--- Log opened Mon Jul 22 00:00:27 2013 | ||
--- Day changed Mon Jul 22 2013 | ||
@sonney2k | lisitsyn, already having the option not to init memory is very helpful | 00:00 |
---|---|---|
lisitsyn | sonney2k: if you realloc too much it may be a wrong data structure | 00:00 |
@sonney2k | so in a loop where you fill an array anyway | 00:00 |
lisitsyn | sonney2k: reserve doesn't init memory | 00:00 |
lisitsyn | resize does reserve doesn't | 00:01 |
@sonney2k | lisitsyn, sure but in cases you don't know how big the thing will be | 00:03 |
lisitsyn | sonney2k: that's for linked list right? | 00:03 |
@sonney2k | lisitsyn, for e.g a tree | 00:03 |
lisitsyn | why to store it in a vector? | 00:04 |
@sonney2k | to save memory | 00:05 |
@sonney2k | you only have one big memory block and indices into the nodes | 00:05 |
lisitsyn | sonney2k: I wouldn't use vector for that then! | 00:05 |
@sonney2k | lisitsyn, what else? | 00:06 |
lisitsyn | sonney2k: it is like when you write your own data structure | 00:07 |
lisitsyn | so you can put some manually controlled memory | 00:07 |
thoralf | sonney2k: examples from libshogun are working, but cmdline_static has 5 segfaulting tests... | 00:14 |
thoralf | #0 shogun::CGUIPreprocessor::add_preproc (this=0x620620, preproc=0x62a2f0) at ui/GUIPreprocessor.cpp:122 | 00:16 |
thoralf | #1 0x00007ffff792c03f in shogun::CSGInterface::cmd_add_preproc (this=0x613e00) at ui/SGInterface.cpp:5263 | 00:16 |
thoralf | #2 0x00007ffff792ecc2 in shogun::CSGInterface::handle (this=0x613e00) at ui/SGInterface.cpp:7965 | 00:16 |
thoralf | #3 0x0000000000408af8 in main (argc=<value optimized out>, argv=0x7fffffffe0f8) at CmdLineInterface.cpp:753 | 00:16 |
@iglesiasg | lisitsyn, sonney2k : PR alarm! | 00:16 |
thoralf | in distance_hammingword.sg | 00:16 |
lisitsyn | haha | 00:16 |
@sonney2k | thoralf, that sounds like a change of Heiko | 00:19 |
@sonney2k | I didn't do anything there | 00:19 |
thoralf | sonney2k: I'm putting away my old laptop now and get some sleep. | 00:20 |
thoralf | sonney2k: If you need more help, you need to wait until tomorrow after work. | 00:21 |
@sonney2k | thoralf, no it is all good! thanks! | 00:21 |
@sonney2k | lisitsyn, whenever you see Heiko/wiking tell them we need tests back working or we will run into a nightmare mess | 00:22 |
lisitsyn | yeah | 00:22 |
* sonney2k is finally entering the nips reviews | 00:23 | |
thoralf | Guys, see you tomorrow. | 00:23 |
thoralf | Sorry, today. ;) | 00:23 |
@iglesiasg | thoralf: good night! | 00:23 |
-!- thoralf [~thoralf@37-5-32-132-dynip.superkabel.de] has quit [Quit: Konversation terminated!] | 00:24 | |
@sonney2k | ohh man why do I only get crappy papers for review | 01:05 |
lisitsyn | sonney2k: again? | 01:08 |
@sonney2k | bah | 01:09 |
@iglesiasg | sonney2k: what are they about? | 01:10 |
@sonney2k | can't say svm, mkl this kind of stuff | 01:19 |
@sonney2k | all submitted just 9 days late | 01:36 |
@sonney2k | alrighty sleep! | 01:37 |
@iglesiasg | good night! | 01:38 |
-!- FSCV [~FSCV@216-230-229-167-colo.oplink.net] has quit [Ping timeout: 268 seconds] | 01:46 | |
-!- FSCV [~FSCV@189.139.250.12] has joined #shogun | 01:49 | |
-!- nube [~rho@49.244.123.105] has quit [Ping timeout: 264 seconds] | 01:51 | |
-!- iglesiasg [~Fernando@83.179.44.135] has quit [Quit: Leaving] | 01:54 | |
-!- nube [~rho@49.244.51.204] has joined #shogun | 02:04 | |
-!- FSCV_ [~FSCV@189.139.138.157] has joined #shogun | 02:14 | |
-!- FSCV [~FSCV@189.139.250.12] has quit [Ping timeout: 248 seconds] | 02:15 | |
-!- FSCV_ [~FSCV@189.139.138.157] has quit [Quit: Leaving] | 02:21 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:33 | |
-!- nube [~rho@49.244.51.204] has quit [Quit: Leaving.] | 02:47 | |
shogun-buildbot | build #400 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/400 | 03:07 |
shogun-buildbot | build #465 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/465 | 04:22 |
-!- gsomix [~gsomix@95.67.172.110] has quit [Ping timeout: 245 seconds] | 06:30 | |
-!- gsomix [~gsomix@95.67.172.110] has joined #shogun | 06:37 | |
-!- lisitsyn [~lisitsyn@92-240-133-94.clients.tlt.100megabit.ru] has quit [Quit: Leaving.] | 07:43 | |
-!- nube [~rho@116.90.239.13] has joined #shogun | 07:46 | |
-!- nube [~rho@116.90.239.13] has quit [Quit: Leaving.] | 08:02 | |
-!- foulwall [~user@2001:da8:215:503:1d6f:dc69:2111:a45a] has joined #shogun | 08:12 | |
-!- gsomix_ [~gsomix@88.200.188.178] has joined #shogun | 08:44 | |
-!- gsomix [~gsomix@95.67.172.110] has quit [Ping timeout: 248 seconds] | 08:47 | |
-!- foulwall [~user@2001:da8:215:503:1d6f:dc69:2111:a45a] has quit [Remote host closed the connection] | 09:14 | |
-!- nube [~rho@116.90.239.13] has joined #shogun | 09:18 | |
-!- lambday [67157d4f@gateway/web/freenode/ip.103.21.125.79] has joined #shogun | 09:29 | |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:21 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 10:21 | |
@iglesiasg | good morning | 10:22 |
-!- foulwall [~user@2001:da8:215:503:d2e:b5f1:ff20:a998] has joined #shogun | 10:32 | |
-!- gsomix_ [~gsomix@88.200.188.178] has quit [Remote host closed the connection] | 10:43 | |
-!- lambday [67157d4f@gateway/web/freenode/ip.103.21.125.79] has quit [Quit: Page closed] | 11:07 | |
-!- Yanglittle [b74040fc@gateway/web/freenode/ip.183.64.64.252] has joined #shogun | 11:16 | |
Yanglittle | excuse me, is there any body? | 11:16 |
Yanglittle | excuse me, is there any body? | 11:16 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 11:41 | |
lisitsyn | hey there | 11:45 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has joined #shogun | 12:01 | |
van51 | hello | 12:01 |
-!- nube [~rho@116.90.239.13] has quit [Quit: Leaving.] | 12:08 | |
-!- lambday [67157f36@gateway/web/freenode/ip.103.21.127.54] has joined #shogun | 12:09 | |
@sonney2k | van51, hey there! | 12:13 |
van51 | sonney2k: hello :) | 12:13 |
@sonney2k | and hey foulwall ! | 12:13 |
@sonney2k | van51, so what would you want to do now? | 12:14 |
-!- foulwall [~user@2001:da8:215:503:d2e:b5f1:ff20:a998] has quit [Ping timeout: 245 seconds] | 12:14 | |
van51 | sonney2k: hmm I think I could add support for quadratic features in the hashed sparse and dense classes | 12:14 |
van51 | sonney2k: don't think it will take long | 12:15 |
@sonney2k | van51, it would be could if you could continue the dotfeatures hacking spree doing things like the random fourier features and lbppyrfeatures etc | 12:15 |
@sonney2k | van51, I finally finished nips reviews and so I can use my spare time for shogun stuff | 12:16 |
van51 | sonney2k: nice :) | 12:18 |
van51 | sonney2k: well I can also continue with those and wait if Benoir or Olivier have another suggestion for extension or something | 12:18 |
-!- thoralf [~thoralf@enki.zib.de] has joined #shogun | 12:19 | |
thoralf | Heyho. | 12:19 |
van51 | sonney2k: btw on another note, for the GSoC we are supposed to submit some answers about our mentor | 12:19 |
van51 | sonney2k: I have three mentors now I think. Who is supposed to be the official one? | 12:20 |
@sonney2k | van51, we are one entity | 12:21 |
@sonney2k | van51, and we are happy with what you are doing so no worries either | 12:22 |
van51 | sonney2k: ok then :) | 12:22 |
@sonney2k | van51, btw what n-gram length did you use in the webspam thing? | 12:22 |
van51 | sonney2k: size 8 | 12:23 |
@sonney2k | ahh ok that explains the good performance | 12:23 |
van51 | sonney2k: do you want me to try with size=3? | 12:23 |
@sonney2k | van51, and the experiment was with preprocessing> | 12:23 |
@sonney2k | ? | 12:24 |
@sonney2k | or with the dot features? | 12:24 |
van51 | sonney2k: no everything was on the fly | 12:24 |
van51 | sonney2k: with dot features | 12:24 |
@sonney2k | so dot | 12:24 |
@sonney2k | but not streaming right? | 12:24 |
van51 | sonney2k: nope, because the current streaming classes read line by line | 12:24 |
van51 | sonney2k: instead of delimiting on \0 eg | 12:24 |
@sonney2k | as in '\n' as separator | 12:25 |
@sonney2k | yes o | 12:25 |
@sonney2k | k | 12:25 |
van51 | sonney2k: do you want me to try a different setting? | 12:25 |
van51 | sonney2k: I run it on another machine so I won't mind | 12:25 |
@sonney2k | did you compare with the results I had in my paper? | 12:25 |
@sonney2k | van51, btw how much memory do 100k examples take? | 12:25 |
van51 | sonney2k: I haven't checked that. I know that it crashes on 150k examples | 12:26 |
@sonney2k | http://sonnenburgs.de/soeren/publications/SonRaeRie07.pdf | 12:26 |
@sonney2k | table 4.4 | 12:27 |
@sonney2k | look at the training time | 12:27 |
@sonney2k | 107661 seconds | 12:28 |
@sonney2k | for 100k | 12:28 |
@sonney2k | auROC 99.64% | 12:28 |
van51 | sonney2k: well it's much faster and for a high number of bits in the hash seems to perform better | 12:29 |
@sonney2k | van51, what did you have? | 12:29 |
van51 | for eg 16 bits in hash | 12:30 |
@sonney2k | van51, I don't have my tomtom mail here so I cannot check | 12:30 |
van51 | sonney2k: it took 11044 seconds and had a auROC of 0.998183 | 12:30 |
@sonney2k | van51, it is a weird that it is better | 12:31 |
@sonney2k | should be the same result | 12:31 |
@sonney2k | 10 times faster it seems - which is good | 12:31 |
@sonney2k | van51, would be interesting to see the result when you use the converter in a preprocessing step - with same sqrt normalization | 12:33 |
@sonney2k | van51, was that liblinear or svmocas btw? | 12:33 |
van51 | sonney2k: svmocas | 12:33 |
@sonney2k | epsilon default? | 12:34 |
van51 | sonney2k: yeap and C=0.01 | 12:34 |
van51 | sonney2k: I have the code here https://gist.github.com/van51/5938097 if you want to check it for anything I missed | 12:34 |
@sonney2k | van51, that explains it | 12:35 |
@sonney2k | van51, I had C=1 in the experiment | 12:35 |
van51 | sonney2k: and you had used tri-grams? | 12:35 |
@sonney2k | van51, so C=0.01 seems to be a better choice | 12:36 |
@sonney2k | no 8-grams | 12:36 |
van51 | sonney2k: ah | 12:36 |
van51 | sonney2k: do you want to compare in the same settings? | 12:36 |
@sonney2k | van51, you could pick e.g. 5k examples and compare | 12:36 |
@sonney2k | but I think it is alright | 12:37 |
van51 | sonney2k: ok I will, just to be thorough :D | 12:38 |
van51 | sonney2k: so about what's next, what do you suggest? quadratic or randomfourier/lbppyr ? | 12:39 |
@sonney2k | van51, quadratic is easy to do and very useful so it is good if you do that too. I don't know how to properly stack things though :/ | 12:40 |
@sonney2k | I mean consider that you have n-gram's and want to use quadratic on top | 12:40 |
@sonney2k | no idea how this could work in a general framework | 12:41 |
@sonney2k | so for now dense/sparse is OK | 12:41 |
van51 | sonney2k: and the user would specify the indices to be paired? | 12:41 |
@sonney2k | I have no masterplan | 12:42 |
@sonney2k | I would say for now do the comparison with C=1 on 5k examples and the quadratic features for dense and sparse | 12:43 |
@sonney2k | maybe sth comes to our mind in the meantime | 12:43 |
van51 | sonney2k: ok, just one question about the quadratic features | 12:44 |
@sonney2k | van51, btw for dense specifying which pairs should be quadratic is likely a good idea to do (sometimes) | 12:44 |
van51 | sonney2k: if I'm not mistaken in vw the user specifies the namespaces to be pired | 12:44 |
@sonney2k | and for quadratic features it makes sense to keep also the linear terms (at least optional | 12:44 |
@sonney2k | ) | 12:44 |
van51 | sonney2k: what's the alternative? to brute-force all possible pairs? | 12:44 |
@sonney2k | so you have the x_1,...x_D and x_1*x_1,x_1*x_2...x_1*x_D, x_2*x_2...x_2*x_D... | 12:45 |
@sonney2k | yes exactly | 12:46 |
@sonney2k | for dense that is simply 2 nested for loops | 12:46 |
@sonney2k | for sparse you do that only over the available dims | 12:46 |
van51 | sonney2k: ok got it | 12:47 |
van51 | sonney2k: will get right on it | 12:47 |
@sonney2k | so for high sparsitye & sparse it is cheap for dense not so | 12:47 |
@sonney2k | but at least you don't need to store D^2 many features in memory | 12:47 |
@sonney2k | van51, btw you could compare it with the result of a polynomial kernel svm on some toy data if you use the factors 2*x_1*x_2 for cross terms | 12:48 |
@sonney2k | which you in the final version can just leave out | 12:48 |
@sonney2k | they are not doing any good but are a necessary evil when using kernels | 12:49 |
@sonney2k | so the polykernel cannot do w/o them | 12:49 |
@sonney2k | van51, questions so far? | 12:49 |
van51 | sonney2k: I don't think so | 12:50 |
@sonney2k | if not I will swim a few km now | 12:50 |
van51 | sonney2k: if something comes up, I'll bug you again later | 12:50 |
van51 | sonney2k: nice, hopefully by the end of the week i'll be swimming as well :D | 12:50 |
van51 | sonney2k: although I don't think I can last that much, more like 50m :p | 12:51 |
@sonney2k | van51, heh weather is nice today no wind no clouds just calm sea | 12:51 |
@sonney2k | and water is cold which makes me swim fast | 12:51 |
@sonney2k | alrighty | 12:51 |
@sonney2k | enough | 12:51 |
@sonney2k | water here I come | 12:51 |
van51 | sonney2k: enjoy then! | 12:52 |
* sonney2k off | 12:52 | |
@iglesiasg | sonney2k, hey! I think there is something going wrong with c++11 in travis clang | 12:58 |
-!- HeikoS [~heiko@176.248.212.166] has joined #shogun | 13:18 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 13:18 | |
thoralf | Hey HeikoS | 13:19 |
@HeikoS | thoralfhi! :) | 13:19 |
@HeikoS | lambday: hi! | 13:19 |
@HeikoS | thoralf: sorry did not yet have time to write the automated testing thing down but will soon! :) | 13:20 |
thoralf | HeikoS: sonney2k told us, if you get to the channel, please try to fix travis. ;) | 13:20 |
@HeikoS | thoralf: yes, will do today | 13:20 |
thoralf | told us to tell you | 13:20 |
@HeikoS | thoralf: thanks! :) | 13:21 |
lambday | HeikoS: hi | 13:21 |
@HeikoS | lambday: how are things? | 13:21 |
lambday | HeikoS: a bit slow this weekend.. :( | 13:21 |
@HeikoS | lambday: any particular issues? | 13:21 |
lambday | HeikoS: hopefully I'll finish COCG_M by tonight | 13:22 |
@HeikoS | lambday: ok, good to hear. what about the triangular solver? | 13:22 |
lambday | HeikoS: no just got occupied with a few other things :( | 13:22 |
lambday | HeikoS: well, turns out that alglib actually uses lapack for that particular thing | 13:22 |
@HeikoS | lambday: I see well we are pretty good in schedule | 13:22 |
@HeikoS | lambday: I see, mmh | 13:23 |
@HeikoS | lambday: what about eigen3 again? | 13:23 |
@HeikoS | I dont like lapack so much | 13:23 |
lambday | HeikoS: eigen3 I guess don't have that | 13:23 |
@HeikoS | lambday: it IS in there, but not exposed in a nice way | 13:23 |
@HeikoS | which is annoying | 13:23 |
lambday | HeikoS: not exposed as in? | 13:24 |
@HeikoS | lambday: cannot access it easily | 13:24 |
lambday | okay | 13:24 |
lambday | but shall I try to study that instead of incorporating lapack routine? | 13:24 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 13:25 | |
@HeikoS | lambday: how for advanced are you with that? | 13:26 |
lambday | with the lapack? | 13:26 |
@HeikoS | http://eigen.tuxfamily.org/dox/TutorialSparse.html#TutorialSparseDirectSolvers | 13:26 |
@HeikoS | Just as with dense matrices, the triangularView() function can be used to address a triangular part of the matrix, and perform triangular solves with a dense right hand side: | 13:26 |
@HeikoS | isnt that helpful? | 13:27 |
lambday | HeikoS: can we set the subdiagonals in sparse easily? | 13:27 |
@HeikoS | lambday: dont know ... | 13:27 |
lambday | well, for lanczos, the system is not that huge | 13:28 |
lambday | we can use dense things | 13:28 |
@HeikoS | lambday: yes before I meant how much of the lapack/alglib integration have you done | 13:28 |
lambday | HeikoS: not started with that yet, was working on COCG_M | 13:28 |
@HeikoS | lambday: okay good, | 13:28 |
@HeikoS | so its worth thinking about it | 13:29 |
@HeikoS | for what do you need the triangular exactly again? | 13:29 |
lambday | for lanczos, it sets the diagonal of a matrix from some values generated in its cg iterations | 13:29 |
lambday | so, its not that huge | 13:29 |
lambday | HeikoS: can we think about importing readymade Lanczos implementation from somewhere? | 13:30 |
lambday | HeikoS: well, that would create another dependency.. CG things are pretty much same | 13:31 |
@HeikoS | lambday: would be good, but also sounds painful ;) | 13:31 |
@HeikoS | lambday: so erlend spent quite some time on alglib design | 13:31 |
@HeikoS | maybe its a good idea just to make it the same | 13:31 |
lambday | HeikoS: well, krylstat simply uses alglib routine for solving the tridiagonal system.. we have to import that method in shogun if we wanna do the same | 13:32 |
lambday | HeikoS: | 13:32 |
lambday | https://github.com/lambday/KRYLSTAT/blob/master/lanczos/eigen/eigen_lanczos.h#L111 | 13:32 |
@HeikoS | lambday: we are already depending on lapack | 13:33 |
@HeikoS | and the I ported quite a few alglib routines already, its a bit work but not hard to to and doesnt create dependencies | 13:33 |
lambday | HeikoS: I was taking a look.. they use quite a few internal data structures which I gotta check how they works... plus they say that they imported that particular routine from lapack | 13:34 |
@HeikoS | lambday: oh .... | 13:34 |
lambday | I was then having a look at our lapack.h | 13:34 |
lambday | if that was already there.. | 13:34 |
lambday | but I didn't get how our lapack works | 13:35 |
@HeikoS | lambday: ha, through chaos :D | 13:35 |
lambday | hehe :D | 13:35 |
@HeikoS | ok, we should check this out a bit more then, I wil lhave lunch now and then be back, will you be around a bit longer? | 13:35 |
lambday | HeikoS: yes I am here | 13:35 |
@HeikoS | lambday: ok cool | 13:36 |
lambday | HeikoS: enjoy your meal :) | 13:36 |
@HeikoS | lambday: ah btw | 13:36 |
@HeikoS | http://arxiv.org/abs/1307.5302 | 13:36 |
lambday | HeikoS: this is another paper? :D | 13:36 |
@HeikoS | lambday: yes thats the nips kernel sampler | 13:37 |
lambday | awesome!!! | 13:37 |
@HeikoS | lambday: nice pictures in there ;) | 13:37 |
@HeikoS | ok going now, see you later | 13:37 |
lambday | HeikoS: I am checking.. ciao :) | 13:37 |
-!- HeikoS [~heiko@176.248.212.166] has left #shogun [] | 13:41 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:788e:31ff:81c6:5bc9] has joined #shogun | 14:07 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 14:07 | |
-!- foulwall [~user@2001:da8:215:503:d2e:b5f1:ff20:a998] has joined #shogun | 14:08 | |
-!- FSCV [~FSCV@50.7.50.60] has joined #shogun | 14:08 | |
-!- nube [~rho@49.244.18.229] has joined #shogun | 14:33 | |
-!- foulwall [~user@2001:da8:215:503:d2e:b5f1:ff20:a998] has quit [Read error: Connection reset by peer] | 14:34 | |
-!- HeikoS [~heiko@nat-180-24.internal.eduroam.ucl.ac.uk] has joined #shogun | 14:47 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:47 | |
lambday | HeikoS: hi | 14:48 |
@HeikoS | loa hi | 14:49 |
@HeikoS | lambday: hi | 14:49 |
lambday | HeikoS: so in the Lanczos, we form this T matrix.. whose dimension is nxn, n=number of cg iteration... | 14:50 |
lambday | HeikoS: now, I can treat that as a dense matrix and run a self adjoint eigen solver | 14:50 |
lambday | because, eigen3 has this tridiagonalization thing... but I don't know if I can use that for solving eigenvalue problems | 14:51 |
@HeikoS | lambday: it *is* in fact dense richt? | 14:51 |
lambday | HeikoS: its tridiagonal... | 14:51 |
lambday | HeikoS: not so dense | 14:51 |
@HeikoS | lambday: ah | 14:52 |
@HeikoS | mmh | 14:52 |
@HeikoS | that would only work for a few thousand iterations then | 14:52 |
@HeikoS | so not so good | 14:52 |
lambday | yep :( | 14:52 |
lambday | I guess that's why erlend used particular method for tridiagonal specifically | 14:53 |
lambday | which is not aparently there in eigen3 | 14:53 |
@HeikoS | lambday: I guess so | 14:53 |
@HeikoS | lambday: lets maybe discuss with him? | 14:53 |
lambday | alright | 14:53 |
@HeikoS | lambday: could you write an email, summarise the problem and possible solutions? | 14:53 |
@HeikoS | maybe he has some new insights | 14:53 |
@HeikoS | lambday: how long do you think until the rest is finished and we need this? | 14:54 |
lisitsyn | I heard Lanczos | 14:54 |
lisitsyn | I have highlight on this word lol | 14:54 |
lambday | HeikoS: this is for the extremal eigenvalues... so even if our COCG_M works, we can't approximate logdet until we have this module working | 14:55 |
lambday | I planned to finish it within next week | 14:55 |
@HeikoS | lambday: ok | 14:55 |
lambday | and then move on towards graph coloring things cause the rest of the stuffs will be almost ready | 14:55 |
@HeikoS | lambday: : might be good to ask sergey :) | 14:55 |
@HeikoS | lambday: the is always my joker ;) | 14:55 |
lambday | lisitsyn: hehe :D | 14:55 |
@HeikoS | lisitsyn: how is it going btw? | 14:55 |
@HeikoS | long time no see | 14:55 |
lisitsyn | HeikoS: hey | 14:56 |
@HeikoS | settled in again? | 14:56 |
lisitsyn | HeikoS: yeah | 14:56 |
lambday | I saw you guys' pics :D | 14:56 |
lambday | lisitsyn: we need Lanczos for large sparse Hermitian matrix | 14:57 |
lambday | HeikoS: shall I cc to Dan too? | 14:58 |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:788e:31ff:81c6:5bc9] has quit [Quit: Ex-Chat] | 15:00 | |
-!- iglesiasg_ [~iglesias@n131-p244.kthopen.kth.se] has joined #shogun | 15:00 | |
@HeikoS | lambday: yes | 15:02 |
@HeikoS | lisitsyn: or do you have an idea? | 15:02 |
@HeikoS | lisitsyn: since we a talked about this before | 15:02 |
@HeikoS | lambday: also cc sergey | 15:02 |
lambday | HeikoS: alright :) | 15:02 |
@sonney2k | van51, no idea if it is good news but I survived ;) | 15:03 |
-!- iglesiasg_ [~iglesias@n131-p244.kthopen.kth.se] has quit [Read error: Connection reset by peer] | 15:03 | |
-!- iglesiasg__ [~iglesias@n131-p244.kthopen.kth.se] has joined #shogun | 15:03 | |
@sonney2k | hey HeikoS ! | 15:03 |
@sonney2k | HeikoS, any plans to get the tests back to greeeen? | 15:03 |
@HeikoS | sonney2k: yes will do that later today | 15:03 |
@HeikoS | sorry for having broken them for so long | 15:04 |
@HeikoS | already spent so much time on fixing that I had to do some other things I postponed ;) | 15:04 |
@HeikoS | NIPS review is coming soon and we have to prepare stuff for that | 15:04 |
@HeikoS | sonney2k: btw http://arxiv.org/abs/1307.5302 | 15:04 |
@HeikoS | lisitsyn: ^ | 15:05 |
lisitsyn | HeikoS: | 15:07 |
@sonney2k | HeikoS, well the alternative is to blacklist all failing tests | 15:08 |
@HeikoS | sonney2k: nono, I am sure I can fix them soon | 15:08 |
@sonney2k | HeikoS, thoralf yesterday figured out that we have some failing examples in static interfaces | 15:09 |
@sonney2k | HeikoS, so these might be due to some fixes in the GUI* classes | 15:09 |
@HeikoS | sonney2k: I hope that wasnt me touching the GUI | 15:09 |
@HeikoS | sonney2k: ah damn | 15:09 |
@HeikoS | sonney2k: ok will have a look into them | 15:09 |
@sonney2k | HeikoS, could be :P | 15:09 |
@HeikoS | which language? | 15:09 |
thoralf | HeikoS: cmdline_static | 15:10 |
@HeikoS | sonney2k: buildbot should tell that right? | 15:10 |
@sonney2k | HeikoS, well get the tests back to work and we will see anyway | 15:10 |
@sonney2k | HeikoS, the buildbot doesn't build the rest if already libshogun & tests fail | 15:10 |
@HeikoS | sonney2k: ah | 15:10 |
@HeikoS | damn | 15:10 |
@sonney2k | HeikoS, btw I want to get rid of the get_feature_type / get_feature_class / get_kernel_type and all these ENUMS | 15:10 |
@HeikoS | so that slipped through | 15:10 |
iglesiasg__ | sonney2k, HeikoS, thoralf: and what's with travis clang and c++11? | 15:10 |
@HeikoS | sonney2k: ok I will work on it soon | 15:11 |
@sonney2k | iglesiasg__, is clang not supporting c++11? What do you mean? | 15:11 |
iglesiasg__ | sonney2k, let me show you | 15:11 |
@sonney2k | HeikoS, ahh and maybe important to I added a check to detect c++11 | 15:11 |
@sonney2k | HeikoS, if available we use atomic<int> for locking in SGReferencedData (this has 0 memory overhead!!) | 15:12 |
iglesiasg__ | sonney2k, https://travis-ci.org/shogun-toolbox/shogun/jobs/9333590 | 15:12 |
iglesiasg__ | sonney2k, huge pile of cc1plus: error: unrecognized command line option '-std=c++11' | 15:12 |
@sonney2k | iglesiasg__, clang is a b0rken mess | 15:13 |
iglesiasg__ | sonney2k, it seems that started with the add cxx11 support: https://travis-ci.org/shogun-toolbox/shogun/builds | 15:13 |
-!- iglesiasg__ is now known as iglesiasg | 15:14 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 15:14 | |
@sonney2k | how can they not fail when they don't support an option | 15:14 |
@sonney2k | iglesiasg, so it seems the build just went fine | 15:14 |
@iglesiasg | sonney2k, mmm I don't think so | 15:16 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 15:17 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 99cfdb8 / src/configure: https://github.com/shogun-toolbox/shogun/commit/99cfdb8a2da062b730a065002b8744e2cd2412ab | 15:17 |
shogun-notifier- | shogun: use -Werror to workaround b0rken clang | 15:17 |
lambday | HeikoS: sent the mail | 15:19 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 79ad3c1 / src/shogun/features/ (2 files): https://github.com/shogun-toolbox/shogun/commit/79ad3c1a07ec1ef8c63790c98a215579d4a7d06e | 15:19 |
shogun-notifier- | shogun: work in the shogun namespace even in the .cpp file for templated classes | 15:19 |
@sonney2k | iglesiasg, indeed | 15:20 |
@HeikoS | lambday: nice thanks, lets see what they say ,we still got some days until COCG is finished | 15:20 |
lambday | HeikoS: yes :) | 15:20 |
lambday | HeikoS: by the way, so the stuffs left as of now are 1)COCG_M 2) Lanczos and 3) probing vectors, right? | 15:21 |
lambday | and preconditioned CG | 15:21 |
@HeikoS | lambday: yes | 15:21 |
@HeikoS | pretty much | 15:21 |
@HeikoS | probing will be some work | 15:21 |
@HeikoS | too :) | 15:21 |
@HeikoS | so thats the hardest parts of the project | 15:21 |
@HeikoS | after that, we will spend some time on testing and documenting | 15:21 |
@HeikoS | maybe even write some static interfaces as long as the R-bindings do not properly work | 15:22 |
lambday | alright :) | 15:22 |
lambday | HeikoS: well one thing... how much accuracy can we expect using Gaussian samples by the way? | 15:22 |
lambday | HeikoS: have you noticed anything in particular? | 15:23 |
@sonney2k | iglesiasg, hopefully fixed... | 15:23 |
@iglesiasg | sonney2k, yes, let's see | 15:23 |
@HeikoS | lambday: well in the limit, 100% | 15:23 |
@HeikoS | and with finite samples, I have no idea ;) | 15:24 |
@HeikoS | lambday: why? | 15:24 |
lambday | HeikoS: just trying to get an idea whether I should get depressed when I write the unit-tests next week and it turns out to be not so accurate :( | 15:24 |
@HeikoS | lambday: ah, no dont worry | 15:24 |
@HeikoS | lambday: in fact, we should not unit test the accuracy too much | 15:24 |
@HeikoS | rather use all basis vectors | 15:25 |
@HeikoS | to get the exact trace | 15:25 |
@HeikoS | lambday: the tests should not take too long | 15:25 |
@sonney2k | lambday, ohh you have seen the picture of us? Are you as shocked as I am how tall HeikoS is? | 15:26 |
@sonney2k | Gunnar must have felt puny! | 15:26 |
@HeikoS | everyone should be shocked! ;) | 15:26 |
* sonney2k already did | 15:26 | |
lambday | sonney2k: lol you are tall too :-o | 15:26 |
lambday | HeikoS: you're 6'4"?? | 15:26 |
@HeikoS | lambday: 2meters | 15:26 |
@sonney2k | 10cm less | 15:27 |
* sonney2k ^ | 15:27 | |
lambday | errr... :D | 15:27 |
@sonney2k | lambday, competition is not good sometimes ;) | 15:27 |
@HeikoS | tall people are awesome, as we always say in the boxing gym >:o | 15:27 |
lambday | haha :D | 15:28 |
@iglesiasg | I felt I was a dwarf next to you guys in Berlin :( | 15:28 |
lambday | iglesiasg: I'd have too :( | 15:28 |
lambday | iglesiasg: average Indian guy :( | 15:28 |
@iglesiasg | lambday, hehe how long is that? I am 1.73 | 15:29 |
@HeikoS | iglesiasg: more wheatbeer is needed then | 15:29 |
-!- gsomix [~gsomix@88.200.188.178] has joined #shogun | 15:29 | |
thoralf | lambday: What does average mean? | 15:29 |
gsomix | thunderstorm! | 15:29 |
@iglesiasg | HeikoS, I was missing that! :) | 15:29 |
thoralf | iglesiasg: That's good european average :) | 15:29 |
lambday | iglesiasg: well, 5'7" | 15:29 |
lambday | so, that too is around 1.72 I guess | 15:30 |
@iglesiasg | lambday, pretty much the same as me then :) | 15:30 |
lambday | iglesiasg: brother :D | 15:30 |
@iglesiasg | thoralf, well living in Sweden and after my visit to Berlin, I think I am below the average :P | 15:30 |
@iglesiasg | but when I get to the subway in Madrid, that feeling changes | 15:31 |
@sonney2k | lambday, if we organize sth next year maybe you can come :) | 15:31 |
lambday | sonney2k: I so wish to.. let's see how things are :) | 15:32 |
shogun-buildbot | build #1492 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1492 blamelist: Soeren Sonnenburg <sonne@debian.org> | 15:32 |
@iglesiasg | sonney2k, the flag work fine for clang travis indeed | 15:33 |
@sonney2k | HeikoS, regarding the failing static interfaces - if one fails all fail - that is the good thing :D | 15:34 |
thoralf | iglesiasg: Wikipedia says for men its 177cm as of 2003. | 15:35 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * b199658 / src/configure: https://github.com/shogun-toolbox/shogun/commit/b1996589b0a12851e12d348a0934d6e2364f6260 | 15:36 |
shogun-notifier- | shogun: fix c++11 error | 15:36 |
shogun-buildbot | build #1491 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1491 blamelist: Soeren Sonnenburg <sonne@debian.org> | 15:37 |
lambday | thoralf: is it global avg? | 15:37 |
@iglesiasg | thoralf, almost there then! | 15:37 |
thoralf | lambday: No, Germany. | 15:37 |
thoralf | lambday: Sorry, Europe. | 15:38 |
thoralf | lambday: Again, sorry. The statistics for 2003 are for Germany only. | 15:38 |
lambday | thoralf: hehe | 15:39 |
lambday | 1.647 m for India according to wiki | 15:39 |
shogun-buildbot | build #1493 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1493 blamelist: Soeren Sonnenburg <sonne@debian.org> | 15:43 |
lisitsyn | iglesiasg: well I was more like dwarf | 15:46 |
-!- travis-ci [~travis-ci@ec2-184-73-148-97.compute-1.amazonaws.com] has joined #shogun | 15:46 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9351577 | 15:46 |
-!- travis-ci [~travis-ci@ec2-184-73-148-97.compute-1.amazonaws.com] has left #shogun [] | 15:46 | |
lisitsyn | ;) | 15:46 |
gsomix | lisitsyn, http://images3.wikia.nocookie.net/__cb20120922121059/lotr/images/c/c4/Balin_-_The_Hobbit.PNG | 15:50 |
gsomix | you need a beard | 15:50 |
lisitsyn | iglesiasg: I don't remember, am I taller? | 15:52 |
lambday | HeikoS: regarding COCG_M, I have a problem and a solution | 16:01 |
lambday | HeikoS: its tough to get all the solution vectors to be returned by our solve method, because the linear operator is real, the vector is real but just because of the complex shifts the solution is complex | 16:02 |
@iglesiasg | lisitsyn, I think you are taller than me yeah | 16:02 |
@iglesiasg | lisitsyn, I am about 173cm | 16:02 |
lambday | HeikoS: so I can't extend the base properly and its hard to fit in the system.. but I want to keep the base same (CLinearSolver) | 16:02 |
lisitsyn | iglesiasg: I am around that too | 16:02 |
lisitsyn | but 2x of you anyway | 16:02 |
lisitsyn | haha | 16:02 |
@iglesiasg | :P | 16:03 |
lambday | HeikoS: so, I am planning to add another level of abstraction - extending CIterativeLinearSolver (subclass of CLinearSolver) to CLinearShiftedFamilySolver... where shifts can be real/complex (templated) | 16:03 |
lisitsyn | iglesiasg: in other dimension | 16:03 |
lisitsyn | :D | 16:03 |
lambday | HeikoS: then, in COCG_M implementation of that, solve just returns the basic solution (similar to what Krylstat does) | 16:04 |
lambday | HeikoS: and there will be two abstract methods for specially handling the shifts.. | 16:04 |
lambday | so, in the COCG_M job, we'll first solve (which computes all of these and stores the solution matrix inside it).. and a subsequent get call will return the actual solution for the whole family | 16:05 |
@iglesiasg | sonney2k, did you have time to have a look at the new LMNN PR? https://github.com/shogun-toolbox/shogun/pull/1270 | 16:06 |
@iglesiasg | sonney2k, I addressed the issues you pointed out a some days ago. I think it should be ready to merge | 16:06 |
lambday | HeikoS: we can use this for CG_M/COCG_M as needed (although not for our project but may be later) - will be flexible | 16:06 |
lambday | HeikoS: the only downside is that, we'll have to remember calling "solve" before we call the "get" each time - but that's not a problem I guess | 16:07 |
lambday | or, we'll move solve inside that "get", call it "solve_shifted" or something, and in COCG_M jobs, call that instead | 16:07 |
lambday | HeikoS: I can finish this implementaion by tonight and send a PR and then may be you can have a look (not much work so I won't get sad if I gotta change it later :D) | 16:13 |
-!- travis-ci [~travis-ci@ec2-67-202-44-165.compute-1.amazonaws.com] has joined #shogun | 16:13 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9351711 | 16:13 |
-!- travis-ci [~travis-ci@ec2-67-202-44-165.compute-1.amazonaws.com] has left #shogun [] | 16:13 | |
lambday | HeikoS: brb | 16:18 |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 16:21 | |
-!- travis-ci [~travis-ci@ec2-67-202-44-165.compute-1.amazonaws.com] has joined #shogun | 16:39 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9352112 | 16:39 |
-!- travis-ci [~travis-ci@ec2-67-202-44-165.compute-1.amazonaws.com] has left #shogun [] | 16:39 | |
-!- iglesiasg [~iglesias@n131-p244.kthopen.kth.se] has quit [Ping timeout: 245 seconds] | 16:51 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:788e:31ff:81c6:5bc9] has joined #shogun | 17:04 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 17:06 | |
thoralf | gsomix: Doxygen gives a lot of warnings for your classes. Mostly because the variables in @param are not matching the signatures. | 17:08 |
thoralf | gsomix: Reproducable when building i.e. python_static | 17:09 |
gsomix | thoralf, woops. | 17:10 |
gsomix | I'll see, thanks. | 17:10 |
thoralf | gsomix: Just check it when you're finished - this always happens with heavily chaning classes :) | 17:11 |
-!- foulwall [~user@dirtycod.es] has joined #shogun | 17:14 | |
-!- lambday [67157f36@gateway/web/freenode/ip.103.21.127.54] has quit [] | 17:16 | |
gsomix | sonney2k, good evening, mr. Soeren. | 18:11 |
gsomix | sonney2k, I have prepared pack of changes for shogun. have you a minute to look at it? | 18:12 |
pickle27 | hey guys | 18:13 |
pickle27 | having the weirdest problem here | 18:13 |
pickle27 | eigen3 will not transpose a matrix | 18:13 |
pickle27 | like the code runs but it doesn't get transposed | 18:13 |
pickle27 | anyone else ever seen this? | 18:13 |
@iglesiasg | pickle27, what are you doing exactly? | 18:14 |
lisitsyn | pickle27: haha nice | 18:14 |
pickle27 | Im computing eigenvectors and then I need to transpose them | 18:14 |
@iglesiasg | pickle27, A.transpose() makes the transpose of A by doesn't assign to it, I believe | 18:14 |
pickle27 | yeah you need to use transposeInPlace() | 18:15 |
lisitsyn | yes A.transpose() is expression | 18:15 |
gsomix | sonney2k, https://github.com/shogun-toolbox/shogun/pull/1271 | 18:15 |
pickle27 | but that doesn't work either | 18:15 |
@iglesiasg | aham! | 18:15 |
gsomix | van51, hey. there are some changes with line reader in my PR. | 18:15 |
@iglesiasg | so transposeInPlace() is like A=A.tranpose() | 18:15 |
pickle27 | yeah it explicitly says in the docs not to use A = A.transpose() | 18:15 |
lisitsyn | pickle27: too weird to be true! | 18:15 |
gsomix | van51, I replaced it with LineReader. more useful class. | 18:15 |
lisitsyn | I use transposeInPlace in my code | 18:16 |
lisitsyn | it works | 18:16 |
pickle27 | it usually works for me too | 18:16 |
van51 | gsomix: you mean FileReader? | 18:16 |
pickle27 | but it isn't in this case | 18:16 |
pickle27 | and I had another time the other day when it wouldn't either | 18:16 |
@iglesiasg | pickle27, x-files! | 18:16 |
pickle27 | hmmmm | 18:21 |
pickle27 | maybe it is transposing ... | 18:22 |
@iglesiasg | pickle27, are you doing array() or anything else on the way? | 18:22 |
pickle27 | its confusing because of the values its a very symmetric thing | 18:22 |
pickle27 | i think I need to sort actually instead | 18:22 |
gsomix | van51, oops, exactly. :) | 18:23 |
van51 | gsomix: ok! thanks for letting me know :0 | 18:23 |
van51 | :)* | 18:23 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 18:36 | |
-!- foulwall [~user@dirtycod.es] has quit [Ping timeout: 248 seconds] | 18:56 | |
lisitsyn | pickle27: iglesiasg: gsomix: van51: guess what | 19:12 |
lisitsyn | ;) | 19:12 |
pickle27 | reports | 19:12 |
@iglesiasg | lisitsyn, almost written! :) | 19:12 |
pickle27 | its coming! | 19:12 |
@iglesiasg | pickle27, race! :P | 19:12 |
pickle27 | iglesiasg: I haven't started writting it yet, trying to fix up some code first | 19:13 |
@iglesiasg | pickle27, aah ok. Next week then ;) | 19:14 |
gsomix | lisitsyn, reports time! http://www.youtube.com/watch?v=E5KC1E5NyR0 | 19:16 |
@iglesiasg | gsomix, dafuq O_O | 19:17 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 19:21 | |
gsomix | iglesiasg, cool cartoon. I'm highly recommend. | 19:21 |
gsomix | cartoons is for strong man :3 | 19:22 |
gsomix | are | 19:22 |
@iglesiasg | lol | 19:22 |
gsomix | I need a book on English grammar =___= | 19:23 |
pickle27 | hey guys | 19:25 |
pickle27 | chasing down a slight bug in my program | 19:25 |
pickle27 | http://pastebin.com/7K9rxHZq | 19:25 |
pickle27 | is this a rotation? | 19:26 |
@iglesiasg | pickle27, mmmm (thinking) | 19:28 |
@iglesiasg | pickle27, I think that if you regard the first matrix as a rotation matrix of angle theta, then the second matrix is the rotation matrix of angle theta plus or minus (not sure about the sign) pi/2 | 19:36 |
pickle27 | thats what I was thinking | 19:36 |
pickle27 | I'll try that out | 19:37 |
pickle27 | okay | 20:26 |
pickle27 | narrowing this down further | 20:26 |
pickle27 | I have sign differences in my eigenvectors between matlab and eigen3 | 20:26 |
pickle27 | any idea how to fix this? | 20:27 |
pickle27 | looks like the eigen3 equivalent of matlabs eig might be shcur | 20:28 |
pickle27 | still has sign mismatch | 20:33 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 256 seconds] | 20:36 | |
@HeikoS | sonney2k: sorry have to postpont builds to tomorrow, but then for sure | 20:38 |
gsomix | van51, hey | 20:43 |
gsomix | I have question | 20:44 |
van51 | gsomix: go ahead | 20:44 |
gsomix | webspam_example is your code? it's now at develop branch& | 20:44 |
gsomix | *& | 20:44 |
gsomix | agrh! | 20:44 |
gsomix | *? | 20:44 |
gsomix | ahh | 20:45 |
gsomix | sorry, no more questions | 20:45 |
van51 | gsomix: yes it's my code | 20:46 |
van51 | gsomix: but it's not submitted | 20:46 |
van51 | gsomix: I lost you somewhere :p | 20:46 |
gsomix | it's just in my local rep, ok. | 20:46 |
van51 | gsomix: ah ok | 20:47 |
gsomix | all ok now | 20:48 |
van51 | gsomix: I also have to adjust it to your new class | 20:48 |
van51 | gsomix: why are you doing this to me | 20:48 |
-!- lisitsyn [~lisitsyn@213.87.128.238] has joined #shogun | 20:48 | |
van51 | :P | 20:48 |
pickle27 | lisitsyn: narrowed down my problem | 20:49 |
pickle27 | sign differences on the eigenvectors found using eigen3 | 20:49 |
pickle27 | isn't the same result as octave's eig() function | 20:49 |
lisitsyn | pickle27: well different signs are kind of ok | 20:50 |
pickle27 | usually yeah but not for Jade apparently | 20:50 |
gsomix | van51, huh. life is pain. >:3 | 20:50 |
lisitsyn | pickle27: why? | 20:50 |
gsomix | van51, btw free testing. | 20:51 |
pickle27 | messes the cumulant matrices | 20:51 |
lisitsyn | pickle27: do you need positive eigenvalues? | 20:51 |
pickle27 | I don't neccessarily need positive ones I just need the same signs as matlab | 20:51 |
lisitsyn | pickle27: are they all positive with matlab? | 20:52 |
lisitsyn | or negative | 20:52 |
pickle27 | lisitsyn: http://pastebin.com/GZXs8RR8 | 20:53 |
lisitsyn | pickle27: what about eigenvalues? | 20:54 |
pickle27 | 0.259276 and 2.15752 in both eigen3 and octave | 20:55 |
lisitsyn | pickle27: but different eigenvectors? weird! | 20:56 |
pickle27 | Im just going to flip the sign if they are all neg to be pos | 20:57 |
@iglesiasg | lisitsyn, is it so weird that there are several eigenvectors for the same eigenvalue? | 21:12 |
lisitsyn | iglesiasg: I think so | 21:16 |
@iglesiasg | lisitsyn, I just looked it up. Let me write it | 21:17 |
@iglesiasg | lisitsyn, Given that v is an eigenvector of a matrix A, then any scaled eigenvector c*v is also an eigenvector for the same eigenvalue. | 21:17 |
@iglesiasg | I thinks it makes pretty much sense | 21:18 |
lisitsyn | iglesiasg: true | 21:18 |
lisitsyn | yeah my bad | 21:18 |
@iglesiasg | lisitsyn, it was good to recall the property :) | 21:18 |
lisitsyn | so signs are whatever :) | 21:18 |
pickle27 | yeah except it is affecting the algorithm ahhhhh | 21:19 |
@iglesiasg | true, the constant can be negative | 21:19 |
@iglesiasg | pickle27, maybe one of the implementations normalize the eigenvectors so they are unit | 21:19 |
lisitsyn | pickle27: then something like positivity is forced | 21:19 |
pickle27 | hmm yeah could be that | 21:19 |
pickle27 | it doesn't make any sense | 21:20 |
pickle27 | I have 2 examples one ported from scikit learn and one that andreas gave me | 21:20 |
pickle27 | in the scikit one eigen gives mainly neg eigenvectors and the orig alg gives positive | 21:21 |
pickle27 | then its the opposite on andreas' example | 21:21 |
lisitsyn | pickle27: I wonder why sign matters here | 21:24 |
lisitsyn | if it operates eigenvectors it should be ready to sign changes | 21:25 |
pickle27 | Im not sure | 21:28 |
lisitsyn | prince is born | 21:34 |
lisitsyn | hehe | 21:34 |
@iglesiasg | lisitsyn, what? | 21:35 |
@iglesiasg | aaah UK? | 21:35 |
lisitsyn | iglesiasg: yeah | 21:36 |
-!- lisitsyn [~lisitsyn@213.87.128.238] has quit [Ping timeout: 240 seconds] | 21:42 | |
gsomix | good night guys | 21:44 |
@sonney2k | van51, please fix the warnings travis reports https://travis-ci.org/shogun-toolbox/shogun/jobs/9352117 | 21:44 |
-!- lisitsyn [~lisitsyn@213.87.128.238] has joined #shogun | 21:44 | |
-!- lisitsyn [~lisitsyn@213.87.128.238] has quit [Client Quit] | 21:44 | |
@sonney2k | van51, btw did C=1 give a more similar result? | 21:44 |
van51 | sonney2k: yea the speedup isn't that bigger with C=1 | 21:46 |
van51 | sonney2k: I think it's about the half in some cases | 21:47 |
@sonney2k | van51, I meant the auROC value... | 21:47 |
van51 | sonney2k: yeah the same pretty much | 21:47 |
van51 | sonney2k: I will make a table so that they can be compared more easily | 21:47 |
@sonney2k | van51, in that case use a more relaxed epsilon please | 21:47 |
@sonney2k | 1e-2 | 21:47 |
@sonney2k | ocas is using a rather tough criterion, svmlight certainly is not as accurate | 21:48 |
-!- nube [~rho@49.244.18.229] has quit [Ping timeout: 246 seconds] | 21:48 | |
van51 | sonney2k: ok I will do that as well and I'll give a summary by tomorrow | 21:48 |
van51 | sonney2k: I have to go for now, but I'll make a PR for that warning when I get back | 21:49 |
van51 | sonney2k: I 'll also have one for quadratic on hashed dense features | 21:50 |
@sonney2k | van51, cool! | 21:51 |
@sonney2k | cu then | 21:51 |
@sonney2k | thoralf, do you have clang installed? | 21:52 |
van51 | cu | 21:55 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has quit [Quit: Leaving.] | 21:55 | |
@sonney2k | iglesiasg, do you have clang? | 21:55 |
@iglesiasg | sonney2k, of course! | 21:56 |
@iglesiasg | best compiler ever :) | 21:56 |
@iglesiasg | have you seen the error/warning messages it gives? I wish gcc worked like that | 21:56 |
@iglesiasg | actually, I just like it because of that, no idea if it is better/worse thatn gcc in other aspects | 21:57 |
-!- HeikoS [~heiko@nat-180-24.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 22:00 | |
-!- nube [~rho@49.244.18.229] has joined #shogun | 22:02 | |
@sonney2k | iglesiasg, could you please run configure with latest shogun and --cxx=clang++ ? | 22:04 |
@sonney2k | iglesiasg, and I agree about the quality of warning/error messages | 22:04 |
@iglesiasg | sonney2k, sure, I thought you were never going to ask :) | 22:04 |
@iglesiasg | sonney2k, should I enable any part in particular? | 22:05 |
@sonney2k | iglesiasg, interfaces= is sufficient | 22:05 |
@iglesiasg | all right | 22:05 |
@sonney2k | iglesiasg, well you know I am on vacations so I am distracted even more | 22:06 |
@iglesiasg | sonney2k, how are your holidays going btw? Lot of sun-bathing and swimming? | 22:06 |
@sonney2k | I think i was only doing a ~2km swim today but the crazy thing is that my feet hurt from the walking back | 22:07 |
@iglesiasg | hehe | 22:07 |
@iglesiasg | sonney2k, I noticed Checking for C++11 Atomic Integer ... no, should it be yes? | 22:08 |
@sonney2k | iglesiasg, which version is it? | 22:09 |
@iglesiasg | sonney2k, clang? | 22:10 |
@sonney2k | yes | 22:10 |
@iglesiasg | Ubuntu clang version 3.0-6ubuntu3 | 22:10 |
@sonney2k | ahh ok too old | 22:11 |
@sonney2k | 3.1 introduce atomic | 22:11 |
@iglesiasg | aham I see | 22:12 |
@iglesiasg | sonney2k, did you want to check that thing or it doesn't matter? | 22:12 |
@sonney2k | and it seems one needs clang3.2 for anything reasonable | 22:14 |
@sonney2k | well would be nice | 22:14 |
@sonney2k | maybe wiking has a newer clang? | 22:14 |
@iglesiasg | aham I see | 22:14 |
@iglesiasg | sonney2k, I can check in my laptop too. I tried in the office computer so far | 22:15 |
@sonney2k | iglesiasg, check clang++ -v first | 22:15 |
@sonney2k | if the version is recent enough | 22:15 |
@iglesiasg | sonney2k, 3.2.1 in my laptop | 22:16 |
@sonney2k | iglesiasg, good enough | 22:17 |
@sonney2k | iglesiasg, so please try | 22:17 |
@sonney2k | it should detect c++11 | 22:17 |
@iglesiasg | sonney2k, it did indeed | 22:18 |
-!- zxtx [~zv@rrcs-76-79-81-162.west.biz.rr.com] has joined #shogun | 22:18 | |
@iglesiasg | sonney2k, I see c++ Atomic integer yes, but no c++11 flag in the list of c++ compiler flags | 22:19 |
@iglesiasg | sonney2k, I am not sure if it should appear there too | 22:19 |
@sonney2k | iglesiasg, well type make :) | 22:20 |
@iglesiasg | sonney2k, compiling... | 22:21 |
@sonney2k | iglesiasg, done? | 22:29 |
@iglesiasg | sonney2k, nope, still compiling | 22:29 |
@iglesiasg | sonney2k, done | 22:33 |
@iglesiasg | sonney2k, what now? | 22:34 |
-!- FSCV [~FSCV@50.7.50.60] has quit [Ping timeout: 245 seconds] | 22:41 | |
@sonney2k | iglesiasg, all good! | 22:42 |
@iglesiasg | sonney2k, all right! | 22:42 |
@iglesiasg | sonney2k, btw, did you have a moment to check the LMNN PR? | 22:42 |
@sonney2k | iglesiasg, no :/ | 22:44 |
@sonney2k | gsomix, has sent a >1kLoC monster PR | 22:45 |
@iglesiasg | sonney2k, hehe I read your comment | 22:45 |
@sonney2k | and he has prio | 22:45 |
@iglesiasg | sonney2k, then for another day maybe, mine is not precisely short actually. I think it is about 1k (including two class headers and two files with unit tests though) | 22:46 |
@sonney2k | iglesiasg, maybe lisitsyn or so has time? | 22:48 |
@sonney2k | I ahve to sleep now due to too nice weather I am occupied swimming at lunch break | 22:49 |
@sonney2k | cu! | 22:49 |
@iglesiasg | sonney2k, he already took a look and said it was fine | 22:49 |
@iglesiasg | sonney2k, bye! | 22:50 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 23:00 | |
shogun-notifier- | shogun: Fernando Iglesias :develop * cefe77a / src/shogun/ (6 files): https://github.com/shogun-toolbox/shogun/commit/cefe77a37bfc55d5144225c712ea783d4c853e4d | 23:00 |
shogun-notifier- | shogun: LMNN implementation based on auxiliary methods hidden in an implementation class. | 23:00 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 2913219 / tests/unit/metric/LMNNImpl_unittest.cc,tests/unit/metric/LMNN_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/29132198cd1ce7d080510ecf0f4d3bcd1f56f5c8 | 23:00 |
shogun-notifier- | shogun: A few unit tests for LMNN | 23:00 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 263c8d7 / / (8 files): https://github.com/shogun-toolbox/shogun/commit/263c8d7d4b13a40bc93b8e9ac5d23457a2a57a93 | 23:00 |
shogun-notifier- | shogun: Merge pull request #1270 from iglesias/feature/lmnn | 23:00 |
shogun-notifier- | shogun: | 23:00 |
shogun-notifier- | shogun: LMNN implementation and a couple of unit tests | 23:00 |
shogun-buildbot | build #1494 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1494 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 23:03 |
shogun-buildbot | build #1495 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1495 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 23:17 |
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has joined #shogun | 23:22 | |
@iglesiasg | hey hushell | 23:22 |
hushell | hey iglesiasg | 23:22 |
hushell | I just got connected | 23:22 |
@iglesiasg | hushell, sure, I think I will be able to have a look at it by tomorrow morning | 23:22 |
hushell | I have a meeting with Patrick tonight | 23:23 |
hushell | maybe he has more comments on it | 23:23 |
@iglesiasg | hushell, hehe tonight when? | 23:23 |
@iglesiasg | I get confused when talking to people in different timezones :) | 23:23 |
@iglesiasg | how many hours from now you mean with tonight? | 23:23 |
hushell | 6 or 7 hours later | 23:23 |
@iglesiasg | ok | 23:23 |
@iglesiasg | hushell, so with tomorrow morning you mean about the time when the meeting for the mid-term evaluation takes place, right? | 23:24 |
hushell | but I don't think there are will be many things need to revise for him | 23:24 |
hushell | What time for you right now? | 23:25 |
@iglesiasg | hushell, 23h25 | 23:25 |
@iglesiasg | or 11:25PM | 23:25 |
hushell | yeah, I guess right :) so I say tomorrow morning your time | 23:25 |
@iglesiasg | hushell, ok got it :) | 23:26 |
hushell | iglesiasg: okay, talk you later | 23:26 |
@iglesiasg | hushell, all right, bye! | 23:27 |
hushell | night! | 23:27 |
@iglesiasg | see you later, people | 23:35 |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:788e:31ff:81c6:5bc9] has quit [Quit: Ex-Chat] | 23:35 | |
-!- travis-ci [~travis-ci@ec2-54-224-68-246.compute-1.amazonaws.com] has joined #shogun | 23:47 | |
travis-ci | [travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9368929 | 23:47 |
-!- travis-ci [~travis-ci@ec2-54-224-68-246.compute-1.amazonaws.com] has left #shogun [] | 23:47 | |
--- Log closed Tue Jul 23 00:00:43 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!