--- Log opened Tue Nov 29 00:00:59 2011 | ||
-!- blackburn [~blackburn@188.168.5.8] has quit [Quit: Leaving.] | 06:28 | |
-!- blackburn [~blackburn@188.168.2.74] has joined #shogun | 11:22 | |
sonne|work | blackburn: what happened? | 12:11 |
---|---|---|
blackburn | sonne|work: nothing, I did git bisect but went to wrong commit I guess | 12:12 |
blackburn | it takes too long to do it again now | 12:12 |
sonne|work | you mean you marked one commit bad that wasn't ? | 12:13 |
sonne|work | (or good) | 12:13 |
blackburn | may be.. | 12:14 |
blackburn | may be I did not clean everything well when recompiling | 12:14 |
blackburn | I could do it again later, but not today | 12:15 |
sonne|work | hmmhh too bad - I guess one should do git clean -dfx or similar | 12:15 |
sonne|work | and use ccache to speed up the procedure | 12:15 |
blackburn | sonne|work: btw I ran GMNPSVM this night | 12:15 |
blackburn | before going sleep hah | 12:15 |
blackburn | and it was one more percent accurate | 12:16 |
sonne|work | I made some progress on the typemap stuff but I would rather want to adjust all typemaps now | 12:16 |
blackburn | 88.8% or so | 12:16 |
sonne|work | as I expected | 12:16 |
sonne|work | I guess you could only gain more when doing these distortions to make the machine learn about invariances | 12:16 |
blackburn | I forgot to add time measure | 12:16 |
sonne|work | anyway back to work | 12:16 |
sonne|work | cu | 12:16 |
blackburn | ok see you | 12:16 |
blackburn | sonne|work: there is a way I can gain still - RoI, custom color/shape kernels, etc | 12:19 |
blackburn | battle is not lost :D | 12:20 |
sonne|work | is this number accuracy? | 12:39 |
sonne|work | how many classes? | 12:39 |
blackburn | sonne|work: 43 | 12:40 |
sonne|work | damn good then already | 12:40 |
sonne|work | will be hard to improve | 12:40 |
blackburn | sonne|work: CNN got 99% :D | 12:40 |
sonne|work | how many examples / class? | 12:41 |
sonne|work | sounds like cheating... | 12:41 |
blackburn | http://benchmark.ini.rub.de/?section=gtsrb&subsection=news | 12:42 |
blackburn | sonne|work: http://benchmark.ini.rub.de/?section=gtsrb&subsection=results | 12:43 |
blackburn | as you can see Commitee of CNNs performed better than human :D | 12:43 |
sonne|work | as usual | 12:43 |
sonne|work | ;-O | 12:43 |
sonne|work | LDA on HOG seems to be doing pretty well | 12:44 |
blackburn | I hope SVM will do the same | 12:44 |
blackburn | these guys went crazy | 12:45 |
blackburn | LDA + HOG + ... + ... | 12:45 |
blackburn | they combined everything :D | 12:45 |
blackburn | sonne|work: is LDA success surprising to you? I do not know much except how it works | 12:46 |
blackburn | hahah | 12:47 |
blackburn | SHOGUN KLTSA applied, took 7.867928s | 12:47 |
blackburn | SKLEARN LTSA applied, took 483.886713s | 12:48 |
blackburn | sonne|work: that's ^ not fair to publish :D | 12:48 |
sonne|work | nice work | 12:48 |
blackburn | why is it so slow.. | 12:49 |
blackburn | sonne|work: the most strangest commit msg I've ever seen: http://code.google.com/p/locally-linear-embedding/source/detail?r=30 | 12:53 |
sonne|work | :) | 12:54 |
blackburn | I guess I won't compare with such strange project :D | 12:54 |
sonne|work | I am impressed that they have 96 % with random forests | 13:06 |
sonne|work | same for LDA + HOG | 13:07 |
sonne|work | not bad | 13:07 |
CIA-33 | shogun: Sergey Lisitsyn master * r6aebf7d / src/shogun/converter/LocalTangentSpaceAlignment.h : Warn that LTSA is slower than KLTSA with linear - http://git.io/rd7uPA | 13:07 |
blackburn | sonne|work: I heard random forests are good for similar tasks | 13:07 |
blackburn | sonne|work: for vehicle (numbers? signs? bad english) recognition | 13:08 |
blackburn | :D | 13:08 |
blackburn | sonne|work: madeleine's issue solved completely, all went well | 13:49 |
blackburn | but we know that java has some leaks | 13:49 |
sonne|work | how did she solve it? | 13:49 |
sonne|work | which leaks? | 13:49 |
sonne|work | too bad that this is not public on the mailinglist | 13:50 |
blackburn | sonne|work: you did, and I | 13:50 |
blackburn | sonne|work: same script in java was eating too much memory | 13:50 |
blackburn | oh I have ten minutes to get myself to job haha | 13:50 |
sonne|work | but where | 13:50 |
sonne|work | hmmhh | 13:50 |
blackburn | sonne|work: I will try to reproduce it later | 13:51 |
blackburn | be right back in twenty minutes haha | 13:51 |
sonne|work | yeah lets get a releasy out | 13:51 |
-!- blackburn [~blackburn@188.168.2.74] has quit [Quit: Leaving.] | 13:51 | |
sonne|work | release out before that | 13:51 |
sonne|work | cu | 13:51 |
sonne|work | I will be away then | 13:51 |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 14:19 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 14:23 | |
blackburn | oh wow | 14:25 |
blackburn | blocker bug on canada project :D | 14:28 |
blackburn | canadian* | 14:28 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Ping timeout: 265 seconds] | 15:45 | |
-!- in3xes_ [~in3xes@180.149.49.230] has joined #shogun | 16:04 | |
-!- in3xes_ [~in3xes@180.149.49.230] has quit [Client Quit] | 16:06 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 240 seconds] | 16:08 | |
-!- blackburn [~blackburn@188.168.2.74] has joined #shogun | 17:59 | |
@sonney2k | canadian? | 20:05 |
blackburn | sonney2k: yeah nevermind :) | 20:06 |
blackburn | recall I work in telecom software company | 20:06 |
-!- supernaught [~hal@host3-196.wifi.ubc.ca] has left #shogun [] | 20:57 | |
@sonney2k | blackburn, how big is wc -l interfaces/python_modular/modshogun_wrap.cxx for you? | 22:22 |
blackburn | 418912 | 22:23 |
@sonney2k | 426712 | 22:23 |
blackburn | sonney2k: are you trying to reduce its size again? | 22:24 |
@sonney2k | I am fixing python typemaps | 22:30 |
blackburn | I see | 22:33 |
@sonney2k | we wrap a lot of stuff that isn't supported | 22:51 |
@sonney2k | e.g. sparsevector | 22:52 |
@sonney2k | or functions that return sgvector of gaussian kernels etc | 22:52 |
blackburn | sonney2k: yeah a lot of useless stuff | 22:54 |
@sonney2k | and I ran into some weird crashes when we disable certain data types | 22:57 |
CIA-33 | shogun: Soeren Sonnenburg master * r4bce5ac / (3 files in 3 dirs): | 23:16 |
CIA-33 | shogun: Overhaul python typemaps | 23:16 |
CIA-33 | shogun: - check that we don't have a list as input | 23:16 |
CIA-33 | shogun: - refactor code to call static functions to only have minimal | 23:16 |
CIA-33 | shogun: functionality inside the actuall typemap | 23:16 |
CIA-33 | shogun: - re-enable boolean type by default to fix crasher - http://git.io/SybNbA | 23:16 |
--- Log closed Wed Nov 30 00:00:59 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!