--- Log opened Sun May 27 00:00:41 2012 | ||
@sonney2k | blackburn, I added some stuff and at least now kernel.kernel() works | 00:02 |
---|---|---|
blackburn | kernel.kernel()? | 00:02 |
blackburn | what's that? | 00:02 |
@sonney2k | kernel.kernel(i,j) | 00:03 |
@sonney2k | gives you k[i,j] :D | 00:03 |
blackburn | ah | 00:03 |
blackburn | kernel(i,j) | 00:03 |
@sonney2k | I don't udnerstand the error though | 00:03 |
blackburn | init one? | 00:03 |
@sonney2k | init? | 00:03 |
@sonney2k | there is no init | 00:03 |
@sonney2k | the error is in get_kernel_matrix() | 00:04 |
@sonney2k | and that is just calling kernel(i,j) | 00:04 |
blackburn | ah so what is the error? | 00:05 |
@sonney2k | RuntimeError: maximum recursion depth exceeded while calling a Python object | 00:05 |
blackburn | argh | 00:05 |
@sonney2k | but at least it still says returning kernel matrix of size 5x5 | 00:07 |
@sonney2k | so it enters get_kernel_matrix | 00:07 |
blackburn | sonney2k: looks like known bug of swig? | 00:07 |
@sonney2k | wait not true | 00:08 |
@sonney2k | that was for the get-kernel_matroix of the linear kernel | 00:08 |
@sonney2k | blackburn, ? | 00:08 |
blackburn | I googled - some issues | 00:08 |
@sonney2k | sure but why does kernel(i,j) work? | 00:08 |
blackburn | different way of call I'd say | 00:08 |
blackburn | ah | 00:09 |
blackburn | no | 00:09 |
@sonney2k | hmmhh, one is using typemaps... | 00:09 |
@sonney2k | it returns SGMatrix | 00:09 |
@sonney2k | kernel(i,j) just double | 00:09 |
@sonney2k | let me try sth | 00:09 |
@sonney2k | blackburn, I now created a simple get_km() | 00:12 |
@sonney2k | returning some sgmatrix | 00:12 |
@sonney2k | directly from directorkernel | 00:12 |
blackburn | works? | 00:12 |
@sonney2k | if that fails we will see | 00:12 |
@sonney2k | compile time is much higher with directors... | 00:12 |
@sonney2k | wrapper is much bigger | 00:12 |
blackburn | hmm | 00:12 |
@sonney2k | blackburn, yes that works | 00:13 |
blackburn | so what is the reason? | 00:14 |
@sonney2k | blackburn, no idea | 00:19 |
@sonney2k | timings: | 00:19 |
@sonney2k | 0.7 sec for a 1000x1000 matrix w/ shogun | 00:19 |
@sonney2k | 15 sec with directors from python | 00:19 |
blackburn | heh | 00:19 |
blackburn | sonney2k: what is the reason between get_km and get_kernel_matrix? | 00:20 |
blackburn | errr | 00:20 |
blackburn | difference | 00:20 |
@sonney2k | blackburn, again no idea | 00:20 |
blackburn | pthreads? | 00:20 |
@sonney2k | btw the example crashes at the end | 00:20 |
blackburn | heh | 00:20 |
@sonney2k | blackburn, well I set threads=1 | 00:20 |
@sonney2k | (of course) | 00:20 |
@sonney2k | it doesn't crash with gdb | 00:21 |
@sonney2k | hah | 00:21 |
@sonney2k | blackburn, also doesn't crash with valgrind | 00:22 |
blackburn | yes that happens sometimes | 00:22 |
@sonney2k | blackburn, do you think gsomix can find out what is wrong? | 00:23 |
blackburn | give a chance :) | 00:23 |
@sonney2k | ok | 00:23 |
@sonney2k | I will mail him the example then | 00:23 |
CIA-113 | shogun: Soeren Sonnenburg master * rd8a7270 / (2 files): add some helper functions for director kernel - http://git.io/nlEDig | 00:29 |
@sonney2k | uhh ohh | 00:29 |
@sonney2k | I have to sleep | 00:29 |
-!- blackburn [~blackburn@31.28.59.65] has quit [Ping timeout: 240 seconds] | 03:36 | |
-!- gsomix [~gsomix@109.169.151.2] has joined #shogun | 08:33 | |
gsomix | sonney2k, I got mail. I got mail. I can try to solve issue in evening | 09:15 |
gsomix | sorry, need to go | 09:15 |
-!- ZiYuan [~ZiYuan@14.151.191.134] has joined #shogun | 10:52 | |
-!- ZiYuan [~ZiYuan@14.151.191.134] has quit [Client Quit] | 10:52 | |
-!- ZiYuan [~ZiYuan@14.151.191.134] has joined #shogun | 10:53 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:53 | |
-!- ZiYuan [~ZiYuan@14.151.191.134] has quit [] | 10:59 | |
-!- ZiYuan [~ZiYuan@14.151.191.134] has joined #shogun | 10:59 | |
-!- ZiYuan [~ZiYuan@14.151.191.134] has quit [Client Quit] | 11:00 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: Lost terminal] | 11:11 | |
-!- blackburn [~blackburn@31.28.59.65] has joined #shogun | 11:27 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 11:40 | |
-!- blackburn [~blackburn@31.28.59.65] has quit [Ping timeout: 260 seconds] | 11:50 | |
-!- gsomix [~gsomix@109.169.151.2] has quit [Quit: Ex-Chat] | 14:14 | |
-!- blackburn [~blackburn@31.28.59.65] has joined #shogun | 19:14 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 19:34 | |
-!- blackburn [~blackburn@31.28.59.65] has quit [Quit: Leaving.] | 19:36 | |
-!- blackburn [~blackburn@31.28.59.65] has joined #shogun | 19:37 | |
-!- gsomix [~gsomix@188.168.5.87] has joined #shogun | 20:16 | |
gsomix | fuf, I have returned to dorm. | 20:16 |
gsomix | time to work | 20:16 |
gsomix | sonney2k, blackburn, hi | 20:17 |
blackburn | hey | 20:18 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:35 | |
blackburn | n4nd0: hey | 20:56 |
blackburn | I tweaked it up to 97.82 :D | 20:56 |
n4nd0 | blackburn: hey man, how is it going? | 20:56 |
n4nd0 | blackburn: wow nice tweak then :) | 20:56 |
n4nd0 | blackburn: what value were you aiming at? | 20:57 |
blackburn | no idea I just needed to overwhelm 97.00% | 20:57 |
n4nd0 | objective achieved then | 20:57 |
blackburn | well.. before (in paper for that russian journal) I stated I 'can' recognize with 97.43 accuracy | 20:57 |
blackburn | so objective is rather 97.43 | 20:57 |
blackburn | but achieved yeah | 20:58 |
n4nd0 | congrats ;) | 20:58 |
blackburn | however I have not too much time now to get thesis itself done | 20:59 |
n4nd0 | blackburn: do you have to write it completely? | 21:02 |
blackburn | yes needs to be 100% done before 7th June | 21:03 |
n4nd0 | blackburn: you write it in Russian, right? | 21:08 |
blackburn | yes | 21:08 |
n4nd0 | :( it would be cool to read it otherwise | 21:08 |
blackburn | hah nothing special | 21:09 |
blackburn | I have nothing but HOG and SVM here | 21:09 |
blackburn | Crammer-Singer SVM + Homogeneous Kernel Map + HOG | 21:09 |
n4nd0 | blackburn: I insist, probably interesting to read ;) | 21:11 |
blackburn | good thing being russian is direct access to vapnik's stuff | 21:12 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 21:15 | |
gsomix | sonney2k, around | 21:41 |
gsomix | *? | 21:41 |
gsomix | hmm, nevermind :) | 21:42 |
@sonney2k | gsomix, any news? | 23:05 |
gsomix | sonney2k, yep. | 23:05 |
gsomix | get_kernel_matrix works. | 23:05 |
blackburn | sonney2k: to redeclare virtuals is the key :D | 23:06 |
gsomix | blackburn, tnx for fast description. :) | 23:08 |
@sonney2k | k | 23:10 |
@sonney2k | no more crashes? | 23:10 |
gsomix | sonney2k, hmm, I have not noticed more. but work in progress, need 15mins. | 23:11 |
blackburn | sonney2k: what do you think about BSVM? | 23:27 |
@sonney2k | blackburn, don't think it is worth it | 23:31 |
blackburn | why? | 23:31 |
@sonney2k | but I haven't read the paper/mail :) | 23:31 |
@sonney2k | gmnpsvm should be doing better | 23:31 |
@sonney2k | we need something massively multiclass | 23:32 |
blackburn | MDM is cool yes but I do not think BSVM is slower | 23:32 |
blackburn | sonney2k: ok then another thing - strprtool has 5 solver for gmnp | 23:33 |
blackburn | solvers* | 23:33 |
blackburn | why do we have only mdm? | 23:34 |
blackburn | do not say ask vojtech :D | 23:39 |
gsomix | sonney2k, ok, works. but this requires cleaning up | 23:55 |
gsomix | btw, feats_train.parallel.set_num_threads(2) -> segfault | 23:56 |
gsomix | sonney2k, can I commit/push it tomorrow? | 23:59 |
gsomix | tired, time to sleep =__= | 23:59 |
--- Log closed Mon May 28 00:00:41 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!