Open in new window / Try shogun cloud
--- Log opened Mon Apr 16 00:00:19 2012
-!- karlnapf [~heiko@host86-180-222-171.range86-180.btcentralplus.com] has joined #shogun00:04
-!- harshit_ [~harshit@182.68.246.67] has quit [Read error: Connection reset by peer]00:05
-!- av3ngr [av3ngr@nat/redhat/x-ujqsetgbcmptahsn] has joined #shogun01:15
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 245 seconds]02:56
-!- karlnapf [~heiko@host86-180-222-171.range86-180.btcentralplus.com] has left #shogun []03:08
-!- vikram360 [~vikram360@117.192.175.234] has joined #shogun04:20
-!- vikram360 [~vikram360@117.192.175.234] has quit [Ping timeout: 245 seconds]07:15
-!- vikram360 [~vikram360@117.192.175.234] has joined #shogun07:25
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun07:57
n4nd0good morning07:58
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]08:11
CIA-64shogun: Soeren Sonnenburg master * r1be6198 / (2 files):08:22
CIA-64shogun: Merge pull request #446 from PhilippeTillet/nearest_centroid08:22
CIA-64shogun: Added doxygen comments - http://git.io/yUFwRA08:22
CIA-64shogun: Philippe Tillet master * r3255541 / (2 files): Added doxygen comments - http://git.io/L0IMDg08:22
CIA-64shogun: Philippe Tillet master * r0dc55d2 / (2 files): Added assertion on feature type, removed store_model_features - http://git.io/z-feJA08:22
-!- gsomix [~gsomix@188.168.128.177] has quit [Ping timeout: 252 seconds]08:24
-!- sonne|work [~sonnenbu@194.78.35.195] has joined #shogun08:48
-!- uricamic [9320543b@gateway/web/freenode/ip.147.32.84.59] has joined #shogun09:19
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun09:42
shogun-buildbotbuild #480 of python_modular is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/48009:51
n4nd0shogun-buildbot: good job09:57
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun10:00
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host]10:00
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun10:00
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun10:00
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection]10:08
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun10:08
n4nd0blackburn: I think you were faster answering :P10:32
blackburnfaster than already? :)10:33
blackburnah10:33
blackburnI see10:33
blackburnlol10:33
n4nd0haha10:33
blackburnmy answer is much more brief10:33
n4nd0I am a literat10:34
n4nd0:P10:35
shogun-buildbotbuild #481 of python_modular is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/481  blamelist: sonne@debian.org, Phil.Tillet@gmail.com10:43
blackburnehm10:46
blackburnwow outofmemoryerror on server with weblogic11:07
n4nd0nice one11:08
-!- av3ngr [av3ngr@nat/redhat/x-ujqsetgbcmptahsn] has quit [Quit: That's all folks!]11:10
blackburnJVM version Java HotSpot(TM) Client VM 1.4.2_05-b04  OS Linux, ver: 2.6.18-194.el511:10
blackburnhuh pretty old one11:10
shogun-buildbotbuild #476 of ruby_modular is complete: Failure [failed test_1]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/476  blamelist: sonne@debian.org, Phil.Tillet@gmail.com11:16
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 276 seconds]11:37
-!- wiking_ [~wiking@78-23-189-112.access.telenet.be] has joined #shogun11:52
-!- wiking_ [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host]11:53
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun11:53
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 240 seconds]11:56
-!- wiking_ is now known as wiking11:56
blackburn24gb and 12 cores o_o11:58
blackburnI wish I could use it to train some svms :D11:58
-!- vikram360 [~vikram360@117.192.175.234] has quit [Ping timeout: 244 seconds]12:00
blackburnuricamic: are you familiar with libocas?12:20
uricamicblackburn: yes12:23
blackburnuricamic: can it solve the following slightly modified problem:12:24
blackburn||w||^2 + w0 w + ..?12:25
blackburnw0 is obtained before so can be considerd as constant vector12:25
blackburnI am lame with it :D12:27
-!- PhilTillet [~Philippe@157.159.42.154] has joined #shogun12:40
uricamicwell I will look at it, but I am afraid that it cannot be used for that12:40
uricamicblackburn: the .. in your question means the rest of the original equation in OCAS solver?12:44
blackburnuricamic: yes sure12:44
blackburnuricamic: actually I'm interested in something similar to crammer-singer12:45
blackburnso even \sum_m ||w_m||^2 + \sum w_m w0_m + C \sum_i \xi_i12:45
blackburns.t. <w_yi,x_i> - <w_m,x_i> <= 1 - \delta_{yi,m} - \xi_i12:46
uricamicblackburn: ok, in that case I think you cannot use libocas without modification, the additional w term will change the dual function12:46
blackburnI hope it is parseaable12:46
blackburnuricamic: hmm is libocas a dual solver?12:46
uricamicI think it is12:48
-!- vikram360 [~vikram360@117.192.186.232] has joined #shogun12:48
blackburnuricamic: it looks for me like it is a primal solver12:52
-!- vikram360 [~vikram360@117.192.186.232] has quit [Ping timeout: 260 seconds]12:53
-!- vikram360 [~vikram360@117.192.160.167] has joined #shogun12:54
uricamicblackburn: sorry my mistake, it is12:54
blackburnwhy I am asking is I am curious how hard it would be to modify it12:55
uricamicblackburn: I am looking on the libocas.cpp code now, and it looks that e.g. in svm_ocas_solver it uses dual function13:08
uricamicit calls libqp_splx_solver and computes the wights of alpha13:08
blackburnyou are right13:09
uricamic*weights13:09
blackburnthat's all confusing13:09
uricamicand the transformation to the primal objective function is done in compute_W, which uses this alpha13:11
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection]13:36
-!- wiking [~wiking@vpnd095.ugent.be] has joined #shogun13:37
-!- wiking [~wiking@vpnd095.ugent.be] has quit [Changing host]13:37
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun13:37
-!- harshit_ [~harshit@182.68.246.67] has joined #shogun14:39
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun14:42
-!- harshit_ [~harshit@182.68.246.67] has quit [Ping timeout: 245 seconds]14:51
blackburntotally DEADLY today14:58
blackburnn4nd0: what's up?14:58
n4nd0blackburn: bah, not much, boring day so far14:59
n4nd0blackburn: what about you?14:59
blackburnn4nd0: totally BORING14:59
blackburnI guess it is the time to rename shogun14:59
n4nd0really? to what?15:00
blackburnn4nd0: I preferred CCCP15:00
n4nd0that stands for?15:00
blackburnn4nd0: USSR15:00
blackburnback in back in15:01
blackburnback in USSR15:01
blackburn;)15:01
n4nd0back to black maybe :P15:01
blackburnn4nd0: it seems you are not the biggest fan of beatles :)15:02
n4nd0blackburn: probably not, but I like them quite much :)15:03
n4nd0blackburn: why so?15:03
blackburnn4nd0: they had this song :)15:04
* n4nd0 is listening Back in the USSR15:05
n4nd0I didn't know about it :)15:05
-!- vikram360 [~vikram360@117.192.160.167] has quit [Ping timeout: 276 seconds]15:08
-!- PhilTillet [~Philippe@157.159.42.154] has quit [Ping timeout: 248 seconds]15:24
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has left #shogun []15:28
-!- emrecelikten [~Anubis@213.74.228.187] has joined #shogun15:47
emreceliktenHello all15:47
blackburnuricamic: is there any speedup with GS?15:52
uricamicblackburn: it depends on the size of kernel matrix, I think15:53
blackburnuricamic: sure with large kernel matrix pinv wouldn't work at all :)15:54
uricamicI have also tried variant which will not compute the whole kernel matrix, but it was very slow15:54
blackburnreally?15:54
uricamicwell, maybe I haven't done that in the most clever way15:55
blackburnah you need each element there15:55
uricamicI took the row of matrix in each step, but I think that this allocation and freeing of memory is killing the performance15:56
blackburnuricamic: how many iterations it requires?15:56
wikinguricamic: hey i was wondering if you'll have some times this week to have a chat about the integration of libqp into shogun15:56
uricamicdepends on the epsilon and also on the initialization15:57
uricamicwiking: libqp?15:57
uricamicwiking: I thought it is already there15:57
wikingwell it's a very immature integration15:57
uricamicblackburn: I will try to think about this GS for KRR, maybe it can be simplified which could bring some performance increase15:58
uricamicwiking: ok15:58
wikingand afaik you'll use it in your bmrm , or?15:58
uricamicwiking: yep, I am using it there15:58
wikingyeah and i'm planning to use it in my latent svm implementation15:59
wikingthus maybe if you have some time we could have a little chat about the libqp integration into shogun15:59
uricamicwiking: sure, I will be on IRC all this week, I think16:00
wikingcool16:00
wikingmaybe then tomorrow16:00
wikingtoday is not the best for me16:00
uricamicsure, neither for me, I am leaving soon16:00
uricamicso tommorow in the morning or afternoon?16:00
blackburnthat day is totally boriinggggg16:01
wikinguricamic: morning means for you?16:01
wikingi mean for me tomorrow before noon around 10am would be fine16:01
uricamicafter 9am16:01
uricamicok16:02
wikingok cool, so how about 10am?16:02
wikingcool16:02
wikingi'll be around16:02
wikingthnx a lot!16:02
uricamicI hope we have the same timezone :D16:02
wikingyep16:02
wikingyou are from .cz right?16:02
uricamicyep16:02
uricamicok, fine :)16:02
wikingdiky16:02
uricamicneni zac :D16:02
wikingyeah about this is my level of czech ... i think i could order some beers as well ;P16:03
wikingdva piva? ;016:03
blackburnwiking: https://github.com/shogun-toolbox/shogun/pull/450 what would you say about my last comment?16:03
uricamic*dve16:03
blackburnsound like russian16:03
wikingah ok16:03
wikingit's different then a bit16:03
blackburndva piva is totally correct russian16:03
wikinghahah yeah i think serbian is closer to russian16:04
wiking;)16:04
uricamicall of these are quite similar16:04
blackburnhowever dve butylki vodki would make it more russian16:04
wikingblackburn: i'll check the comment just a sec16:04
wikinguricamic: yeps16:04
blackburnuricamic: are you able to understand russian then? ;)16:04
wikinguricamic: this is how i survive when i'm in .cz ;)16:04
uricamicwhenever I hear some russian counting I feel like hearing czech :)16:04
uricamicblackburn: only a little16:04
blackburnmy friend studying in cz teached me how to count in czech16:05
blackburnhowever I do not remember that :(16:05
uricamicfor example some words are completely the same, but some have exactly the opposite meaning :)16:05
blackburnuricamic: how does it sound like?16:05
wikingfor sure they understand: jedan, dva, tri, cetri, pet....16:05
wiking;)16:05
wikingmaybe there's an -i on the end ;P16:05
uricamicjeden, dva, tri, ctyri, pet16:05
wikingsee it's very similar :)16:06
blackburnodin dva try chetyre pyat schest sem' vosem' devyat'16:06
wikingblackburn: your problem is with how to find out the size of an array?16:06
blackburnwiking: it is not a problem of mine ;)16:06
uricamiccounting for sure, but for example word "voni" has the exactly opposite meaning in russian than in czech :)16:06
blackburnwiking: do you agree it *can't* work at all?16:06
wikingi mean this should be fine: sizeof(vec)/sizeof(variable that the array is made of)16:06
blackburnuricamic: voni? some smells?16:06
wikingthat's totally legal in ansi c16:07
uricamicin czech it means that something smells nice, and in russian it means that somethins smells badly16:07
blackburnuricamic: yeap16:07
wikingof course you'll start having troubles16:07
blackburnwiking: I am afraid it would not work still16:07
wikingit will16:07
wikingexcept if it's a pointer16:07
wiking;)16:07
blackburnit is16:08
blackburnhttp://pastebin.com/VYvPnkiP16:08
blackburnit would work only for static allocated arrays afaik16:08
wikingyeah as said pointers will make you feel bad ;)16:08
wikingsince you'll get back the pointer size16:08
blackburnso zero length alway16:08
blackburns16:08
wikingno pointer size = 4 or 816:08
wikingdepends on the arch16:08
blackburnah nevertheless it is not correct16:09
wikingyes with pointer you'll have troubles16:09
wikingand this syntax is anyways bogus16:09
wiking(sizeof array)/sizeof(int);16:09
wikingit should be like sizeof(array)/sizeof(int)16:09
wikingif16:10
wiking;)16:10
wikingblackburn: you can check your fav help as well, i.e. stack overflow; quote: "It is worth noting that sizeof doesn't help if you are dealing with a pointer to an array:"16:10
-!- harshit_ [~harshit@182.68.246.67] has joined #shogun16:11
wikingand of course the other problem here is what if you change your function arg in the future16:11
blackburnwiking: hah why stack overflow is my fav help?16:11
wikingblackburn: just joking ;)16:11
wikinganyhow if in future you make a function arg type change you are (not you but any 3rd person) most likely going to forget to change the type in  /sizeof(.)16:12
wikingso you'll end up with a bad vector size obviously16:12
wikingso it's far from being safe16:12
blackburnwiking: I was really surprised to see that here16:12
wikingmuch better to pass the vector size via the function arg, as anywhere else in shogun ;P16:12
blackburnI mean this way of obtaining a size of array16:13
wikingblackburn: yeah this is a dirty hack from ansi c ;)16:13
harshit_blackburn: how can I make my current fork exactly same as upstream, I don't want to rebase I want to get to the current state of upstream master16:13
wikingblackburn: it works until you work on the stack... as soon as you start using the heap you'll have problems16:13
blackburnwiking: in my example it is on stack but doesn't work16:13
blackburnharshit_: why don't you want to rebase?16:14
-!- ishaanmlhtr [~chatzilla@101.63.121.9] has joined #shogun16:14
harshit_I have made some changes in my fork that I want to delete ..16:14
wikingblackburn: true that :)16:14
blackburnharshit_: did you commit it then?16:14
harshit_yeah :(16:15
harshit_so do I need to delete my current fork16:15
harshit_and fork again ?16:15
blackburnno you can reset your state and push with --force16:15
harshit_reset ?? how to do that ?16:16
blackburnharshit_: something like git reset --hard upstream/HEAD16:16
blackburnwiking: am I right?16:16
wikingblackburn: yeah16:16
wikingharshit_: have you commit?16:16
blackburnhe did16:17
wikingi mean your code16:17
wikinginto your own branch16:17
harshit_wiking: yeah I have commit some changes16:17
wikingaaaah16:17
wikingwell then it's a bit tricky16:17
blackburnwiking: I can't see any other way but hard reset and force push16:17
wikingsince you'll repo is going to be a bit fucked up with you do a rebase16:17
wikingwh16:17
wikingblackburn: you can remove commites16:17
wiking*commits16:17
blackburnwiking: how?16:18
wikingjust you need to do a rebase16:18
wikingwhich is usually a pain in the ass16:18
wikingif somebody is following your repo16:18
harshit_why not delete and fork again ?16:18
blackburnharshit_: yes it is the simplest way16:18
wikingit's simple :)16:18
harshit_:)16:18
wikingblackburn: yours would do that16:19
wikinggit reset --hard upstream/HEAD16:19
wikingthat'd remove the commits16:19
blackburnnot really remove but jump to another state16:19
wikingand what happens with the prev commits? :)16:19
wikingi mean that are after HEAD?16:20
blackburnwiking: hmm are they really deleted from index?16:21
wikingyep16:22
wikingyou'll never get those commits back16:22
wikingunless you've saved them somewhere... other branch or somethin16:22
blackburnbut can't you access it by sha1?16:22
wikingnot in that branch no16:23
blackburnwiking: it seems we spend days solving puzzles16:23
wiking:D16:24
-!- vikram360 [~vikram360@117.192.160.167] has joined #shogun16:27
-!- ishaanmlhtr [~chatzilla@101.63.121.9] has quit [Quit: ChatZilla 0.9.88.2 [Firefox 9.0.1/20111220165912]]16:33
-!- emrecelikten [~Anubis@213.74.228.187] has quit [Ping timeout: 245 seconds]16:49
@sonney2kuricamic, around?17:03
@sonney2kuricamic, how many iterations does GS need to get eps precision?17:04
-!- harshit_ [~harshit@182.68.246.67] has quit [Ping timeout: 264 seconds]17:07
blackburnsonney2k: I am afraid he is afk17:08
blackburn(as he told he is leaving)17:09
uricamicsonney2k: I am still here17:09
blackburn:D17:09
blackburnthat's nice then17:09
uricamicI am now working on your comments17:10
uricamicbut I will have to leave soon17:10
@sonney2kblackburn, yeah I guess so too... I've studied his code and I guess it needs some kernel caching routines to be fast w/ GS17:10
blackburnsonney2k: here already17:10
@sonney2kblackburn, unfortunately we don't have that in CKernel ... we only have crappy svmlight kernel cache business there17:10
blackburnsonney2k: I mean uricamic is here17:11
blackburn:)17:11
@sonney2kso we in fact need one general solution that one can use with distances etc too17:11
blackburnbut yes I agree we need nice caching17:11
@sonney2kuricamic, hey17:11
uricamicsonney2k: I am working on the comments17:12
uricamicsonney2k: then I will do the benchmarks17:12
blackburnsonney2k: someone could do that while gsocing may be ;)17:12
@sonney2kuricamic, so any idea how many GS iterations you need?17:15
uricamicsonney2k: I haven't measure it in this implementation, but when I was trying my first implementation in MATLAB it was around 700 for eps=1e-517:16
@sonney2kuricamic, that is all I wanted to know - ok then you will need some kind of cache and clever strategy / way to iterate over examples for this to be any fast then17:19
uricamicsonney2k: yes17:20
uricamicI really have to go now, see you17:23
-!- uricamic [9320543b@gateway/web/freenode/ip.147.32.84.59] has quit [Quit: Page closed]17:23
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed]17:32
-!- emrecelikten [~Anubis@176.33.63.94] has joined #shogun17:32
-!- blackburn [~qdrgsm@188.168.2.179] has joined #shogun18:23
-!- akshayms [75c0dcbc@gateway/web/freenode/ip.117.192.220.188] has joined #shogun18:41
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun18:52
@sonney2kn4nd0, hey! good evening!18:52
n4nd0sonney2k: hey there!18:53
n4nd0good evening guys18:53
* sonney2k pastes in all the timezones into the timeanddate meeting planner18:53
n4nd0meeting planner, sounds cool18:54
-!- harshit_ [~harshit@182.68.246.67] has joined #shogun18:55
@sonney2kn4nd0, yeah dammed we have 6 timezones or so19:04
n4nd0oh19:04
n4nd0I can adapt my timezone if necessary :P19:05
@sonney2kwe are basically dead already because we have one mentor in NY19:07
@sonney2kand one in melbourne19:07
@sonney2kand some spread inbetween19:07
@sonney2kso it is like either the NY guys suffer or the ozzies19:08
n4nd0sonney2k: who is the mentor in NY?19:09
@sonney2kchristian widmer19:10
n4nd0ok19:10
@sonney2kUTC 11 hrs or 12hrs seems to be the only viable time19:11
@sonney2kwell in one week the officially accepted students will be announced19:16
@sonney2kso we should have some kickoff meeting that informs about how things work and they way we want them to work :)19:17
n4nd0:)19:17
n4nd0sonney2k: by the way, one doubt about JL's cover tree19:17
n4nd0sonney2k: so when m_q != 1, since the cover tree doesn't output the neighbors ordered by distance19:18
n4nd0it would be required to compute the distance to the k neighbors and sort them19:19
n4nd0something similar to what is done in the method to classify for different k19:20
n4nd0when you said on github19:21
n4nd0"I would just speed up the most common q=1 case with the covertree."19:21
n4nd0do you mean that this computation of the distances and sorting shouldn't be done right?19:21
n4nd0i.e. no cover tree + m_q != 119:21
@sonney2kn4nd0, I am not sure what you are saying19:23
@sonney2kI meant m_q !=1 -> no cover tree19:23
n4nd0haha ok19:23
@sonney2kslow19:23
n4nd0ok19:23
@sonney2kbut if you want to do the sorting - feel free :)19:23
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun19:24
n4nd0it is ok like you say19:24
@sonney2kit would be the same thing because you dont' have to sort :)19:24
n4nd0you just condensed all the stuff I said in a couple of sentences :D19:24
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 248 seconds]19:27
-!- wiking_ [~wiking@huwico/staff/wiking] has quit [Ping timeout: 240 seconds]19:28
-!- harshit_ [~harshit@182.68.246.67] has quit [Read error: Connection timed out]19:50
-!- harshit_ [~harshit@182.68.246.67] has joined #shogun19:51
blackburnsonney2k: being chris in NY doesn't lead to any inconveniences I think ;)19:51
blackburnhe tends to appear randomly anyway19:53
blackburnsonney2k: n4nd0: I can't agree m_q != 1 should be w/o covertree..19:54
-!- akshayms [75c0dcbc@gateway/web/freenode/ip.117.192.220.188] has quit [Quit: Page closed]19:55
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun19:57
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host]19:58
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun19:58
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit]20:02
@sonney2kblackburn, it will still be a problem to have a meeting with everyone20:02
blackburnsonney2k: did you know we want to send you a present - book on STL? ;)20:03
-!- gsomix [~gsomix@83.234.54.52] has joined #shogun20:03
gsomixhi20:03
blackburnprishel20:03
blackburnsonney2k: encourage gsomix with my favourite russian sentence of you20:04
@sonney2kprishel?20:04
blackburnsonney2k: 'came'20:04
blackburnsonney2k: but my fav is about work20:05
gsomixblackburn, idi rabotay?20:05
@sonney2kblackburn, very good - I realize that you can easily take over the project when I leave :D20:06
blackburnsonney2k: because of?20:06
blackburn:)20:06
blackburnsonney2k: I am still waiting you to get gsomix to work!20:07
blackburn:(20:09
@sonney2kso gsomix work !20:12
blackburnsonney2k: "rabotatsch"!20:12
blackburneh what a sadness :)20:12
gsomixyou are funny, guys :)20:13
gsomixblackburn, add expression20:14
blackburngsomix: which one you would like?20:14
gsomix"rabotatsch, bldjad'!"20:14
blackburnoh I didn't think about obscene stuff yet20:15
gsomixhuh20:15
gsomixblackburn, but you are right - it's time to work.20:19
blackburnyou got my implicit message right ;)20:19
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun20:36
-!- nickon [~noneedtok@kotnet-147.kulnet.kuleuven.be] has joined #shogun20:41
blackburngsomix: what is you are working on?20:47
gsomixblackburn, SGVector, khm20:48
blackburnah right20:48
@sonney2kblackburn, you wanted him to start working on caches right :D20:48
blackburnsonney2k: no that's damn hard I think20:49
blackburnI just have holes in my head and my thoughts slip away through it20:49
gsomixblackburn, tomorrow I learn about researches that are conducted at our university.20:53
blackburnah how to drink 2 litres of vodka?20:54
blackburnwhat is the role of optics when you are drunk?20:54
blackburnhow to burn up cucumber with laser to have vodka more comfortably?20:54
@sonney2kblackburn, vision correcting glasses for drunks!20:54
blackburnsonney2k: I imagine its surface20:55
blackburnkind of perlin noise lol20:55
@sonney2kblackburn, tell him about the animal tests they are conducting!20:55
blackburnsonney2k: you mean 2nd semester of optics for me?20:55
-!- PhilTillet [~Philippe@157.159.42.154] has joined #shogun20:56
blackburnsounds like cruel animal test20:56
@sonney2kyeah animals == students!20:56
gsomixblackburn, is it a future topic of my bachelor diploma? T___T20:56
@sonney2kthat's the devastating truth20:56
blackburngsomix: yeah some vodka stuff20:56
blackburnsonney2k: do you know they reject any work which title has no 'vodka' word?20:57
@sonney2kI knew it!20:57
gsomixit's time to become a physicist. no optics! T____T20:57
@sonney2ksomeone named vodking told me!20:57
n4nd0:D20:58
blackburnsonney2k: my bachelor's work is title20:58
blackburnsupport vodka machine recognition20:58
blackburntitled*20:58
blackburnSVM20:58
blackburnI wish it was vector but they forced me to call it support vodka machine20:59
@sonney2kvodka really is the key to everything...21:00
@sonney2kwithout vodka R.I.P.21:00
@sonney2kno fun, no gsoc, no shogun, no theorems, no kernels, no support vodka machines!21:00
blackburnyou catch things21:01
@sonney2kyeah I am fast in this21:01
@sonney2kthat's because we had a brewing seminar for x-mas party at tomtom21:01
@sonney2kand you know once your blood got replaced by vodka there is nothing to worry about!21:02
blackburnsonney2k: heh21:02
gsomixblackburn, should I offer vodka to my future supervisor?21:03
blackburngsomix: to sobolev?21:03
gsomixyep21:03
blackburnhe seems to not be a big fan of it21:03
gsomix:(21:04
blackburngsomix: consider kolomiets for that21:05
n4nd0does that work in Russia, to buy things for the teachers?21:05
gsomixblackburn, ohoho.21:05
n4nd0some friends tried it once in my school (for real!) but it didn't work21:05
blackburnn4nd0: to get exam passed?21:05
n4nd0blackburn: yeah, for example21:05
blackburnyes pretty often unfortunately21:06
blackburnmostly money however21:06
n4nd0the story in my university is quite fun actually21:06
n4nd0they put a serrano on the hood of the car21:06
blackburnn4nd0: half of my group at univ bribed teacher to get physics passed21:07
blackburn6000 roubles = 200$ each21:07
n4nd0and went to the office with more stuff21:07
gsomixblackburn, congac (brandy) for Alimenkov21:07
blackburnha21:07
blackburnn4nd0: did it work?21:08
n4nd0blackburn: no21:08
blackburnn4nd0: as for your very first question21:08
blackburnit works to bribe anyone in russia21:09
blackburnit is just a matter of different cost ;)21:09
n4nd0haha I see21:10
n4nd0I actually shouldn't laugh, sad thing indeed21:10
blackburnn4nd0: police there tends to put a bottle in your ass if they don't like you :D21:11
blackburnI am not joking21:11
blackburnthree cases recently one has died21:11
gsomixblackburn, brrr >__< this story...21:11
n4nd0:O21:12
n4nd0for real? hard to believe21:12
n4nd0blackburn: I am not saying you're lying, that could be misunderstood :S21:12
blackburnn4nd0: yes in kazan'21:12
blackburnthey put a bottle of champagne right to the ass21:12
n4nd0not good21:13
gsomixblackburn, horror stories about Russia21:15
blackburnhah why not21:15
-!- PhilTillet [~Philippe@157.159.42.154] has quit [Remote host closed the connection]21:17
gsomixblackburn, I sleep badly, hehe21:18
blackburnn4nd0: remember you said to come up with some paper sounds like something good?21:31
n4nd0blackburn: yes21:31
emreceliktenI think one of you guys recommended a good dataset website before, what was it?21:31
n4nd0very good actually21:31
blackburnemrecelikten: mldata.org21:31
emreceliktenblackburn: Yep, that's it. Thanks21:31
blackburnn4nd0: I think it would be hard and time consuming to come up with strong journal article21:31
n4nd0blackburn: why?21:32
blackburnhowever there is arxiv still21:32
blackburnand we could place some kind of paper there :)21:32
n4nd0aham21:32
n4nd0I didn't know about arxiv21:32
blackburnhmm it is openaccess and rather easy reviewed21:33
blackburnso you won't be rejected because this paper does no revolution in SO :)21:33
n4nd0yeah sure, it could be a nice place to start with21:33
blackburnI would be happy to collaborate later on that21:33
n4nd0blackburn: :)21:34
n4nd0it looks that they not require the common format with two columns of text per page21:35
n4nd0blackburn: by the way, did you or any of you blog about gsoc last year?21:36
blackburnno21:36
blackburnI actually do not think it is so useful though21:36
n4nd0I think it is a good idea21:36
blackburnsonney2k: what do you think about adv ourselves and shogun using arXiv? :)21:39
blackburnand blogs21:39
@sonney2kwell we have a meta-blog21:40
blackburnyours?21:40
@sonney2khttp://shogun-toolbox.org/planet/21:40
@sonney2keveryone that wants to be on it - send me the url21:41
@sonney2kand voila21:41
blackburnah21:41
blackburnyes21:41
@sonney2kand please regularly write something21:41
@sonney2kI am damn lazy so I won't write more than once a year or so21:41
@sonney2kcrazy .. .https://fgiesen.wordpress.com/21:42
* sonney2k gets tired21:42
blackburnsonney2k: and what about arxiv stuff?21:42
@sonney2kblackburn, publish as much as you want :D21:45
blackburnsonney2k: hrr I am asking whether it is a good idea21:46
@sonney2kblackburn, depends what you want to achieve...21:50
@sonney2kif you want to become professor it is worth nothing21:51
blackburnsonney2k: me? professor?21:51
@sonney2kbut to e.g. apply to something related to ML (also research) it can help21:51
@sonney2kblackburn, you drink vodka so you qualify21:52
blackburndon't you think I am too stupid^W programmer to be a professor?21:52
@sonney2kand you know the 'dawai dawai rabotatch' slogan pretty well21:52
blackburnfinally you said that21:52
@sonney2kwellyou have to stop programming of course and let your slaves^H^H^Hstudents to this21:52
blackburnsonney2k: no seriously no chance21:54
n4nd0sonney2k: but now, while we are studying, isn't it a good idea to do some programming job to become professor in the future?21:54
blackburnbrb21:54
n4nd0I meant job, like work or collaboration as we do here21:54
-!- karlnapf [~heiko@host86-177-177-13.range86-177.btcentralplus.com] has joined #shogun21:58
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection]21:58
-!- wiking [~wiking@vpnb164.ugent.be] has joined #shogun21:58
-!- wiking [~wiking@vpnb164.ugent.be] has quit [Changing host]21:58
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun21:58
@sonney2kn4nd0, all profs did real work at some point - they started out as researchers, group leaders, ...22:07
@sonney2kso yes it helps + you need to slowly become some kind of leader (like backburn :D22:07
@sonney2khey karlnapf22:07
karlnapfsonney2k, hi there22:08
n4nd0sonney2k: thank you :)22:08
karlnapfsonney2k, good to catch you, wanted to talk about subsets once more22:08
karlnapfI am currently working on moving all the stack stuff to a separate class22:09
-!- harshit_ [~harshit@182.68.246.67] has quit [Remote host closed the connection]22:09
karlnapfwhich does everything, instead of doing this by hand in every class22:09
karlnapfsonney2k, I think this should solve all subset stuff once and for all22:10
karlnapfI think it will also be possible to migrate the old plain Subsets into that new system22:11
-!- PhilTillet [9d9f0a0e@gateway/web/freenode/ip.157.159.10.14] has joined #shogun22:20
PhilTilletHello22:20
karlnapfhi22:21
@sonney2kkarlnapf, sounds great22:23
@sonney2k(as usual)22:23
karlnapfit will also be easier than its now22:23
@sonney2kI guess you should just do it then and I can only hope that you are only 150% overloaded :)22:23
karlnapftell me about it :)22:24
karlnapfWell I cannot compete with you?22:24
karlnapfehm not ? ,  !22:24
karlnapfwhat about you and your family? are things getting more relaxed?22:25
@sonney2kI can sleep at night now!22:28
@sonney2k(since about the week before easter)22:28
@sonney2kand that is a big big big +22:28
@sonney2kwhich reminds me22:28
@sonney2kI should really sleep now22:28
@sonney2kkarlnapf, nice talking to you if only briefly22:28
@sonney2kcu22:28
karlnapfsonney2k, ok enjoy :)22:29
blackburnkarlnapf: hey there22:30
karlnapfblackburn, hi, how is russia? ;)22:31
-!- nickon [~noneedtok@kotnet-147.kulnet.kuleuven.be] has quit [Quit: ( www.nnscript.com :: NoNameScript 4.22 :: www.esnation.com )]22:32
blackburnkarlnapf: stays the same I guess22:32
blackburnkarlnapf: btw I had some crazy thought22:32
karlnapfblackburn, tell me22:33
blackburnwhat if I write some things to you in german and you will check/correct it? :D22:33
karlnapfyou wanna learn German?22:33
blackburnyes22:33
blackburnI know you are busy so 'no' would be ok too ;)22:34
karlnapfnice, I like that.22:34
karlnapfWell what about in June?22:34
blackburnyes sure22:35
karlnapfafter my exams?22:35
blackburndeal22:35
karlnapfThen I am happy to help you22:35
karlnapfI encourage everybody in learning German :)22:35
karlnapfwhy you want to do that?22:35
blackburnkarlnapf: I could help you with russian if you want22:35
blackburnkarlnapf: I want to visit deutschland and probably stay there22:35
karlnapfWould be an idea, but thats probably not before winter, to much stuff to do before22:36
karlnapfreally?22:36
karlnapfany concrete plans?22:36
blackburnand damn you are all germans22:36
blackburnno but may be chris would call me to be his phd, who knows? ;)22:36
karlnapfoh cool :)22:37
karlnapfmaybe we will meet then22:37
karlnapfI am applying to the MPI22:37
blackburnheh who knows again22:38
blackburnit would be useful anyway22:38
blackburnkarlnapf: to talk is the only way I guess22:38
karlnapfyes indeed.22:39
blackburnless or more you now understand my english and should work for german22:39
karlnapfyeah you will get that22:39
karlnapfand you had it in school so should be easy22:40
blackburnit was pretty bad in school22:40
karlnapfmmh, but now you have me correcting you :(22:40
karlnapf:822:40
blackburnheh22:40
blackburnI have known more from rammstein songs than from school22:41
karlnapfhehe22:41
karlnapframmstein22:41
karlnapfcrazy guys22:41
karlnapfonce saw them live22:41
karlnapffire all over the place22:41
blackburnit is easy to distinguish their german words for me22:41
karlnapfbut nobody talks like them here ;)22:42
blackburnyes I hope so22:42
blackburnand actually their songs are about some crazy shit22:42
blackburnI was pretty surprised translated zwitter22:43
blackburnkarlnapf: eins, zwei, drei, fier, f?nf, sechs, sieben, acht, neun, zehn, zw?lf!22:44
blackburnI know that22:44
karlnapfvier22:44
blackburnohh22:44
karlnapfand zehn elf zw?lf :)22:44
blackburnhahahah22:44
blackburn95%22:44
blackburnkarlnapf: v as f is a bit confusing22:45
blackburnv?gel22:45
karlnapfyeah many people confuse these22:45
karlnapfbtw it owuldnt be too bad if you would start learning it here22:45
karlnapfI think many people in t?bingen talk english all day22:46
karlnapfArthur Gretton was there a few years and cannot speak German22:46
blackburnkarlnapf: ah yes but I am rather curious than planning something22:46
blackburnit is always useful to speak diff langs22:46
karlnapfyea true22:47
karlnapfvery nice for traveling22:47
karlnapfI want to get into more languages some time22:47
n4nd0good night guys22:48
blackburnn4nd0: see you!22:49
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]22:49
blackburnkarlnapf: are you applying for Phd at MPI?22:49
karlnapfn4nd0, good night22:49
karlnapfblackburn, yes I think I will do that in a few month22:49
karlnapf(among other places)22:50
blackburnkarlnapf: but what about gretton?22:50
karlnapfalso an idea22:50
blackburnI thought you are going to work with him22:50
karlnapfbut I can only apply in Dec 201222:50
karlnapffor Sept 201322:50
karlnapfpretty late22:50
karlnapfbut perhaps there is a way of bridging this gap22:50
karlnapfdont really know yet22:51
karlnapfI wrote to some people22:52
karlnapfbut I want to concentrate on my work currently, thats more important22:52
blackburnkarlnapf: I see22:53
gsomixgood night, guys22:54
blackburnanother one bites the dust22:54
karlnapfgsomix, gn822:54
karlnapfhehe ;)22:54
blackburngsomix: spokoynoy drochi!22:54
gsomixblackburn, :322:54
blackburnuh that's kind of hard to understand joke22:55
blackburnkarlnapf: so you want to do bioinf?22:57
blackburneverybody does this thing :D22:57
karlnapfblackburn, mmmh I was a bit more sure about this a year ago22:58
blackburnbut actually only in shogun neighborhood probably22:58
karlnapfcurrently I am more interested in kernel stuff and bayesian stuff22:58
karlnapfbut who knows22:58
karlnapfthis MMD thing might be a nice direction22:58
blackburnwhoa I always hated that bayesian stuff22:58
blackburnnetworks, etc22:58
karlnapfoptimization, kernels, statistics22:58
karlnapfI kind of like the gaussian processes22:58
blackburnprobably this is because I do not understand it well22:59
karlnapfwhich is just bayesian linear regression in the feeature space22:59
blackburnkarlnapf: yeah it seems a lot of people like GPs22:59
blackburnchris does afaik22:59
karlnapfno surprise, they are quite recent23:00
karlnapfand really cool23:00
karlnapfnot cool for large scale stuff23:00
karlnapfat least out of the box23:00
karlnapfI havent tried them with some kernel matrix approximation23:00
blackburnkarlnapf: is it only for regression? are classification results so well too?23:01
karlnapfyes23:01
karlnapfcheck out this book by carl rasmussen23:01
karlnapfawesome23:01
karlnapfcurrently studying this stuff for exams23:01
blackburnI've glanced over it before23:02
karlnapfI fell there is so much stuff23:04
karlnapfhave no idea how to decide on anything23:04
karlnapfbut probably thats handled by the people one konws23:04
karlnapfknows23:04
blackburnkarlnapf: I just realized that I will suggest to study GPs to my gf23:06
karlnapfwhat???23:07
blackburnshe was excited about random processes studies23:07
karlnapflol, she likes that?23:07
blackburnyes but not very familiar with ML still23:07
blackburnher bachelor's work is about multiclass svms23:07
karlnapfhehe, funny, you are a ML couple23:08
blackburnbut she doesn't like it so much23:08
blackburnI have to mentor this while professors here are lame idiots23:08
karlnapflol23:09
wikingyeeey i have a competitor !23:10
karlnapfyou cannot escape the ML then23:10
blackburnkarlnapf: why?23:10
blackburnwiking: which?23:10
karlnapfblackburn, just joking23:10
wikingblackburn: saw the mails on shogun mailing list for latent svm23:10
karlnapfwiking, whats happening=?23:10
blackburnkarlnapf: but I didn't get the joke :D23:11
karlnapfwell you work all day on ML stuff and then you return home and continue working on ML stuff, and then youre talking about ML stuff in bed :D23:11
blackburnhahah23:12
blackburnkarlnapf: I wish I was working on ML23:12
karlnapfwhat are you currently doing?23:12
blackburnkarlnapf: working at netcracker and studying some crazy shit23:13
blackburnbut I am going to finish with both activities23:13
blackburni.e. leave job for summer and pass exams :D23:13
karlnapfnice23:14
blackburnkarlnapf: I am pretty tired with development in such huge company23:14
karlnapfwhy?23:15
blackburnfor example oneliner fix took a week recently23:15
blackburnjust because of being "enterprise"/size/etc23:15
karlnapfoh that sucks23:15
blackburnkarlnapf: another example is just what I do23:17
karlnapfI am currently a bit annoyed by studying for exams .... so much stuff23:17
blackburne.g. today I fixed backpropagation of component associations related to component instances with vertical links while decomposing two service orders23:18
karlnapfich verstehe nur Bahnhof23:18
blackburnyou understand ...23:19
karlnapf"I understand just train-station" means: I dont understand anything ;)23:19
blackburnhah23:19
blackburnkarlnapf: that is the problem23:19
blackburnI used to learn stuff that makes no sense in outer world23:19
karlnapfhow do you mean that?23:20
blackburnI've learnt nothing but some netcracker-specific stuff23:20
karlnapfoh23:20
karlnapfmmh23:20
karlnapfbut you always train your mind when you do this23:21
blackburnI wouldn't say so23:21
karlnapfjipiie my shogun compiles again23:21
blackburnyou are right it is not harmful23:21
blackburnbut not very useful23:21
karlnapfwell its soon over :)23:21
karlnapfblackburn, I have to leave now23:22
blackburnyeah I prefer to full-time in gsoc23:22
blackburnok23:22
karlnapfgf wants to have a tea :)23:22
karlnapftake care and good night!23:22
blackburnhah yeah23:22
blackburngood night23:22
karlnapf :) bye23:22
-!- karlnapf [~heiko@host86-177-177-13.range86-177.btcentralplus.com] has left #shogun []23:22
wikinganybody knows in latex what to do when your limits for sum are not really under and above the sum sign?23:38
wikingah ok it's \limits23:40
blackburnwiking: not really under??23:41
wiking$\sum_{a=1}^n$23:41
wikingit will be not the same as $\sum\limits_{a=1}^n$23:41
blackburnwhile it is inlined it would be compact23:41
blackburnthere is a way to force \limits there23:41
blackburnsome option IIRC23:42
-!- PhilTillet [9d9f0a0e@gateway/web/freenode/ip.157.159.10.14] has quit [Ping timeout: 245 seconds]23:55
emreceliktenHey guys, do you think it's possible for an undergraduate to write a paper completely on his/her own?23:59
--- Log closed Tue Apr 17 00:00:09 2012