Open in new window / Try shogun cloud
--- Log opened Thu Sep 06 00:00:17 2012
shogun-buildbotbuild #89 of nightly_default is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/8903:39
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun07:17
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]07:48
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun08:07
sonne|workblackburn: we forgot heiko in NEWS!08:48
-!- uricamic [~uricamic@2001:718:2:1634:c1c6:c21e:4fd:ba07] has joined #shogun08:51
-!- uricamic1 [~uricamic@2001:718:2:1634:c1c6:c21e:4fd:ba07] has joined #shogun08:55
-!- uricamic [~uricamic@2001:718:2:1634:c1c6:c21e:4fd:ba07] has quit [Read error: Connection reset by peer]08:55
blackburnsonne|work: oh well08:56
sonne|workwe better mention him in 2.0.1 :D08:56
blackburnsonne|work: 9 critical level tickets here08:59
blackburnI start to hate that08:59
sonne|workhttp://freecode.com/projects/shogun/date_metrics09:04
blackburnsonne|work: LLTSA author? heheh09:51
sonne|worknice right :)10:21
sonne|workblackburn: I thought this is the motivation you need.10:22
blackburnyeah kind of10:22
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun10:35
blackburnsonne|work: and what is number of downloads?10:36
sonne|workno idea10:37
sonne|workwe don't monitor this10:37
sonne|workusually between 100 and 300 unique visitors per days (has the website)10:37
blackburnsonne|work: we've got no mentions on any website this time :(10:42
sonne|workblackburn: we didn't write any blog post yet...10:43
sonne|workblackburn: and we don't have nice figures - so it is hard for me to blog this time10:45
-!- cronor [~cronor@fb.ml.tu-berlin.de] has joined #shogun10:47
blackburnsonne|work: I have free 3 days from tomorrow may be I'll come with something10:55
blackburnfree from that damned10:55
blackburnsupport10:55
blackburnn4nd0: can HMM recognize strategy of flipping a coin?11:07
blackburni.e. if we learn it on11:07
blackburn010101010111:07
blackburnor11:07
blackburn1111110000011:07
blackburnor 00011100011100011:08
blackburnsee what I mean?11:08
n4nd0blackburn: do you always use the same coin?11:08
n4nd0?11:08
blackburnn4nd0: oh, wait11:09
blackburnI think it is not a fair coin :D11:09
n4nd0I mean if the probability of tail and head is always the same11:09
blackburnwhat I meant is can it learn to recognize that?11:09
blackburn3 zeros, 3 ones11:09
n4nd0is the prob. of tail and head always the same?11:09
blackburn4 zeros, 4 ones11:09
blackburnetc11:09
blackburnlet it be not a coin anymore11:10
n4nd0??11:10
blackburn:D11:10
n4nd0haha ok...11:10
blackburnforget what I said before, from the beginining11:10
blackburnI show you one or two fingers11:10
blackburnwith some strategy I know11:10
blackburncan you learn a HMM that will adapt to my strategy?11:10
n4nd0yes11:11
n4nd0there is a couple of details though11:11
n4nd0the thing would be like11:11
n4nd0sometimes you follow one strategy where the prob. of 1 finger is x and of two fingers is 1-x11:11
n4nd0other times you follow another strategy where the prob. ...11:12
n4nd0you may change between strategies11:12
blackburnI am just thinking of graphical example of HMMs11:12
n4nd0the only thing I can see is the fingers you show me11:12
n4nd0that can be modelled easily with an HMM11:12
n4nd0blackburn: do you mean to put some nice related to HMMs in the webpage or in graphical examples?11:14
blackburnn4nd0: webpage/blog post11:14
n4nd0aham ok11:14
n4nd0they use HMMs in bioinformatics quite a bit11:15
n4nd0and they have nice plots11:15
n4nd0maybe sonne|work can point out something useful for this11:15
blackburnn4nd0: yeah but we would need 4 state model11:15
blackburnor we end with limited thymine and adenine alphabet :D11:16
blackburnn4nd0: btw do you know why 4 aminoacids are there?11:16
n4nd0I don't know anything about aminoacids11:17
sonne|workblackburn: what is wrong with the loaded cube example (that we already have)?11:19
blackburnsonne|work: ?11:19
sonne|workI created an example where 3 dices (all loaded) are drawn11:19
blackburnn4nd0: a little more than 3 is the optimal base for a numeral system11:20
sonne|workthe HMM is used to learn the distribution that one can draw from each dice and when which dice is used11:20
blackburnsonne|work: I've never seen that example :D11:20
blackburnwhere?11:20
sonne|workoctave_static/graphical/hmm.m11:21
sonne|workexample is probably 8 years old :D11:21
blackburnsonne|work: well, there is nothing wrong with it except I've never seen that working :)11:26
blackburnI'll run it once I get home11:27
sonne|workyeah old code in static if's11:27
blackburnthat would take a few hours today, switching city11:27
blackburnI wish my mom gave me a teleport11:27
-!- cronor [~cronor@fb.ml.tu-berlin.de] has quit [Quit: cronor]11:58
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed]14:20
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]14:42
-!- uricamic1 [~uricamic@2001:718:2:1634:c1c6:c21e:4fd:ba07] has left #shogun []14:43
-!- dan_ [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has joined #shogun15:08
dan_Hi, I am trying to install shogun with MOSEK and CPLEX support but ./configure fails to find those lib15:09
dan_How can I add "/usr/local/Aimms/3.12.6/Solvers" to shogun links ? thats the path where libmosek.so, libcplex.so are. Thanks.15:10
sonne|workdan_: ./configure --libs15:11
dan_thanks but MOSEK and CPLEX support are still not found ..15:15
dan_Is there a configure output example with CPLEX support ?15:49
-!- hoijui [~hoijui@141.23.65.251] has joined #shogun15:52
sonne|workdan_: sry I don't have either...15:53
dan_thanks anyway15:53
-!- dan_ [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has quit [Quit: Page closed]15:53
-!- jogin [86600232@gateway/web/freenode/ip.134.96.2.50] has joined #shogun16:32
-!- jogin [86600232@gateway/web/freenode/ip.134.96.2.50] has quit [Client Quit]16:34
-!- jogin [86600232@gateway/web/freenode/ip.134.96.2.50] has joined #shogun16:35
joginHello, I have some data for training and afterwards testing by python interface. As I see in shogun examples, I need my data in *.dat format. And that's my question - how can I create *.dat format? Can I find somewhere a sample?17:01
@sonney2kjogin, no just load your data with python and pass a numpy matrix17:13
@sonney2k(and use the python_modular interface!!)17:13
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | Version 2.0.0 just released | When asking questions please be patient or use the mailing list. We always answer but it might take a while. | Channel logs http://bit.ly/eQDk8Q17:16
-!- heiko [~heiko@82.132.139.20] has joined #shogun17:16
-!- heiko [~heiko@82.132.139.20] has left #shogun []17:16
jogin@sonney2k: So, which object class to use for loading the data with Python?17:32
-!- jogin [86600232@gateway/web/freenode/ip.134.96.2.50] has quit [Quit: Page closed]18:05
-!- blackburn [~blackburn@109.226.119.190] has joined #shogun18:31
-!- hoijui [~hoijui@141.23.65.251] has quit [Ping timeout: 245 seconds]19:31
@sonney2kblackburn, alive?19:45
blackburnsonney2k: yeap19:45
blackburnsonney2k: what's up? wanna make me fixing that crash? :)19:46
@sonney2kblackburn, yes!19:56
@sonney2kdo it!19:56
-!- hoijui [~hoijui@141.23.65.251] has joined #shogun19:59
blackburnokayy let me pack my code for gsoc now and I will check20:00
blackburnlocality preserving projections right?20:03
@sonney2kblackburn, yes - fails when arpack is not available20:04
blackburnsonney2k: ahh okay20:05
blackburneasy tehn20:05
blackburnsonney2k: so you will visit NY?20:05
@sonney2kblackburn, not easy - it does sth in the atlas call20:11
blackburnsonney2k: hah20:11
@sonney2k...which fails20:11
blackburnsonney2k: still - easy20:12
@sonney2kthen please do it20:12
blackburndoing20:12
@sonney2kit is blocking me doing things on the new bb20:12
blackburnoh blocker issue20:12
blackburnI fscking love it20:12
blackburn:D20:12
blackburnQA always says the issue I work on blocks everything20:13
blackburnwe can't start nuclear war!20:13
@sonney2kluckily you slow down the process :)20:16
@sonney2kso war can wait :D20:17
@sonney2kblackburn, I just figured out why I got tons of errors on the web site20:17
blackburnsonney2k: why?20:17
@sonney2ksqlite does auto_increment per default on primary (integer!) keys20:17
@sonney2kbut mysql does not20:17
blackburnehm20:21
blackburnthen enable it20:21
blackburn:)20:21
@sonney2kthat is whay I did20:25
blackburnsonney2k: ah so it is just a bad matrix20:37
blackburnnot a real bug20:37
blackburnsonney2k: here you go20:40
CIA-35shogun: Sergey Lisitsyn master * r2e41877 / examples/undocumented/libshogun/converter_localitypreservingprojections.cpp : Better data matrix for LPP example - http://git.io/2FlrvQ20:40
@sonney2kblackburn, ?20:42
@sonney2kwhat is a bad matrix?20:42
blackburnsonney2k: should work now20:42
blackburnsonney2k: 0 1 2 3 4 5 6 ...20:42
blackburnis a bad feature matrix20:42
blackburnfor *any* dim red algorithm20:42
@sonney2kso data was too tough for this example to ever work?20:42
blackburnyes20:42
blackburnunsolvable generalized eigenvalue problem20:42
blackburnwe could regularize but I am lazy to integrate regularization to LPP20:43
blackburnand I am not sure it is not an idiot way20:43
@sonney2kyeah ok good enough now20:44
@sonney2kI fixed a few issues on the website20:44
blackburnsonney2k: I'll try to come up with pics and may be skeleton of post20:48
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun20:48
blackburnmighty fernando20:49
@sonney2kn4nd0, hey there20:50
-!- hoijui [~hoijui@141.23.65.251] has quit [Ping timeout: 260 seconds]20:50
n4nd0sonney2k: hi20:50
n4nd0hi all :)20:50
@sonney2kI guess it is time that you both blackburn, n4nd0 do some edits on the website yourself20:50
blackburnsonney2k: give me give me20:50
n4nd0aham ok20:51
n4nd0sonney2k: what do you want to see new?20:51
blackburnsonney2k: may be we could afford to put ocr example online?20:51
blackburnhow to make an octave plot stay?20:53
shogun-buildbotbuild #8 of rpm1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/820:54
blackburn--persist okay20:55
blackburnsonney2k: n4nd0: have you seen that ticket? http://developer.pidgin.im/ticket/1530820:56
@sonney2kblackburn, yay! you did it ;)20:57
blackburnsonney2k: I've run hmm graphical exmaple20:58
blackburnsonney2k: what is going on here?20:58
blackburnshould estimated path be equal to true path?20:58
blackburnsonney2k: https://dl.dropbox.com/u/10139213/shogun/Screenshot%20from%202012-09-06%2022%3A58%3A51.png20:59
blackburnI am a kind of lobotomy victim too so hard to understand21:00
blackburnmeanwhile in the scikits ML guys discuss large scale data - 1G21:01
blackburnlol21:01
@sonney2kblackburn, yes please add the flash thing under the applications link21:03
@sonney2k(good exercise btw :)21:03
@sonney2kblackburn, regarding hmm21:03
blackburnsonney2k: no I meant may be21:04
@sonney2kperfect result!21:04
blackburnput it online really?21:04
@sonney2kblackburn, not maybe do it!21:04
blackburn*active*21:04
@sonney2kactive?21:04
@sonney2kn4nd0, I want to see fixes ... the site has many bad/broken links and missing info21:04
blackburnyes, I think how difficult is to make an java application or so21:04
blackburncan we run a java application out there?21:05
@sonney2kblackburn, no just the flash back is sufficient!21:05
@sonney2kno java crap21:05
blackburnwhat is not a crap?21:05
@sonney2k(flash crap is bad enough)21:05
@sonney2kexactly21:05
@sonney2kblackburn, regarding hmm - it shows the states 1...3 corresponding to when which cube was drawn21:06
blackburnsonney2k: so hmm is not broken?21:06
@sonney2kblackburn, so it is *perfect* result21:06
@sonney2kwhy should it be broken?21:06
@sonney2konly parallel hmm did not work21:06
blackburnno, tests are failing still21:07
@sonney2kahh and note that it cannot know whether dice 2 is dice 1 or vice versa21:07
@sonney2kso things might be swapped (just have different name)21:07
@sonney2kand distributions should also match pretty well21:07
blackburnI feel braindamaged21:08
blackburn:D21:08
blackburntime to do something stupid21:08
blackburnahh I wanted to make edrt use eigen321:08
@sonney2kblackburn, but don't commit this in master branch...21:10
@sonney2kbugfix mode for now21:10
blackburnsonney2k: I will once it is ready21:10
blackburnin month or two21:10
blackburn:D21:10
blackburnrecall it is my prolonged gsoc project?21:10
@sonney2kwe should probably really create a dev branch an do things there21:11
@sonney2kblackburn, but more important we need a polished web site21:11
blackburnyeah I received a strange passphrase21:11
blackburnfor schtazi21:11
blackburnknock-knock who's there21:12
@sonney2kn4nd0, too of course21:12
@sonney2kheiko already tried to do some edits :D21:13
n4nd0it will be my first contact with django!21:14
@sonney2kn4nd0, there is no django contact here ... just editing pages21:14
@sonney2k*argh* I overwrote the doc page with the unfixed one21:14
@sonney2kand there is no versioning nor undo :/21:15
blackburnsonney2k: what is username for me? :D21:15
@sonney2ksergey21:16
@sonney2kblackburn, you didn't read your emails *tsss*21:17
blackburnsonney2k: can you find a mail where you mentioned I am a sergey?21:17
blackburn:D21:17
blackburnI received only schtazi passphrase21:18
n4nd0blackburn: where do you log in?21:18
blackburnadmin/21:18
blackburnn4nd0: ^21:18
blackburnadmin!!!/!!!21:18
blackburn:D21:18
@sonney2kheh21:20
@sonney2kn4nd0, for example we need links to the *latest* documentation, the current one and basically all older versions...21:21
@sonney2k0.10.0/  1.0.0/  1.1.0/  2.0.0/  2.0.1/  current@  latest@21:21
n4nd0ok21:21
n4nd0sonney2k: why is it there a current and latest version?21:40
n4nd0is latest cutting edge code in the repo and current the latest release?21:41
blackburnexacltly21:41
n4nd0blackburn: thanks ;)21:42
blackburnsonney2k: do you know what is a 'position paper'?21:42
blackburnfor conference21:42
blackburnah found in description21:43
blackburnA position paper presents an arguable opinion about an issue. The goal of a position paper is to convince the audience that your opinion is valid and worth listening to, without the need to present completed research work and/or validated results. It is, nevertheless, important to support your argument with evidence to ensure the validity of your claims. 21:43
n4nd0sonney2k: here we go with links to the different doc versions21:48
n4nd0http://shogun-toolbox.org/page/documentation/information21:48
n4nd0time to bed21:53
n4nd0good night!21:53
blackburnn4nd0: may you have no segfaults in your dreams21:54
n4nd0haha21:54
n4nd0thanks21:55
blackburn:D21:55
n4nd0I won't leak either :P21:55
n4nd0exit21:55
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]21:55
blackburnmen do not leak21:55
blackburnwe've got to reduce some code I think21:58
-!- hoijui [~hoijui@dslb-088-074-107-101.pools.arcor-ip.net] has joined #shogun22:26
-!- blackburn [~blackburn@109.226.119.190] has quit [Quit: Leaving.]22:57
--- Log closed Fri Sep 07 00:00:10 2012