Open in new window / Try shogun cloud
--- Log opened Fri Jun 01 00:00:41 2012
-!- zxtx [] has quit [Ping timeout: 245 seconds]01:32
-!- blackburn [~blackburn@] has left #shogun []02:17
-!- av3ngr [] has joined #shogun03:08
-!- emrecelikten [~emrecelik@] has joined #shogun03:33
-!- emrecelikten [~emrecelik@] has quit [Client Quit]03:33
-!- av3ngr [] has quit [Ping timeout: 252 seconds]05:54
-!- av3ngr [~av3ngr@] has joined #shogun06:29
-!- av3ngr [~av3ngr@] has quit [Ping timeout: 246 seconds]07:20
-!- av3ngr [~av3ngr@] has joined #shogun07:33
-!- uricamic [~uricamic@2001:718:2:1634:e851:f250:a5d5:cb86] has joined #shogun08:48
-!- av3ngr [~av3ngr@] has quit [Quit: That's all folks!]09:31
-!- zxtx [] has joined #shogun09:32
-!- zxtx [] has quit [Ping timeout: 246 seconds]09:37
-!- zxtx [] has joined #shogun09:53
-!- gsomix [~gsomix@] has quit [Quit: Ex-Chat]10:16
-!- n4nd0 [] has joined #shogun10:33
-!- blackburn [~blackburn@] has joined #shogun11:08
-!- blackburn [~blackburn@] has quit [Ping timeout: 248 seconds]12:09
-!- nickon [] has joined #shogun12:27
-!- n4nd0 [] has quit [Quit: Lost terminal]12:57
-!- n4nd0 [] has joined #shogun12:58
-!- uricamic [~uricamic@2001:718:2:1634:e851:f250:a5d5:cb86] has quit [Quit: Leaving.]13:20
-!- pluskid [~pluskid@] has joined #shogun15:00
-!- n4nd0 [] has quit [Ping timeout: 244 seconds]15:14
-!- pluskid [~pluskid@] has quit [Ping timeout: 245 seconds]16:14
-!- pluskid [] has joined #shogun16:14
-!- eric_ [2e1fd566@gateway/web/freenode/ip.] has joined #shogun16:30
eric_hi all16:30
-!- pluskid [] has quit [Ping timeout: 250 seconds]16:51
-!- gsomix [~gsomix@] has joined #shogun16:55
-!- blackburn [~blackburn@] has joined #shogun17:01
-!- pluskid [~pluskid@] has joined #shogun17:09
-!- pluskid [~pluskid@] has quit [Client Quit]17:11
-!- n4nd0 [] has joined #shogun17:50
-!- eric_ [2e1fd566@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]18:15
blackburnn4nd0: hey how is it going?18:25
-!- nickon [] has quit [Quit: ( :: NoNameScript 4.22 :: )]18:34
n4nd0blackburn: ups sorry, I didn't hear it19:24
n4nd0blackburn: fine and for you?19:24
blackburnfine some thesis things to complete still - hope to switch to slep again19:28
n4nd0blackburn: good19:28
n4nd0I'll be back later!19:29
-!- n4nd0 [] has quit [Ping timeout: 252 seconds]19:35
-!- gsomix [~gsomix@] has quit [Ping timeout: 250 seconds]20:09
-!- gsomix [~gsomix@] has joined #shogun20:32
-!- n4nd0 [] has joined #shogun21:08
-!- gsomix [~gsomix@] has quit [Ping timeout: 245 seconds]21:15
-!- gsomix [~gsomix@] has joined #shogun21:19
shogun-buildbotbuild #256 of nightly_all is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #167 of nightly_default is complete: Success [build successful]  Build details are at
n4nd0blackburn: around?22:12
n4nd0blackburn: ok, I have got a doubt about the opt. problem I have to solve for SO22:13
n4nd0but it might be about something general22:13
n4nd0let me ask you in case you know it ok?22:13
blackburnjust ask but I do not promise I can help :)22:13
n4nd0yeah sure, I understand ;)22:14
blackburnmy opt. background is pretty lame22:14
n4nd0in that pdf, page 222:14
n4nd0algorithm 1, line 1122:14
n4nd0so there we have a quadratic program22:15
n4nd0so do you see the part with the sum22:15
n4nd0so my point is that in that sum there are no unknowns22:16
n4nd0since we are using the training data22:16
n4nd0that is the part indexed as x_i and y_i22:16
blackburnwhat is Psi?22:16
n4nd0and also results that we have previously stored in Gamma_i22:16
blackburnw/o ()22:16
n4nd0is an error22:17
n4nd0Psi should go there with ()22:17
n4nd0and it refers to the values we have stored in Gamma_i22:17
blackburnokay so what is your question?22:17
n4nd0ok wait a moment ...22:18
n4nd0I might have realized of it :P22:18
n4nd0all right22:19
n4nd0it was an stupid thing I have just realized22:19
n4nd0sorry :S22:19
blackburnheh :D22:19
blackburnthat's always happen when you try to describe something22:19
n4nd0haha yeah22:20
n4nd0I forgot about the w there22:20
n4nd0in the sum everything is not known22:20
n4nd0since we have precisely there the optimization vector22:20
* n4nd0 is afk22:25
blackburnhmm i do not understand much in that algorithm :D22:27
-!- CIA-113 [] has quit [Ping timeout: 245 seconds]22:36
-!- CIA-9 [] has joined #shogun22:46
CIA-9shogun: Sergey Lisitsyn master * rf100cae / src/shogun/features/StringFeatures.cpp : Added debug error instead of assert in StringFeatures -
-!- gsomix [~gsomix@] has quit [Ping timeout: 252 seconds]23:04
* n4nd0 is back23:12
n4nd0blackburn: tell me if you want to know something about the algorithm23:17
n4nd0blackburn: it'll probably help to my understanding trying to explain it :D23:17
blackburnok but a little bit later :)23:17
CIA-9shogun: Sergey Lisitsyn master * re915d01 / (2 files): Fixed GNB zero variance handling -
-!- heiko1 [] has joined #shogun23:46
heiko1blackburn, sonney2k, whats the best way to do matrix multiplication in shogun?23:48
blackburnheiko1: cblas_dgemm :)23:49
blackburndepends what is best for you23:49
heiko1is there a wrapper method available?23:49
heiko1just a plain multiplication23:49
blackburnwhat kind of wrapper do you need?23:49
heiko1in place would be nice23:49
heiko1sg matrix for example23:49
blackburnmatrix-matrix inplace??23:50
heiko1probably not when I think about it :)23:50
blackburnok I actually don't mind a wrapper23:50
heiko1oh yes, my matrices are square23:50
heiko1that would be possible in place23:50
heiko1anyway, dont really care23:51
blackburnI can't mind this23:51
blackburnyou would have to store matrix still23:51
heiko1is cblas_dgemm the method to use?23:51
blackburnI don't think inplace is possible at all23:51
blackburnwell yes23:52
heiko1its possible under certain condiutions (which are met in my case) but anyway23:52
heiko1i will try to do a general wrapper23:52
blackburnokay actually main reason why I do not use it23:52
blackburnis some variety of options like transposes23:53
blackburnand sizes23:53
blackburnactually one may end with wrapper with parameters exactly just like dgemm23:53
heiko1so you think directly using it would be better?23:53
heiko1I would love to have one for code reading ease23:54
blackburnno I don't mind wrappers23:54
heiko1kk then23:54
blackburnbut still you would need at least two parameters (transposes)23:54
heiko1ok I will include these23:54
blackburnask if you are stucked with params of blas23:54
heiko1ask :)23:55
heiko1where is an example call in SHOGUN that I can  use?23:55
blackburnlet me describe it for you there23:55
heiko1actually, seeing the blas reference23:56
heiko1I think I will directly use that23:56
blackburnheh ok23:56
heiko1you are right23:56
heiko1a wrapper would have to contain all this stuff :)23:56
blackburnthe only thing we can hide23:57
blackburnis LDA, LDB, LDC23:57
blackburnand N,M,K23:57
-!- heiko2 [] has joined #shogun23:59
-!- blackburn is now known as heiko323:59
--- Log closed Sat Jun 02 00:00:41 2012