20:58 |
wiking | because as far as i can see now | 20:58 |
wiking | the training time is never parallel | 20:58 |
wiking | so it's always running on one cpu | 20:59 |
wiking | but somehow when i use one or the other way of constructing the HIK kernel | 20:59 |
wiking | the training time is very different | 20:59 |
wiking | so let's say one way is like 10-20 seconds | 20:59 |
wiking | the other way it's at least 2-3 minutes | 20:59 |
blackburn | could u please paste the code somewhere? | 20:59 |
wiking | ok sure | 20:59 |
wiking | letme pastebin it | 21:00 |
wiking | but note that i'm using java api | 21:00 |
blackburn | well should not make any difference | 21:00 |
blackburn | but who knows :D | 21:00 |
wiking | ok | 21:03 |
wiking | here's the snippet of the code: http://pastebin.com/2pGUjKw4 | 21:03 |
wiking | and depending on line 35 boolean you'll get slow or 'norma'/fast training time | 21:04 |
blackburn | hmm I see | 21:05 |
blackburn | the same cache size! | 21:05 |
blackburn | strange it could act in a different way | 21:06 |
wiking | yeah | 21:06 |
wiking | so i've tried to set everything the same | 21:06 |
wiking | just the way i construct the kernel is different | 21:06 |
wiking | i've checked the cpp code | 21:06 |
wiking | i cannot see the difference | 21:06 |
blackburn | me neither | 21:07 |
wiking | so now the only thing is somewhere in the java modular stuff | 21:07 |
wiking | something happens | 21:07 |
wiking | i've tried to turn on debugging mode | 21:07 |
blackburn | no, can't be true | 21:07 |
wiking | but i couldn't get there | 21:07 |
wiking | for some reason the inheritance thingy is not working in java | 21:08 |
wiking | i.e. i cannot call kernel.parallel.get_num_cpus () | 21:08 |
blackburn | really? | 21:08 |
blackburn | bad,bad | 21:08 |
wiking | i have to do first a kernel.get_global_parallel | 21:08 |
wiking | and then do the thingy there | 21:08 |
blackburn | do you think it is parallel issue? | 21:09 |
wiking | nah not really | 21:09 |
wiking | as said | 21:09 |
wiking | i've tried to monitor the number of threads | 21:09 |
wiking | it's always 4, i.e. as many core i have | 21:09 |
wiking | and in train time i never get more than 1 threads running | 21:10 |
blackburn | LaRank is not parallelized so it is ok | 21:10 |
wiking | so it's not really the parallelization problem | 21:10 |
wiking | it's more like, why does it run slower | 21:10 |
blackburn | in fact we have no svm with parallel training | 21:10 |
blackburn | that's really strange | 21:11 |
blackburn | sorry I have to go now, but I'll check a little bit later | 21:11 |
wiking | i think some flag is not turned on in some if branch | 21:11 |
wiking | but i'll try to do first the same thing with a silly | 21:11 |
wiking | c++ code | 21:11 |
wiking | if i can get the same thing | 21:12 |
wiking | ok no worries thnx! | 21:12 |
blackburn | are you subscribed to the mailing list? | 21:12 |
blackburn | I would suggest you to create github issue or write to the mailing list | 21:12 |
blackburn | if you want | 21:12 |
wiking | ah ok i'll do the github issue then | 21:16 |
blackburn | thanks | 21:16 |
wiking | nw | 21:16 |
blackburn | it would be easier then to take care :) | 21:17 |
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has joined #shogun | 21:25 | |
blackburn | n4nd0: what is the problem with matlab? | 21:30 |
n4nd0 | blackburn, so there were a couple of issues | 21:30 |
n4nd0 | blackburn, the one that I was concerned about was that the lpboost example does not run | 21:31 |
n4nd0 | blackburn, it says sth like classifier LPBOOST not found | 21:31 |
blackburn | n4nd0: matlab static? | 21:32 |
n4nd0 | blackburn, I have looked it up on google and it seems that cplex is required for it to work so I am with that right now | 21:32 |
n4nd0 | blackburn, yes, I am using matlab static this far | 21:32 |
blackburn | ah yes | 21:32 |
n4nd0 | blackburn, should it be an issue static or modular just to try the example out? | 21:32 |
n4nd0 | blackburn, I found that example under matlab_static so I assumed that it should work there also | 21:32 |
n4nd0 | there too* | 21:33 |
blackburn | well should work indeed | 21:33 |
blackburn | but cplex is required I guess | 21:33 |
blackburn | lemme check | 21:33 |
n4nd0 | let's see then if I managed to get cplex to work | 21:33 |
n4nd0 | yeah I found that in this forum | 21:34 |
blackburn | you could check it by yourself too :) just take a look on classifier/LPBoost.h | 21:34 |
blackburn | there is a #ifdef HAVE_CPLEX | 21:34 |
blackburn | so just install cplex and ./configure, etc | 21:34 |
n4nd0 | didn't think of looking it up in the code :-O | 21:35 |
n4nd0 | the other problem I found is that all the data directories are empty | 21:36 |
n4nd0 | I made first a defauult installation: ./configure, make, make install | 21:36 |
n4nd0 | and later another one with ./configure interfaces=matlab_static | 21:37 |
n4nd0 | but those directories are empty in any case | 21:37 |
blackburn | ah | 21:39 |
blackburn | do | 21:39 |
blackburn | git submodule update --init | 21:39 |
blackburn | IIRC | 21:39 |
blackburn | or git submodule --init update | 21:39 |
blackburn | that's a typical problem for sure | 21:40 |
blackburn | it is because shogun and shogun-data are two different repos | 21:40 |
n4nd0 | mmm I installed it from the compressed directory so I must first activate the repo | 21:41 |
blackburn | uh oh | 21:41 |
blackburn | then I would suggest you to git clone shogun :) | 21:41 |
blackburn | no idea how to activate it from compressed one | 21:41 |
n4nd0 | all right | 21:42 |
n4nd0 | I will take a look to that once I am done with the cplex issue | 21:42 |
n4nd0 | thank you very much :D | 21:42 |
blackburn | you're welcome :) | 21:43 |
blackburn | wiking: did you get any progress with this strange issue? :) | 21:43 |
wiking | blackburn: none yet... postponed it... now i'm using the lil hack which is faster... since i would really like to get to the point where i can try out MKL on multi class :) that was the main reason i've started to write my shogun wrapper in my code... ;) | 21:44 |
blackburn | wiking: okay sure | 21:44 |
wiking | aaha coool | 21:45 |
wiking | i've just realized that i can iterate through the features i've appended to the CombinedFeatures | 21:46 |
blackburn | I feel butthurt remembering that there was a leak in java interface :D | 21:46 |
blackburn | it was reproduced with HUGE mkl script | 21:46 |
wiking | uff | 21:47 |
blackburn | okay heading to bed right now | 21:47 |
blackburn | see you | 21:47 |
wiking | hahahah | 21:47 |
wiking | let's see what happens | 21:47 |
wiking | i really wouldn't like to rewrite now the whole thing in python or c++ :> | 21:48 |
wiking | anyhow gnite | 21:48 |
blackburn | nooo | 21:48 |
blackburn | bug is not here anymore | 21:48 |
blackburn | I hope | 21:48 |
blackburn | it was CustomKernel issue | 21:48 |
blackburn | but consider using python anyway, it is much cooler for shoguning :) | 21:48 |
wiking | hahahahah | 21:49 |
wiking | i'm using mahout as well | 21:49 |
wiking | so i need a common ground :> | 21:49 |
blackburn | I've never used it | 21:49 |
wiking | it's cool | 21:49 |
blackburn | ah yeah makes sense | 21:49 |
wiking | but it's mainly developed for text processing | 21:49 |
blackburn | is it for distributed things? | 21:49 |
wiking | yep | 21:50 |
wiking | based on hadoop | 21:50 |
wiking | but for me that's just for acquiring the features | 21:50 |
blackburn | I see | 21:50 |
wiking | so it's good otherwise with shogun | 21:50 |
blackburn | but they don't have many svms :D | 21:50 |
wiking | until now i was using weka :> | 21:50 |
wiking | but it's kind of like sucks as well with their limited kernels... | 21:51 |
blackburn | what do you mean saying limited? | 21:51 |
wiking | well like i really wanted HIK | 21:51 |
wiking | and now i really hope i'm getting some boost with MKL | 21:51 |
blackburn | with which features do you use HIK? | 21:51 |
wiking | ah i'm doing image classification | 21:51 |
blackburn | color histograms? | 21:52 |
wiking | sift histogram | 21:52 |
wiking | so basically bag of visual words... | 21:52 |
blackburn | I used to do similar thing | 21:52 |
wiking | ah and what's now? :) | 21:52 |
blackburn | actually I'm going to implement some IKSVM a little bit later | 21:52 |
wiking | mmm paper? | 21:53 |
blackburn | http://acberg.com/papers/mbm08cvpr.pdf | 21:53 |
blackburn | this one | 21:53 |
wiking | ash a cvpr one :) | 21:54 |
wiking | cool i'll give it a read | 21:54 |
wiking | btw | 21:54 |
wiking | any plan for latent svm | 21:54 |
wiking | ? | 21:54 |
blackburn | not really, I don't know much about latent svms yet | 21:56 |
blackburn | I used to implement dim reduction things so I'm pretty new to svms still hah | 21:56 |
blackburn | okay now going to sleep really :) | 21:56 |
blackburn | see you later | 21:56 |
wiking | laterz | 21:57 |
blackburn | we should exchange experience later :) | 21:57 |
-!- blackburn [~qdrgsm@83.234.54.5] has quit [Quit: Leaving.] | 21:57 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun | 22:19 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds] | 22:23 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has quit [Ping timeout: 248 seconds] | 22:24 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 22:31 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit] | 22:31 | |
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 265 seconds] | 22:58 | |
-!- n4nd0 [~n4nd0@s83-179-44-135.cust.tele2.se] has joined #shogun | 23:10 | |
--- Log closed Fri Feb 10 00:00:19 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!