--- Log opened Mon Jun 13 00:00:21 2011 | ||
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 00:00 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 246 seconds] | 00:04 | |
@sonney2k | i can imagine... | 00:13 |
---|---|---|
-!- in3xes_ is now known as in3xes | 00:15 | |
serialhex | the insanity that some people code... though this is kinda interesting: http://github-high-scores.heroku.com/shogun-toolbox/shogun/high_scores/ | 04:59 |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 06:02 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 276 seconds] | 06:06 | |
-!- in3xes_ is now known as in3xes | 06:17 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 260 seconds] | 07:15 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 07:27 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Ping timeout: 244 seconds] | 09:41 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has joined #shogun | 09:58 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Ping timeout: 240 seconds] | 10:04 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has joined #shogun | 10:17 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Ping timeout: 276 seconds] | 10:29 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has joined #shogun | 10:47 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Ping timeout: 240 seconds] | 10:59 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has joined #shogun | 11:06 | |
CIA-18 | shogun: Soeren Sonnenburg master * r78b5538 / src/.generate_link_dependencies.py : | 11:18 |
CIA-18 | shogun: test if doxygen is installed and only then add the *_doxygen.i files as | 11:18 |
CIA-18 | shogun: deps - http://bit.ly/lmg2Fb | 11:18 |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Ping timeout: 246 seconds] | 11:28 | |
CIA-18 | shogun: Soeren Sonnenburg master * r9ec240c / examples/undocumented/java_modular/ClassifierLibsvmMinimalModular.java : modified example to work with jblas DoubleMatrix - http://bit.ly/m38Zt5 | 11:37 |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has joined #shogun | 11:40 | |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 15:07 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 15:11 | |
-!- alesis-novik [~alesis@78-56-236-240.static.zebra.lt] has quit [Quit: Leaving] | 16:29 | |
-!- blackburn [~blackburn@188.122.224.102] has joined #shogun | 17:16 | |
blackburn | http://dl.dropbox.com/u/10139213/share/image.png | 17:26 |
blackburn | mds faces example ^ :) | 17:27 |
-!- in3xes_ is now known as in3xes | 17:47 | |
@sonney2k | blackburn, heh :) | 19:53 |
@bettyboo | <:*) | 19:53 |
blackburn | sonney2k: that picture rocks http://dl.dropbox.com/u/10139213/share/image2.png | 19:53 |
blackburn | unfortunately I can't integrate it into shogun | 19:54 |
blackburn | but I have an idea | 19:54 |
blackburn | going to make my own dataset I can integrate | 19:54 |
blackburn | sonney2k: is it ok for you? | 19:54 |
@sonney2k | no not ok | 19:55 |
@sonney2k | fantastic! | 19:55 |
blackburn | sonney2k: I think I'll make photos of my girlfriend :D | 19:55 |
blackburn | just as these ones | 19:55 |
@sonney2k | blackburn, she will probably be very pleased | 19:56 |
blackburn | sonney2k: I hope so :) | 19:56 |
blackburn | let's try isomap on these pics | 19:57 |
@sonney2k | blackburn, what similarity measure are you using? | 19:58 |
blackburn | just euclidian distance between matrices :) | 19:59 |
@sonney2k | impressive results... | 19:59 |
blackburn | in a minute we will see what isomap makes for this dataset | 20:00 |
blackburn | and after I'll try LLE | 20:00 |
blackburn | oh isomap made it pretty similar | 20:02 |
blackburn | lle turn | 20:02 |
-!- f-x [~user@117.192.221.133] has joined #shogun | 20:04 | |
blackburn | sonney2k: which version of matplotlib do we require for shogun? | 20:05 |
@sonney2k | blackburn, at least 0.99.3 I would say | 20:05 |
blackburn | Idk exactly but now I'm using some thing (OffsetImage) that isn't in 0.99 | 20:05 |
blackburn | at least it wasn't in default ubuntu repo matplotlib | 20:06 |
blackburn | I had to install 1.0.1 manually | 20:06 |
blackburn | the reason why - it seems to be the only way to place images on graph plot | 20:06 |
blackburn | wow, LLE had segfault | 20:07 |
blackburn | bad bad | 20:07 |
@sonney2k | blackburn, well you could do a test in your example for matplotlib 1.0.x and then complain if it is not there | 20:09 |
blackburn | ah okay | 20:10 |
@sonney2k | doesn't matter too much for this example I would say | 20:10 |
blackburn | sad news - LocallyLinearEmbedding is bad now | 20:11 |
blackburn | going to debug | 20:11 |
@sonney2k | f-x, any news regarding the benchmarking? | 20:13 |
f-x | sonney2k: using fread(), should i try to read the whole file in one go? | 20:14 |
@sonney2k | well with sth. like filesize == 3times your memory | 20:15 |
@sonney2k | it will be hard | 20:15 |
@sonney2k | just read huge chunks | 20:15 |
@sonney2k | until eof | 20:15 |
blackburn | haha found error | 20:16 |
f-x | i think that can be done with a small modification in the io_buf code from vw | 20:16 |
@bettyboo | grin | 20:16 |
f-x | just changing the default buffer size from 64 KB to something like 10 MB | 20:16 |
blackburn | sonney2k: I forgot that get_feature_vector *creates* new vector | 20:16 |
blackburn | sonney2k: so with 4000d examples it wastes memory a little ;D | 20:17 |
@sonney2k | blackburn, wait get_feature_vector creates new vector? | 20:23 |
blackburn | yes | 20:23 |
@sonney2k | well it doesn't necessarily | 20:23 |
blackburn | sonney2k: I just used feature_matrix | 20:24 |
blackburn | with SGMatrixâ„¢ | 20:24 |
blackburn | so problem disapperead | 20:24 |
blackburn | tfu | 20:24 |
@sonney2k | blackburn, yeah but then it doesn't create a new vector / matrix etc... | 20:24 |
blackburn | with SGMatrix no | 20:24 |
blackburn | but with get_feature_vector - yes | 20:24 |
@sonney2k | ahh | 20:24 |
@sonney2k | I understand | 20:24 |
blackburn | sonney2k: | 20:25 |
blackburn | *dst=(ST*) SG_MALLOC(vlen*sizeof(ST)); | 20:25 |
blackburn | memcpy(*dst, vec, vlen*sizeof(ST)); | 20:25 |
blackburn | that's how it does | 20:25 |
@sonney2k | that code will die soon | 20:25 |
@sonney2k | at least for python I've converted all the typemaps | 20:25 |
blackburn | SimpleFeatures.h:261 | 20:25 |
blackburn | I really like new SGMatrix | 20:25 |
blackburn | I'm now using it where possible | 20:26 |
@sonney2k | now I need to do it for octave / R and then I can start replacing these copy crap things | 20:26 |
@sonney2k | double** stuff | 20:26 |
blackburn | sonney2k: the only thing - I don't use it if I need some matrices to multiply, subtract and so on | 20:26 |
blackburn | what do you think about it? should sgmatrix be used everywhere where matrices are used? | 20:27 |
blackburn | segfault again! damn | 20:28 |
@sonney2k | blackburn, it makes sense to go that route... | 20:29 |
blackburn | sonney2k: with sth like SGMatrix.dot(A,B) and so on? | 20:29 |
@sonney2k | multiply | 20:30 |
@sonney2k | yes | 20:30 |
blackburn | I see | 20:30 |
@sonney2k | A.multiply(B) | 20:30 |
@sonney2k | but only the things that you need - we don't want to compete with the many matrix / vector libraries here | 20:31 |
@sonney2k | sigh | 20:31 |
@sonney2k | sploving is not here and I haven't seen him for a long time | 20:31 |
blackburn | sonney2k: I just want ARPACK :) | 20:31 |
@sonney2k | as you wish... | 20:32 |
blackburn | nothing more :) | 20:32 |
blackburn | because Lanzcos is very crucial | 20:32 |
@sonney2k | just don't forget you have another exam | 20:32 |
blackburn | ah yeap | 20:32 |
blackburn | but now I want to find out why LLE faults | 20:33 |
blackburn | oh | 20:36 |
blackburn | I guess I was drunk writing it | 20:36 |
f-x | sonney2k: which functions are you looking to have reimplemented in StreamingAsciiFile? the get_*_vector() functions? | 20:36 |
* sonney2k compiles kernel 3.0rc2 | 20:37 | |
blackburn | aren't you like stable? :) | 20:37 |
@sonney2k | blackburn, I though Russians are non-operational without their Vodka ;-) | 20:39 |
blackburn | yeah something like that :) | 20:40 |
blackburn | there are so many matrices that I use gnote for checking if I not delete some | 20:41 |
@sonney2k | blackburn - but Linus' kernels are always stable | 20:43 |
@sonney2k | :) | 20:44 |
blackburn | exactly but I mean I thought you use old&stable soft :) | 20:44 |
@sonney2k | blackburn, the desktop environment is much more complex than the kernel | 20:46 |
@sonney2k | f-x, yeah that are the ones we currently need | 20:47 |
f-x | sonney2k: okay, so i guess i don't need to change much in the code which is currently StreamingFile.cpp | 20:47 |
@sonney2k | f-x, well we need efficient function though | 20:48 |
@sonney2k | f-x so your buffered line reader is highly needed | 20:48 |
f-x | sonney2k: that will be used, though i cannot think of something more efficient than the readto() from VW for now | 20:50 |
blackburn | hehe N*m_k -> N*dim | 20:52 |
blackburn | oh I finally fixed it | 20:58 |
blackburn | sonney2k: please merge pull request with fixes :) | 21:03 |
blackburn | it is bug-free now hehe | 21:03 |
-!- f-x [~user@117.192.221.133] has quit [Remote host closed the connection] | 22:12 | |
@sonney2k | Re | 22:26 |
@sonney2k | blackburn, I think it would be more readable to define some float64_t* matrix=feature_matrix.matrix | 22:34 |
@sonney2k | and then use this | 22:34 |
blackburn | why? | 22:34 |
blackburn | it is used only twice | 22:34 |
blackburn | sonney2k: well then if we will have 3-4 matrices it doesn't make sense do it this way | 22:37 |
blackburn | so I think it is more readable | 22:38 |
@sonney2k | I understand you have many matrices :) | 22:41 |
@sonney2k | I am just not sure if feature_matrix.matrix[neighborhood_matrix[j*N+i]*dim+k]; is optimized | 22:42 |
blackburn | why not? | 22:43 |
blackburn | we should check because in the case it is not optimized SGMatrices shouldn't be used internally | 22:44 |
blackburn | sonney2k: how can I check it? | 22:46 |
@sonney2k | well you could just measure the speed | 22:46 |
@sonney2k | well SGMatrices can of course be used | 22:47 |
@sonney2k | but the dereferencing feature_matrix.matrix could be costly | 22:47 |
@sonney2k | I am just not sure | 22:47 |
blackburn | okay will check it in a minute | 22:47 |
@sonney2k | http://www.fefe.de/know-your-compiler.pdf does only mention an aliasing problem | 22:49 |
@sonney2k | thanks | 22:50 |
blackburn | sonney2k: how can I get assembly code? | 22:51 |
@sonney2k | blackburn, I would just measure speed | 22:51 |
blackburn | yes I would rather too but just curious | 22:51 |
blackburn | you have experience I guess :) | 22:52 |
@sonney2k | blackburn, http://www.delorie.com/djgpp/v2faq/faq8_20.html | 22:52 |
-!- blackburn [~blackburn@188.122.224.102] has quit [Quit: Leaving.] | 22:52 | |
-!- blackburn [~blackburn@188.122.224.102] has joined #shogun | 22:53 | |
* sonney2k reboots into the new kernel... | 22:56 | |
@sonney2k | brb *hopefully* | 22:56 |
blackburn | sonney2k: making shogun + testing lle + music playing + etc = fire | 22:56 |
blackburn | hehe | 22:56 |
blackburn | good luck | 22:56 |
blackburn | fire on your applet :) | 22:56 |
blackburn | sonney2k: clock() is unreliable a bit, can you suggest me a better way to measure time? googling didn't help | 23:15 |
blackburn | well with float64_t* - average is 60000 clocks | 23:19 |
blackburn | same without | 23:20 |
blackburn | have to check assembly | 23:25 |
@sonney2k | blackburn, epic fail | 23:28 |
blackburn | what's up? | 23:28 |
@sonney2k | I am now back to a version before the version I used to run before | 23:28 |
@sonney2k | nvidia drivers don't understand 3.0.x | 23:29 |
blackburn | ah | 23:29 |
CIA-18 | shogun: Sergey Lisitsyn master * r9552aef / src/libshogun/preprocessor/LocallyLinearEmbedding.cpp : Crucial fixes for drunk-made mistakes in LocallyLinearEmbedding - http://bit.ly/koZjxa | 23:32 |
blackburn | sonney2k: I haven't ever thought that gcc makes so sophisticated assembly code | 23:32 |
blackburn | sonney2k: about time measure - any ideas? | 23:32 |
@sonney2k | blackburn, use shogun's CTime | 23:33 |
@sonney2k | get_curtime() | 23:33 |
@sonney2k | resoluition is microseconds | 23:33 |
blackburn | it uses clock() | 23:34 |
blackburn | just like I did | 23:34 |
@sonney2k | nope | 23:34 |
@sonney2k | it uses gettimeofday | 23:34 |
blackburn | ah | 23:34 |
blackburn | I see | 23:34 |
blackburn | will try | 23:34 |
blackburn | sonney2k: about ~0.055 without additional pointer | 23:37 |
blackburn | about 0.055 with | 23:37 |
blackburn | sonney2k: so seems no diffrerence | 23:38 |
@sonney2k | very good | 23:38 |
blackburn | hehe it produced all sources into .s and .ii files | 23:41 |
blackburn | with assembly code | 23:41 |
@sonney2k | :) | 23:41 |
@sonney2k | for some reason 1 of my cpu cores vanished?! | 23:42 |
blackburn | how? :) | 23:49 |
@sonney2k | no idea all slow now | 23:50 |
blackburn | which kernel do you use now? | 23:53 |
@sonney2k | some .38 | 23:59 |
--- Log closed Tue Jun 14 00:00:23 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!