IRC logs of #shogun for Sunday, 2011-04-24

--- Log opened Sun Apr 24 00:00:36 2011
-!- sploving [sploving@210.77.26.145] has joined #shogun02:15
-!- sploving [sploving@210.77.26.145] has left #shogun []02:29
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun06:35
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has left #shogun []06:48
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun08:28
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds]08:31
-!- Ziyuan [~Ziyuan@116.22.51.87] has joined #shogun08:42
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 252 seconds]09:03
-!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun09:07
* siddharth exam :|09:21
-!- Ziyuan [~Ziyuan@116.22.51.87] has quit [Read error: Connection reset by peer]10:12
-!- Ziyuan [~Ziyuan@116.22.51.87] has joined #shogun10:12
-!- Ziyuan [~Ziyuan@116.22.51.87] has quit []11:03
-!- Ziyuan [~Ziyuan@121.33.190.190] has joined #shogun11:06
-!- Ziyuan [~Ziyuan@121.33.190.190] has quit [Remote host closed the connection]11:24
-!- Ziyuan [~Ziyuan@116.22.51.87] has joined #shogun11:24
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun12:17
-!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 260 seconds]12:25
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed]12:40
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun13:34
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed]14:31
-!- sploving [sploving@210.76.202.198] has joined #shogun14:31
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun14:39
-!- sploving [sploving@210.76.202.198] has quit [Ping timeout: 250 seconds]14:58
-!- sploving [sploving@210.76.202.198] has joined #shogun15:11
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 253 seconds]15:14
-!- Ziyuan [~Ziyuan@116.22.51.87] has quit []15:31
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun15:34
-!- sploving [sploving@210.76.202.198] has quit []15:44
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed]16:35
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun17:17
serialhexman!  my computer is down for a day and a half and no updates to shogun!  shame on you all!!!  :P17:28
serialhexsonney2k: i just read the back logs & i hope everyone feels better soon!17:28
blackburneh, no updates but some pulls ;)17:33
serialhexblackburn, yeah i read about your ROC troubles... did you get it all sorted out??17:36
blackburnseems so :)17:37
serialhexsounds good!17:37
blackburnnow doing some work on numerical methods, arghh, I hate it17:37
serialhexwhat work??17:37
* serialhex likes numerical methods17:37
blackburnhave to numerical integrate :)17:38
@bettyboo:)17:38
josipjust do the trapezoid method :D17:38
josipfor ROC17:38
serialhexthats not too hard, what integration method you using??17:39
blackburnnow finishing last part, kantorovich's method for integrating improper integrals17:39
* serialhex has never heard of kantorovich or his method if integration17:39
blackburnthat is not hard at all, but many methods and spent much time to write a report17:39
serialhexyou know blackburn, you dont have to do your schoolwork, gsoc project, _and_ our gsoc projects too... :P17:40
@bettyboo;> serialhex17:40
blackburneh? don't understand ;)17:41
serialhexin other words: dont work yourself to death!!! :D17:41
blackburnah, I just want to free my time on next week, I don't have to do this work now17:43
serialhexok, i can understand that, but still, you're not a machine!17:44
serialhex...or are you?17:44
* blackburn looks at himself17:45
blackburnno, seems no ;)17:45
serialhexok, good!17:46
blackburneh! it's cold on street17:47
blackburnwant to run a little :)17:48
@bettyboorahaha17:48
blackburndone that mthfcking report17:51
serialhexit's nice here... though i kind of miss a nice brisk day where it's not too cold & theres no wind, then you can run forever!17:51
serialhexcool!17:51
serialhexFFS!!! WHY  do all these journals have to charge an INSANE amount of money for A FSCKING PAPER!!!!!!!!!!17:52
blackburn:D17:54
serialhexOOOH!!!! it's like $30+ for ONE FREAKING PAPER!!!  and all i'm doing is trying to find out about kantorovich and his improper integral method!!!17:55
blackburnhttp://www.mathnet.ru/php/archive.phtml?wshow=paper&jrnid=tm&paperid=1054&option_lang=eng17:56
blackburnthere is no english version though17:57
blackburn:)17:57
serialhexooh!!! ,lets see if google can translate!!!17:57
serialhexoh, it's a pdf... :/17:58
blackburnyeap, 3317:58
blackburn1933 I mean17:59
blackburn+8C and I'm going to freeze my balls some :D18:00
serialhexhmm... given time i _might_ be able to divine something out of this from just the math...18:00
serialhexdamn!18:00
serialhex46.4f!!!! ffs thats cold!!!!18:01
* blackburn is out for some 'jogging' ;)18:01
serialhexhave fun!18:02
blackburnrhh! not so cold when running :D18:36
serialhexthats good to know!!19:04
blackburntime for shogun development!19:11
-!- siddharth_ [~siddharth@117.211.88.150] has joined #shogun19:12
serialhexgood evening siddharth_!19:15
siddharth_good evening19:15
siddharth_wassup19:15
serialhexnot too much, catching up on the things i missed out while i was wonrking on my computer19:16
serialhextrying to install gentoo, but i think i messed up the install19:17
siddharth_I see19:18
serialhexwhat about yourself?19:18
siddharth_just came back from dinner19:18
siddharth_so tomorrow is "the" day19:20
serialhexyes it is!19:20
serialhexi wonder who will get the slots19:21
siddharth_fingers crossed :)19:22
serialhexsame here!19:22
blackburnit will be at 12-00 GMT-8?19:23
serialhexi think so lemme check my calander19:23
serialhexyeah, it says 3pm on my calendar, and it'll be 12pm there, so you're right siddharth_19:24
serialhexbrb, laundry19:25
blackburnahaha I completely f-cked up with that ROC19:53
blackburnauthor of perfmeasure too :D19:53
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun20:01
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | We participate in GSoC with 5 slots | Timeline http://bit.ly/gy7Pdi | Channel logs http://bit.ly/eQDk8Q20:21
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | We participate in GSoC - student acceptance notification countdown http://goo.gl/JPGx3 5 | GSoC timeline http://bit.ly/gy7Pdi | Channel logs http://bit.ly/eQDk8Q20:23
blackburnhehe20:24
* sonney2k goes fixing a wheel on the baby buggy20:24
* sonney2k dammed epic wheel fixing fail - now for the same thing again21:19
blackburngood luck to you :)21:20
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun21:49
blackburnsonney2k: hurray! http://dl.dropbox.com/u/10139213/share/eq.png21:54
* sonney2k *sigh* another fail22:01
* sonney2k gives up22:01
@sonney2kcould be right...22:01
blackburnsonney2k: it is equal to one mldata-utils made22:02
@sonney2kahh - then great :)22:03
@sonney2kis this in the code from the pull request yesterday?22:03
blackburnyeap, I fixed it and then fixed the fixes :D22:03
@sonney2kblackburn, sounds like what I usually do :)22:04
@sonney2kblackburn, I am looking at your pull request right now22:04
blackburntoday was tangled with some issues with it22:04
@sonney2kdo you want comments here or there?22:04
blackburnI don't, just for you ;)22:05
@bettyboo:Q22:05
blackburnwill remove it just before you will merge22:05
@sonney2kblackburn, I don't understand...22:05
blackburnah22:06
blackburnHEHEEH22:06
blackburnwas thinking about comments in .cpp22:06
blackburn:D22:06
blackburnsonney2k: as you wish22:06
@sonney2kblackburn, then here: after for loops, please have one blank line22:07
blackburnokay22:07
@sonney2kblackburn, add a link or reference to the fawcett paper in the documentation22:08
blackburnsonney2k: to .cpp or .h?22:08
@sonney2k.h22:08
@sonney2kin .cpp I don't care that much about documentation22:09
@sonney2konly if you do hard to grasp stuff comments are welcome - otherwise having comments there hints that your code is unreadable.22:10
blackburnsonney2k: should I remove it in .cpp?22:10
@sonney2kblackburn, no it is ok22:13
@sonney2kblackburn, then if you do a for loop and a more complex if statement inside it (or another for loop inside) - please use { } to make it more readable22:14
@sonney2k(line 67)22:14
blackburnfixed22:15
@sonney2kblackburn, the pos_count*neg_count is dangerous - please use a cast to float64_t(pos_count)*neg_count22:17
blackburnsonney2k: okay22:18
blackburnsonney2k: is it ok?22:18
blackburnTom Fawcett "An Introduction to ROC analysis"22:18
@sonney2kyes22:18
blackburnshould I write some more information, e.g. journal or etc22:18
@sonney2kblackburn, yeah - copy the reference from wikipedia22:19
blackburnhehe, good idea22:19
@sonney2kthe get_ROC will crash - but that is ok for now - until the swig issue is fixed22:19
blackburnsonney2k: why it will?22:20
@sonney2kblackburn, memory not copied22:20
blackburnah.. I had a lot of problems during testing and etc22:20
@sonney2kblackburn, well if you want to have it work reliably you have to copy22:21
@sonney2klook at e.g. line 361 in SimpleFeatures.h22:21
@sonney2kfor how to do it22:21
blackburnsonney2k: oh, thank you22:21
blackburnwill do it now22:21
@sonney2kbut as I said at some point we have to fix this22:22
blackburnsonney2k: why it use malloc?22:24
blackburnah22:25
blackburnsorry ;)22:25
@sonney2kblackburn, python uses it to alloc things...22:25
blackburnsonney2k: meant why not new22:26
@sonney2kblackburn, python is not c++22:26
blackburneh.. okay22:26
blackburnsonney2k: have you any more remarks?22:29
blackburnhm.. it fails now22:29
@sonney2kblackburn, also spaces after if ()22:31
@sonney2k{22:31
@sonney2k}22:31
@sonney2kerr new lines22:31
blackburnsonney2k: done22:36
blackburnpushed22:36
@sonney2klooks good22:45
@sonney2kI was for the last 10 minutes looking for the merge help buttong - which apparently disappeard http://support.github.com/discussions/site/3548-merge-help-button-missing22:46
blackburnsomeone stealed it?22:46
CIA-110shogun: lionelc master * rbd60ad6 / (3 files): minor changes -- fixed whitespace issues - http://bit.ly/eI6CnY22:48
CIA-110shogun: Soeren Sonnenburg master * r7e1214f / (22 files): Merge branch 'master' of https://github.com/lionelc/shogun - http://bit.ly/hDAtci22:48
CIA-110shogun: Christian Widmer master * re6cec42 / (4 files): merged old changes from local svn repo into git repo - http://bit.ly/ffm6K622:48
CIA-110shogun: Soeren Sonnenburg master * rf75fb0b / (4 files): Merge branch 'master' of git://github.com/cwidmer/shogun - http://bit.ly/gOScuF22:48
blackburnthe big merge22:49
CIA-110shogun: Soeren Sonnenburg master * rea04aef / (3 files in 2 dirs): Merge git://github.com/SergeyLisitsyn/shogun - http://bit.ly/hLejTa22:49
blackburnsonney2k: thanks for remarks and for example of copying especially, I was struggling two days long with some memory errors22:52
blackburn22 hours!22:53
@sonney2kblackburn, ask if you get stuck - better than wasting time22:58
@sonney2kbut right I might not always be responsive :(22:58
blackburnsonney2k: I tried but you wasn't there sometimes ;)22:58
blackburnand at least I did other work I should do22:59
blackburnsonney2k: is PRC similar to ROC?23:01
@sonney2ksorry about that...23:01
blackburnsonney2k: you don't have to sorry you have your affairs23:01
@sonney2kblackburn, just implement the  python code for now... I don't really know how to properly deal with the case when multiple outputs are the same23:01
blackburnmay be the same?23:03
blackburnwe have to ask fawcett about this :D23:03
blackburnsonney2k: do we need DET?23:03
@sonney2kblackburn, I mean I know that one can translate any point in ROC space to PR space23:03
@sonney2kblackburn, wasn't DET not just 1-auROC or so?23:04
blackburnit is, but graph will be not same, may be one couldn't want to translate it this way23:04
@sonney2kblackburn, I don't care about it23:04
blackburnok, so I will not too23:05
blackburnomg23:05
blackburnprecision=numpy.cumsum(pmap)/(numpy.arange(len(pmap),dtype=numpy.double)+1.0)23:05
blackburnwhat a code! ;)23:05
blackburnsonney2k: is there any way to fix some comment without new pull?23:14
@sonney2kblackburn, just tell me which one and I hack it in23:15
@sonney2khmmhh seems like what I've just pushed doesn't build23:15
blackburnsonney2k: ROCEvaluation.h line 7023:16
blackburnplease fix it to23:16
blackburn/** number of ROC graph points */23:16
blackburnsonney2k: hm. what error?23:16
@sonney2kblackburn, some ifdef / etc not in your code23:16
@sonney2kbut cwidmer's...23:17
blackburnhehe23:17
@sonney2kyour potential mentor =:-)23:17
blackburnaha23:18
blackburnassume he forgot to test it ;)23:18
@sonney2kseems to be a common problem23:19
@sonney2kwould be your duty to teach him then ;-)23:20
blackburn:D23:20
blackburnsonney2k: btw, are applications working on your shogun?23:22
@sonney2kyou mean there are some real applications? like the ocr thingy on the homepage?23:22
blackburnaha, I will write an error just after compile it23:23
@sonney2kI guess most of them are now broken due to us changing too much23:23
blackburnsome issue with stringfeatures, iirc23:23
@sonney2kcould very well be23:23
blackburnshogun is compiling very slow on my notebook!23:24
@sonney2kI expect many more changes anyways so I won't bother about these yet23:24
@sonney2kblackburn, use ccache and ./configure  --enable-debug --disable-optimization23:24
@sonney2kand only the interface you need23:24
@sonney2kotherwise it is hell(tm)23:24
blackburni don't know what is ccache :)23:25
blackburndebug is enabled23:25
blackburnsonney2k: you said about 'steppiness' of ROC I showed you yesterday or so23:26
@sonney2kblackburn, install and compile shogun twice and you will realize what ccache does :)23:26
@sonney2kyes23:26
blackburnseems it will be 'steppy' almost in every case23:27
blackburnbecause it will have diag line only if there are equal labels23:27
blackburnsonney2k: that picture I showed you today with ROC is for some noised random labels23:27
blackburnit is 'steppy' if zoomed in23:28
@sonney2kblackburn, ahh right - the convex hull thingy is just sth that can be done on top - the steppy one is the one that belongs to the classifier23:28
blackburnKonrad said me about some averaging of ROC graph23:29
blackburnif it was he :D23:29
@sonney2kblackburn, mlsec ... yes23:29
blackburnif averaging is necessary I could work on it too23:30
@sonney2kthat issue could come up when we do cross validation23:30
@sonney2knot sure if it is necessary right now though23:30
blackburnok. it's good to me, because I want to get rid of evaluation and start working on CGraph class23:31
@sonney2kblackburn, yes do what you prefer to work on!23:31
blackburnbut at first will implement PRC :)23:31
@sonney2kthanks!23:31
blackburnI have to finish or it will be unfinished23:32
blackburnall things from perfmeasures are now implemented more 'modular', but not PRC23:32
blackburnlast one :)23:33
blackburnit is compiling still..23:33
blackburnhurray23:38
@sonney2kbtw, mlsec mentioned yesterday (chat logs rule!) that roc shouldn't be computed on discrete outputs - true in general but not when you want to test that everything is ok :)23:41
@bettyboo*g*23:41
blackburnsonney2k: but what's up with ifdef in Chris' commits?23:42
@sonney2kblackburn, I have fixed this but not yet pushed...23:42
@sonney2kdid you pull again?23:42
blackburnsonney2k: I mean he don't fixed anything23:42
blackburnwith #ifdef23:43
blackburnsonney2k: what you mean?23:43
@sonney2kI don't know what you are saying23:43
* sonney2k is confused?23:43
blackburn:D23:43
blackburnsorry23:43
blackburnI mean Chris don't fixed any preprocessor things23:43
blackburnwhy it is failing now?23:43
blackburndidn't*23:44
@sonney2kno idea....23:44
blackburnwill fetch it and test23:45
blackburn21 hours left23:45
blackburn:D23:45
blackburnШ рфму штыефддув23:48
blackburndamn23:48
blackburnI have installed ccache23:48
blackburnbut it is slow though23:48
blackburnsonney2k: ah.. unterminated #ifdef23:56
@sonney2kblackburn, did you enable it? yeah23:56
blackburnenable what?23:56
@sonney2kccache?23:57
blackburnnope, how I should? :D23:57
@sonney2kread the docs :)23:57
@sonney2kof ccache23:57
@sonney2k/usr/share/doc/ccache/README.Debian in my case23:57
blackburnthank you23:58
--- Log closed Mon Apr 25 00:00:37 2011

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!