--- Log opened Mon May 09 00:00:50 2011 | ||
blackburn | I see | 00:01 |
---|---|---|
blackburn | sonney2k: do you plan to change internals for SGVector, etc usage? | 00:15 |
-!- blackburn [~qdrgsm@188.122.225.203] has quit [Quit: Leaving.] | 01:29 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 05:09 | |
-!- blackburn [~qdrgsm@188.122.225.203] has joined #shogun | 10:14 | |
@sonney2k | blackburn, would love to | 10:31 |
@sonney2k | not sure if I have time to do it now. | 10:31 |
blackburn | me too | 10:31 |
blackburn | but I think we could do it during summer | 10:32 |
@sonney2k | maybe a first step is to put the struct definition in DataType.h | 10:32 |
blackburn | I think we should make definition and then convert one algo to its usage | 10:32 |
@sonney2k | there is one problem though | 10:32 |
blackburn | which? | 10:33 |
@sonney2k | typemaps need to be adjusted | 10:33 |
blackburn | ah, yeap | 10:33 |
@sonney2k | so that means we have to add another round of typemaps | 10:33 |
@sonney2k | (to not break anything inbetween) | 10:33 |
@sonney2k | and then gradually remove the others | 10:33 |
blackburn | so well when I will have holidays (mid of june - .. ) | 10:33 |
blackburn | may be I will do it for GaussianNaiveBayes | 10:34 |
blackburn | with typemaps, etc | 10:34 |
blackburn | if everything will work fine - we will convert others gradually | 10:34 |
blackburn | sonney2k: can you recommend some BLAS/ATLAS tutorial? | 10:37 |
@sonney2k | blackburn, it may actually be simpler to do the typemaps... | 10:37 |
@sonney2k | with SGVector etc | 10:38 |
@sonney2k | blackburn, I only ever used this lapack / blas overview pdf that you find in the net | 10:38 |
@sonney2k | there is a book though | 10:38 |
blackburn | ok | 10:39 |
@sonney2k | blackburn, I am currently finishing the malloc / realloc / free -> SG_MALLOC etc transition | 10:43 |
blackburn | where? | 10:43 |
blackburn | sonney2k: I need to use atlas for solving systems of linear equations, right? | 10:51 |
@sonney2k | yes | 10:51 |
blackburn | and cblas for other routines, like dot product | 10:51 |
@sonney2k | otherwise it will be tough | 10:51 |
blackburn | sonney2k: do we have example of atlas usage in shogun? I got stucked with setting it up | 10:52 |
@sonney2k | blackburn, PCACut.cpp or dot product in lib/Mathematics.h or in pr_loqo.cpp | 10:53 |
blackburn | seems PCACut use only blas.. | 10:54 |
blackburn | same thing in Mathematics.h | 10:56 |
@sonney2k | compute_eigenvectors in Math uses dsyev | 11:02 |
@sonney2k | that is eigenvectors/values for symmetric doubl ematrix | 11:02 |
blackburn | ah! | 11:03 |
blackburn | thank you | 11:03 |
blackburn | sonney2k: what is wrap_**? | 11:06 |
blackburn | ah, I see | 11:07 |
blackburn | it is in lapack.h | 11:07 |
blackburn | sonney2k: the problem was related to my bad atlas installation | 11:12 |
@sonney2k | blackburn, yeah there are problems with when things are available in atlas or lapack only and different os / libs make a difference | 11:12 |
@sonney2k | ok | 11:12 |
blackburn | sonney2k: question now is should I add needed functions to lib/lapack.h or can I use it straight? | 11:12 |
@sonney2k | blackburn, preferably add it to lapack.h (wrap_*) and then use it | 11:13 |
@sonney2k | in case one has to change it later it will be easier to modify it there | 11:13 |
blackburn | aha! I see | 11:14 |
blackburn | thanks! you helped me very much | 11:14 |
-!- [1]warpy [~warpy@bzq-79-183-44-64.red.bezeqint.net] has joined #shogun | 12:50 | |
-!- [1]warpy [~warpy@bzq-79-183-44-64.red.bezeqint.net] has quit [Ping timeout: 260 seconds] | 13:01 | |
CIA-90 | shogun: Soeren Sonnenburg master * r7890ec2 / (10 files): remove roc eval functions from static interfaces - http://bit.ly/mNbmUQ | 13:07 |
CIA-90 | shogun: Soeren Sonnenburg master * r6e92985 / (59 files in 13 dirs): Convert all memory functions to use the SG_* ones. - http://bit.ly/mbsY2g | 13:07 |
CIA-90 | shogun: Soeren Sonnenburg master * ra861f66 / examples/undocumented/matlab_and_octave/classifier_libsvmmulticlass.m : fix multiclass example - http://bit.ly/lXAmmo | 13:07 |
CIA-90 | shogun: Soeren Sonnenburg master * r84e585e / (src/libshogun/lib/memory.cpp src/libshogun/lib/memory.h): | 13:07 |
CIA-90 | shogun: Introduce SG_MALLOC, SG_FREE, SG_REALLOC. | 13:07 |
CIA-90 | shogun: These replacements to the stdlib functions will throw exceptions in case | 13:07 |
CIA-90 | shogun: of errors (never return NULL). - http://bit.ly/mUySQo | 13:07 |
blackburn | sonney2k: what is the purpose of SG_MALLOC, SG_FREE, SG_REALLOC? | 13:17 |
blackburn | ah I see | 13:18 |
blackburn | in new code we should use SG_MALLOC, right? | 13:18 |
@sonney2k | blackburn, usually new[] new() | 14:03 |
@sonney2k | but never malloc / realloc / free but the SG_* things instead | 14:03 |
blackburn | I see | 14:04 |
@sonney2k | it just has the advantage that you never need to check for NULL ptr's | 14:05 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 14:48 | |
-!- blackburn [~qdrgsm@188.122.225.203] has quit [Quit: Leaving.] | 15:35 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 16:35 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 17:58 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 18:00 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 18:01 | |
-!- [1]warpy [~warpy@bzq-79-183-44-64.red.bezeqint.net] has joined #shogun | 18:15 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 18:41 | |
-!- [1]warpy [~warpy@bzq-79-183-44-64.red.bezeqint.net] has quit [Ping timeout: 240 seconds] | 19:36 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 20:14 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 20:16 | |
-!- Daniel [50fa9238@gateway/web/freenode/ip.80.250.146.56] has joined #shogun | 20:40 | |
Daniel | Hi Sonney | 20:40 |
Daniel | Sonney2K? | 20:41 |
Daniel | You had told me to use ./configure --interfaces=libshogun,csharp to compile the csharp typemap when it was ready | 20:41 |
Daniel | but I got "Unknown interface: csharp" | 20:42 |
Daniel | this was on the code where you had added the csharp folder | 20:43 |
Daniel | I emailed you | 20:48 |
Daniel | take care | 20:48 |
-!- Daniel [50fa9238@gateway/web/freenode/ip.80.250.146.56] has quit [Quit: Page closed] | 20:48 | |
-!- blackburn [~qdrgsm@188.168.4.172] has joined #shogun | 21:02 | |
@sonney2k | Daniel no I told you to use csharp_modular | 21:24 |
@sonney2k | sigh too late | 21:29 |
blackburn | what's up, sonney2k? | 21:29 |
@sonney2k | blackburn, please have a look at the datatype | 21:30 |
CIA-90 | shogun: Soeren Sonnenburg master * ra996050 / src/libshogun/lib/DataType.h : draft definition for vector / matrix datatype - http://bit.ly/mOSESQ | 21:30 |
@sonney2k | blackburn, I think we should rename things to SGVector SGMatrix, SGSparseVector, SGSparseMatrix, SGString | 21:30 |
blackburn | of course, I don't like T* | 21:30 |
@sonney2k | blackburn, but have a look first | 21:31 |
blackburn | already | 21:31 |
blackburn | sonney2k: what you think about implementing operations for it using blas, for example dot product | 21:32 |
blackburn | do you* | 21:33 |
blackburn | ehhrr my english is becoming worse every day :D | 21:33 |
@sonney2k | blackburn, makes sense - but we shouldn't over do it - there are many matrix/vector packages around that probably do a better job then us | 21:35 |
blackburn | anyway it is better way | 21:36 |
blackburn | because we can change all the things just changing implementation of sgvector | 21:36 |
@sonney2k | on the other hand it is very convenient and more clean - we could just implement the functions that are needed anyways | 21:36 |
@sonney2k | it will probably make some functions in *features and *math* obsolete | 21:37 |
blackburn | sonney2k: what about +, * overloading? I think it is a bad practice | 21:38 |
@sonney2k | lets not do that now | 21:39 |
blackburn | it looks undetermined | 21:39 |
blackburn | sonney2k: do we really need template matrix type? | 21:53 |
blackburn | I mean almost all of them are float | 21:54 |
@sonney2k | blackburn, almost all yes - but not e.g. images | 22:02 |
blackburn | may be we could have some shorthand name | 22:02 |
blackburn | SGFloatMatrix or so | 22:02 |
blackburn | ohhh overloading [] for matrices could be goooood | 22:06 |
blackburn | z_transposed[dim*j+d] -= feature_vector[d]; | 22:11 |
blackburn | pfff | 22:11 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 22:30 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 22:32 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 22:34 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 22:36 | |
@sonney2k | blackburn, overloading [] for 2d is not possible - | 22:38 |
blackburn | ah.. yeah, forgot | 22:38 |
@sonney2k | I guess you want sth like set / get functions | 22:38 |
blackburn | yeap because that X*Y+Z is pretty unreadable if it is not your own code | 22:39 |
@sonney2k | and it can probably be done more efficiently using some matrix lib like eigen anyways | 22:39 |
@sonney2k | btw having SGFloatmatrix etc is not really necessary - typemaps can have type specific implementations if needed | 22:40 |
blackburn | ok | 22:41 |
blackburn | I don't know which is most efficient, but we should use that one :) | 22:41 |
@bettyboo | interesting | 22:41 |
@sonney2k | going to bed now | 22:47 |
@sonney2k | l8r | 22:47 |
blackburn | good night | 22:48 |
@sonney2k | the mighty diaper master leaves the room | 22:48 |
blackburn | :D | 22:50 |
--- Log closed Tue May 10 00:00:52 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!