--- Log opened Fri May 13 00:00:57 2011 | ||
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 00:15 | |
-!- blackburn [~qdrgsm@188.168.4.251] has quit [Quit: Leaving.] | 00:18 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 240 seconds] | 01:29 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 02:06 | |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 03:37 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 05:18 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 05:20 | |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 06:23 | |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 08:48 | |
-!- warpy [~warpy@bzq-79-183-44-64.red.bezeqint.net] has joined #shogun | 08:51 | |
CIA-31 | shogun: Soeren Sonnenburg master * r88994f7 / (71 files in 17 dirs): | 10:36 |
---|---|---|
CIA-31 | shogun: Introduce SGVector, SGMatrix, SGString, SGSparseMatrix data types. | 10:36 |
CIA-31 | shogun: And add a get_labels() function for SGVector testing. - http://bit.ly/lghujb | 10:36 |
CIA-31 | shogun: Heiko Strathmann master * r764d28d / src/libshogun/base/SGObject.cpp : remove redundant save/load for model selection parameters - http://bit.ly/kFh5AR | 10:45 |
CIA-31 | shogun: Heiko Strathmann master * rffeca12 / examples/undocumented/libshogun/parameter_set_from_parameters.cpp : added copyright - http://bit.ly/iMyXmr | 10:45 |
CIA-31 | shogun: Heiko Strathmann master * r5bc70da / examples/undocumented/libshogun/Makefile : added parameter iteration example - http://bit.ly/iiQFe2 | 10:45 |
CIA-31 | shogun: Soeren Sonnenburg master * ra7c1e48 / src/libshogun/features/Labels.h : introduce set_labels based on SGVector<float64_t> data type - http://bit.ly/lrrFU6 | 10:45 |
CIA-31 | shogun: Soeren Sonnenburg master * rec32295 / (6 files in 3 dirs): Merge branch 'master' of git://github.com/karlnapf/shogun - http://bit.ly/khzbZ5 | 10:45 |
@sonney2k | warpy, you wanted to ask sth? | 12:00 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 240 seconds] | 14:00 | |
CIA-31 | shogun: Soeren Sonnenburg master * r29dfea9 / (3 files in 3 dirs): add get/set vector in/out typemap for python - http://bit.ly/iTTFPA | 16:16 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 17:41 | |
-!- blackburn [~qdrgsm@188.168.3.72] has joined #shogun | 17:55 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 18:02 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 20:37 | |
@sonney2k | warpy, ? | 20:52 |
@sonney2k | blackburn, I did the SGVector thing for one examples - CLabels | 20:53 |
blackburn | sonney2k: aha, I have seen | 20:54 |
@sonney2k | blackburn, it seems to work from python already | 20:55 |
blackburn | nice! | 20:55 |
@sonney2k | I realized that we need this fro ndim arrays too | 20:56 |
blackburn | sonney2k: what do you think about implementing CLLE this way? | 20:56 |
@sonney2k | ehh how is this related? | 20:57 |
blackburn | I could use SGMatrix instead of float64_t :) | 20:58 |
@sonney2k | yes but algorithmwise there is no difference... | 21:01 |
blackburn | it could be more readable | 21:01 |
@sonney2k | hmmhh not much | 21:02 |
blackburn | ok | 21:03 |
@sonney2k | anyways I will try to do the python typemaps for matrix... | 21:03 |
warpy | what about c# ? :P | 21:03 |
blackburn | but the problem is if we will use *only* SGVector - we have to change internals for its usage | 21:04 |
blackburn | i.e. if get_feature_vector will return SGVector we would rather use it | 21:04 |
blackburn | I'm not sure I understand concept | 21:05 |
blackburn | are SG* structures purposed for interfaces? | 21:05 |
@sonney2k | warpy, well I won't do more except the draft - I was hoping you continue work on this... | 21:07 |
@sonney2k | blackburn, we have to change serialization code to only work with SGVector/Matrix etc and indeed all classes to internally use these things | 21:08 |
warpy | i was talking about the code daniel did | 21:10 |
warpy | didnt he send it to you | 21:10 |
warpy | the alpha version | 21:10 |
warpy | to put inside the git repo | 21:10 |
@sonney2k | warpy, did he finally get the example to run? | 21:10 |
warpy | i think so. | 21:10 |
blackburn | sonney2k: so would you like if I will change algos I implement(ed) for its usage? | 21:10 |
@sonney2k | warpy, I answered but he didn' treply | 21:10 |
warpy | he told me we can start using more than just the simple basic example | 21:11 |
warpy | because it compiled and returned "1" or something, correct ? | 21:11 |
@sonney2k | yes | 21:11 |
@sonney2k | blackburn, yes :) | 21:12 |
blackburn | sonney2k: okay | 21:13 |
warpy | was that yes for me ? | 21:17 |
warpy | ah yes i wanted to ask you | 21:17 |
warpy | how does swig/shogun deal with sparse matrices | 21:17 |
warpy | as far as we see the java code that we trnaslated to c# only deals with arrays and matrices. | 21:18 |
warpy | is there something special that we need to do to support sparse matrices ? | 21:19 |
@sonney2k | warpy, yes | 21:22 |
@sonney2k | warpy, is there any sparse matrix library in c#? | 21:24 |
@sonney2k | blackburn, the good news is that typemaps are drastically simplified | 21:52 |
@sonney2k | we now just need in and out typemaps | 21:53 |
@sonney2k | and no more %apply stuff | 21:53 |
blackburn | aha, I see, moreover it is more comprehensive | 21:53 |
@sonney2k | as soon as someone returns a SGVector or SGMatirx etc these things are swig wrapped | 21:53 |
@sonney2k | same for inputting | 21:53 |
@sonney2k | the only problem I have is that it is hard to do without breaking things at the moment | 21:54 |
@sonney2k | I think one would need to a) write all the new typemaps for all languages | 21:54 |
@sonney2k | b) add set/get functions to all relevant classes | 21:55 |
@sonney2k | (that use SGVector & friends) | 21:55 |
@sonney2k | c) add functions to do the serialization with SGVector etc | 21:55 |
blackburn | who it could be? :D | 21:55 |
@sonney2k | d) translate all classes to use SGVector internally | 21:55 |
@sonney2k | converting typemaps is piece of cake (I just did matrix for python) | 21:56 |
@sonney2k | adding set/get functions too | 21:56 |
@sonney2k | serialization is something heiko could help | 21:57 |
blackburn | aha, d) is for me :) | 21:57 |
@sonney2k | d) is potentially more demanding | 21:57 |
blackburn | anyway we don't need to rush d), right? | 21:58 |
blackburn | sonney2k: LLE for the moment works ~20s on 10000 | 21:58 |
blackburn | with no eigenproblem solving though | 21:58 |
@sonney2k | there is no rush for anything - but it will give us much more readable code | 22:00 |
blackburn | I mean I could do it for sure, but I haven't much time right now :( | 22:01 |
@sonney2k | blackburn, well you cannot start before a-c are done | 22:02 |
blackburn | -ac? | 22:02 |
@sonney2k | and believe me - you have much more time than I have currently... | 22:02 |
blackburn | a-c? | 22:02 |
@sonney2k | a)-c) | 22:02 |
blackburn | ah | 22:02 |
blackburn | sonney2k: hrr, looks like you think I'm relaxing :) | 22:05 |
@sonney2k | blackburn, no I don't think you are. | 22:06 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 22:06 | |
blackburn | sonney2k: SystemError: Out of memory error, tried to allocate 800000000 bytes using new[]. | 22:07 |
@sonney2k | We should postpone this discussion until you too have 2 babies at home :) | 22:07 |
blackburn | :D | 22:07 |
* sonney2k counts the zeros | 22:07 | |
blackburn | float64_t 10000*10000 matrix | 22:07 |
blackburn | I really need sparse matrix :D | 22:08 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 22:08 | |
@sonney2k | if the matrix is sparse then yes :) | 22:09 |
blackburn | sonney2k: 5.690s for 5000 examples | 22:09 |
blackburn | not so bad this time | 22:09 |
blackburn | but memory complexity is terrifying | 22:09 |
blackburn | in each row of that NxN matrix only k<<N values aren't zero | 22:10 |
* blackburn is in the crape, russia is beaten in 1/2 | 22:11 | |
@sonney2k | ok then sparse could work | 22:17 |
@sonney2k | but no idea if all the algos you need are available for sparse matrices | 22:17 |
blackburn | sonney2k: only eigenvalues | 22:18 |
@sonney2k | I never did that fro sparse matrices... | 22:18 |
blackburn | me too but I don't know any other way | 22:19 |
blackburn | sonney2k: nevermind, on last stage it is not sparse | 22:20 |
@sonney2k | heh | 22:20 |
@sonney2k | anyways I wouldn't optimize before you have a set of baseline algos | 22:21 |
blackburn | W is sparse matrix, then (I-W)'*(I-W) is computed and eigenvectors of it are feature vectors | 22:21 |
blackburn | (I-W)'*(I-W) is exactly don't sparse :) | 22:22 |
@sonney2k | good night! | 22:34 |
blackburn | good night.. | 22:38 |
blackburn | LLE is almost working now | 22:38 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 260 seconds] | 23:32 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 23:59 | |
--- Log closed Sat May 14 00:00:58 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!