Open in new window / Try shogun cloud
--- Log opened Tue Mar 26 00:00:16 2013
-!- FSCV_ [~FSCV@187.210.54.166] has joined #shogun00:09
-!- FSCV [~FSCV@187.210.54.166] has quit [Read error: No route to host]00:09
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]00:36
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]01:46
-!- FSCV_ [~FSCV@187.210.54.166] has quit [Quit: Leaving]03:35
shogun-buildbotbuild #333 of nightly_default is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/33303:43
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has joined #shogun07:33
-!- fendow [deb20af4@gateway/web/freenode/ip.222.178.10.244] has joined #shogun08:40
fendowIn the chi-2,what's the meaning of the cache size?08:42
-!- fendow [deb20af4@gateway/web/freenode/ip.222.178.10.244] has quit [Quit: Page closed]08:51
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun08:59
-!- fendow [deb20af4@gateway/web/freenode/ip.222.178.10.244] has joined #shogun09:02
fendowIs there any body?09:03
n4nd0fendow: Sure09:03
fendowI have a question when using kernel_chi2, what's the meaning of the cache size?09:04
n4nd0fendow: I have not used that kernel in particular, I will take a look now in a moment. But let me do a wild guess09:04
n4nd0It should be the number of computed values that are stored in memory in case they need to be computed again, they are just looked up in memory instead of re-computed09:05
fendow0_009:07
n4nd0fendow: ?09:07
n4nd0fendow: it is not really the number of values, according to the doc in Kernel.h the cache size is measure in MB09:08
fendowok,I know~~09:09
fendowthanks.09:10
n4nd0you are welcome09:11
-!- kakashi_ [~kakashi_@nltk/kakashi] has quit [Read error: Operation timed out]09:45
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun09:56
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has quit [Ping timeout: 245 seconds]10:46
fendowIs there any matlab interfaces for a series of modelselection_grid_search functions?10:51
n4nd0fendow: Shogun's interface for matlab is an static interface and unfortunately most of the features in Shogun cannot be accesssed from static interfaces10:54
sonne|workfendow: no, but you could use octave_modular for that10:55
-!- heiko [~heiko@nat-172-192.internal.eduroam.ucl.ac.uk] has joined #shogun11:31
sonne|workheiko: I've converted the mc examples to just use SGVector/SGMatrix's load functions11:32
sonne|work(not yet committed everything)11:32
sonne|workn4nd0: about SO learning11:33
sonne|workn4nd0: kernels are a no go speed wise11:33
sonne|workit is just too slow that is why the coffin trick 'aka' emulating the kernel as he calls it, is the only reasonable way11:34
-!- fendow [deb20af4@gateway/web/freenode/ip.222.178.10.244] has quit [Quit: Page closed]11:34
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]11:35
heikosonne|work: I saw one PR, good!11:37
heikosonne|work: do we have our git process documented somewhere? people on ithub are asking how to commit, and I answered quite vaguely :)11:38
-!- lambday [b613104d@gateway/web/freenode/ip.182.19.16.77] has joined #shogun11:39
sonne|workheiko: I am not there yet - IMHO we need to first get tests running11:40
sonne|workthen we switch11:40
heikosonne|work agreed!11:40
sonne|work(tests running or blacklisted that is)11:40
heikoyeah11:41
heikosonne|work I asked Oli stegle to co-supervise the GPC gsoc project11:42
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun11:42
sonne|workheiko: what did he say?11:42
heikoyes11:42
heikosonne|work, but only for consulting11:42
sonne|workthat is what mentoring is about :)11:43
heikoso I wanted to ask you, what do you think, is it too much if I am involved in two projects where each of them has a co-mentor?11:43
heikosonne|work, well he has not too much time thats the point :) but I thought it might increase chances to make the project work if he is involved a bit11:43
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 252 seconds]11:43
heikosonne|work, whats the state of the application, can I help there somewhere? last time I asked blackburn told me to wait a bit since you were editing11:45
blackburnheiko: hey11:45
sonne|workheiko: knowing you, that certainly is not enough - you need to write your PhD thesis while doing all that and write a NIPS paper!11:45
heikoblackburn: hi :)11:45
blackburnyes that should be dicussed11:45
blackburnsonne|work: do you mind updating it on gdocs and synchronizing?11:45
sonne|workblackburn: I didn't do anything11:45
sonne|work(on the proposal)11:46
blackburnsonne|work: I know you put a dummy - I updated it with shared doc content11:46
sonne|workI would start with what we had last year11:46
heikosonne|work:  haha ! well no, should be better this summer, probably submit stuff to nips and have a job but these are the same thing11:46
blackburnsonne|work: are you unhappy with current stuff?11:47
sonne|workblackburn: haven't had a look but why reinvent the wheel?11:47
blackburnsonne|work: well things are changing11:48
blackburnheiko: do you have last year thing?11:48
sonne|worksome minor adjustments11:48
heikoblackburn: you sent it on google I think11:48
blackburnheiko: anyway lets work on gdocs thing11:49
blackburnmay be I should put last year stuff there to merge it too11:50
heikoblackburn: ok, I will do some this evening11:50
heikooh btw guys, I got accepted for my phd position/funding this week :) will start in Sept11:51
blackburnheiko: congrats!11:51
heikothanks :) very happy about that!11:52
blackburnheiko: what will you do?11:52
heikoblackburn: kernels! :D11:52
blackburnheiko: oh btw do you remember gsoc ends on 27, sept?11:52
heikoblackburn: the only constraint is major: machine learning, minor: computational neuroscience11:52
heikoblackburn: I know thats when my program starts11:52
blackburnheiko: what a coincidence :)11:53
heikoyeah11:53
blackburnheiko: so do I understand right you are free to do what you want?11:54
blackburnheiko: this colouring guy is insane :)11:55
heikoblackburn: well there is no mandatory list or something - however, supervisor is arthur gretton so thats why I said kernels11:55
heikoblackburn:  why?11:55
blackburnheiko: 'I can't wait' :D11:55
blackburn2 hours11:55
blackburn:D11:55
heikooh yeah11:55
heikowell, thats good :)11:55
heikoblackburn: chinese and indian people work harder :)11:56
blackburnheiko: yeah we are totally lazy here11:56
blackburncomparing to11:56
sonne|workno one works harder than heiko and blackburn11:56
heikoblackburn:  check this11:56
heikohttp://yuzhuoran.wordpress.com/learning-list-for-2012-fall/11:56
blackburnheiko: 86-101 hours/week?11:57
blackburn:D11:57
blackburnI'd rather die11:57
heikohaha :)11:57
blackburnheiko: I used to follow a schedule when I was young and grass was green11:58
heikoblackburn, "when I was young" tztztz11:58
blackburnheiko: oh yeah when I was 1711:58
blackburn:D11:58
heikooh wow :) I wasnt organised at all when I was 1711:58
blackburnheiko: I had something11:58
blackburnlike11:59
blackburn2 hours of calculus11:59
blackburn1.5 hours of linear algebra11:59
heikowow11:59
blackburnmay be it helped me later I don't know11:59
blackburnbut I had got more relaxed after an year may be :)11:59
heikowow thats early12:00
heikoprobably good12:00
blackburnheiko: but you were at school when you was 17 right?12:00
heikoyes, more or less12:00
blackburnit was university for me12:00
heikostarted computer-science when I was 23 ;)12:01
blackburnwhat have you studied before?12:01
heikomusic, but only 1.5 yrs, and thats not really studying12:01
blackburnI am with computer stuff from the very beginning :D12:02
heikoblackburn:  now you know the reason why I sometimes fail so badly when trying to do these technical things with computers - aka git problems ;)12:04
blackburnheiko: git can be troublesome for anyone..12:06
heikoor calling lapack functions12:06
blackburnheiko: if you spent some time on lapack you could get used to that too (I did)12:06
heiko I just cannot control my anger and then always ask you12:07
heikojoking :)12:07
blackburnbut lapack is not the thing one should write in 201312:07
heikoblackburn: indeed, eigen3 is the cool stuff!12:07
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun12:07
blackburnheiko: eigen slow downs compilation though12:07
blackburnsonne|work wouldn't like it :)12:07
n4nd0sonne|work: aham, ok I understand12:07
heikoblackburn:  I dont care  :)12:07
blackburnheiko: tapkee compiles in a few minutes because of heavy eigen3 usage12:08
heikoblackburn: development time also counts12:08
blackburnheiko: yeah true12:08
heikoblackburn: and bugs12:08
blackburnheiko: yeah it allowed me and n4nd0 to create fastest dim reduction stuff12:10
blackburn:)12:10
blackburnheiko: btw have you heard about t-sne?12:11
heikoblackburn:  no what is it?12:11
blackburncool DR method12:11
blackburnheiko: minimizes KL or so12:12
n4nd0blackburn: speaking of which, could you generate html doc from the md file?12:12
heikocool12:12
blackburnn4nd0: ahh yeah will do this evening12:12
n4nd0blackburn: good :)12:12
heikooh blackburn, we have a cool idea here: find a projection of data that maximises the MMD/HSIC between data12:12
blackburnheiko: the author, laurens van der maaten, released it under BSD 4 clause12:13
heikowhich means you map to a space where data is statistically farest apart/ most independent12:13
blackburnso it is available in tapkee12:13
heikoblackburn: cool!12:13
heikowell done  :)12:13
blackburnheiko: is that a method?12:13
blackburnor idea?12:13
heikoblackburn: idea12:13
blackburnlets work it out and write a paper then :D12:13
heikoyes! :)12:13
blackburnheiko: sometimes I think my destiny is to do dimension reduction stuff :D12:14
heikoblackburn: it probably is :D12:14
blackburnI have been into it for a few years already probably12:14
blackburnheiko: about spectral stuff12:15
blackburnheiko: are you sure eig() calls arpack?12:15
heikoit doesnt12:15
heikobut "eigs" does12:15
blackburnheiko: is there partial eigendecomposition in matlab?12:15
blackburnahh12:15
heikoblackburn: I had some problems with that12:16
blackburnheiko: a few algorithms from 2000s are based on smallest eigenvectors12:16
heikomemory excesses on generalised eigenproblems12:16
blackburnnot only LLE but laplacian eigenmaps and etc12:16
heikoblackburn: very good that you told me12:16
heikoI didnt know you knew about that stuff12:16
heikowill probably ask you12:16
heikoand also I think I will go for ARPACK for now12:16
blackburnheiko: yes I am messing with arpack all the time12:16
blackburnheiko: I am watching FEAST solver too12:17
blackburnheiko: planning to integrate it during gsoc with help of student12:17
heikocool!12:17
heikogood12:17
heikowe will use that12:17
heikofor the big-gaussian thing12:17
blackburnheiko: the best thing about ARPACK and FEAST is reverse communication interface (RCI)12:17
n4nd0blackburn: what does the feast solver have special?12:17
blackburnn4nd0: it should be faster than arpack (I think)12:18
blackburnit uses some complex magic12:18
blackburn:D12:18
n4nd0haha dark powers12:18
blackburnheiko: btw check this out12:19
blackburnheiko: STL, RCI, sonne|work's COFFIN12:19
blackburnit is all the same12:19
blackburndecoupling data and algorithms12:19
heiko?12:19
blackburnheiko: in stl algorithms are somewhat decoupled from data storage12:20
blackburnin RCI solvers only matrix-vector product is needed - that doesn't require to provide full matrix to solver12:21
lambdayhello.. this is soumyajit... I have configured shogun with eigen3... if I want to write a sample code with eigen specific task, e.g. calculating log-determinant of a dense matrix, wouldn't including <shogun/mathematics/eigen3.h> suffice?12:21
blackburnin COFFIN you separate data storage and svm training12:21
heikoblackburn: yes, thats the whole point of all sparse methods right?12:21
blackburnheiko: not really sparse - it is generic12:21
heikohi lambday12:21
blackburnarpack works as well for dense and sparse12:22
lambdayhello heiko12:22
blackburnas well as coffin12:22
heikoblackburn: how do you include eigen?12:22
heikoI just do #include <Eigen/Dense> for example12:23
blackburnheiko: better do shogun/mathematics/eigen3.h12:23
heikolambday: then yes!12:23
blackburnthere is a good reason to do that12:23
lambdayheiko: hmm that works.. but I am wondering why just including eigen3.h doesn't work :-/12:23
n4nd0lambday: why it doesn't work? what error do you get??12:24
blackburnheiko: do you include eigen directly?12:24
heikoblackburn: I did for the GP, yes, should I change? what is the good reason?12:24
blackburnheiko: https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/mathematics/eigen3.h12:24
blackburnheiko: check defines12:24
blackburnwe could put there more defines like EIGEN_NO_DEBUG (IIRC)12:25
blackburnwe would need to reproduce them again and again12:25
heikoblackburn: ok, agreed!12:25
lambdayn4nd0: do I have to give the include directory while compiling for eigen3 when I am just using shogun's eigen3.h?12:25
heikowill change12:25
blackburnheiko: alright thanks12:26
blackburnlambday: if it is detected you don't have to add anything12:26
lambdayI am getting errors for both "g++ eigen_test1.cpp -lshogun" and "g++ -I /usr/local/include/eigen3 eigen_test1.cpp -lshogun"12:26
blackburnahh12:26
blackburnyes then you would have to include it12:26
lambdayblackburn: that's what I thought... its detected since i configured it using --enable-eigen3 with --include=/path/to/eigen/headers12:26
blackburnlambday: paths are provided only for shogun sources - why do you develop a separate file?12:27
lambdayblackburn: umm.. I didn't understand.. which file? eigen_test1.cpp is just a test file12:28
lambdaywhen I put <Eigen/Dense> and compile it with "g++ -I /usr/local/include/eigen3 eigen_test1.cpp -lshogun" it works12:28
blackburnlambday: sure that should work this way12:29
blackburnI mean you'd have to put all the includes shogun uses if you use it this way12:29
blackburnif you needed something else you'd need to put it too12:29
n4nd0I am guessing your HAVE_EIGEN3 variable is not defined12:29
n4nd0aaah ok sure if you use g++ then those variables are not defined12:30
heikolambday: when you are doing this log-det method, add it to the CStatistics class and use that code from you example program -this way you dont have these problems12:30
lambdayn4nd0: isn't that done automatically? or I have to define it somewhere?12:30
n4nd0lambday: well if you use g++ how is that going to be done automatically? :)12:31
lambdayheiko: okay :)12:31
blackburnshogun uses its own structure of makefiles which put all paths and etc12:31
lambdayn4nd0: hmmm12:31
heikolambday: you then have to re-compile shogun when you change something, but if you install a compiler cache, that is very fast12:31
n4nd0it is not like installing shogun is going to change the default behaviour of the C++ compiler12:31
n4nd0that would be inconvenient I think hehe12:32
lambdayheiko: log-det for dense matrix is working as per my code.. checked with a small pd symm matrix and log-det(A) and log(det(A)) are giving same results12:33
heikolambday: that is good12:33
heikolambday: this is more of a warming up, the symamd stuff for sparse matrices is a little more complicated, but not vastly12:33
heikolambday: integrate into shogun, write example and a few tests, document, and send pull request then we will merge it12:34
lambdayheiko: alright.. will keep you updated..12:34
heikolambday: nice! thanks12:35
lambdayheiko: I have never used compiler cache :(12:36
heikolambday: its very easy to install if you are under linux12:36
lambdayokay.. yes I am using linux only.. I'm trying it.12:36
heikoccache is the debian(/ubuntu package and the man-page contains a little guide how to install12:36
heikowhich is done via creating 3 symbolic links, and done12:37
blackburnhint: add CXX="ccache g++" to your .bashrc12:37
lambdayand one more thing, where exactly do we have to define the HAVE_EIGEN3 variable to make <shogun/mathematics/eigen3.h> work?12:37
blackburnlambday: shogun defines it on ./configure12:37
blackburnand put it to shogun/base/config.h12:37
n4nd0lambday: you should get it defined automatically running configure12:38
n4nd0if shogun detects your eigen3 of course12:38
lambdayheiko: I am using fedora linux, ccache should be usable on this also I guess12:38
heikoi guess so12:38
n4nd0but this HAVE_EIGEN3 variable will work using the Makefiles12:39
lambdayblackburn: so we have to do that manually? because I configured shogun as "./configure --interfaces=cmdline_static --enable-eigen3 --includes=/usr/local/include/eigen3" and it showed that it found eigen3 in the config msg12:40
lambdayn4nd0: hmmm... let me configure it once more... something must have went wrong :-/12:40
n4nd0lambday: I don't think something went wrong12:41
n4nd0the thing is you give variables to the c compiler using something like12:41
n4nd0g++ -DHAVE_EIGEN3 blah blah blah12:41
n4nd0but of course you don't want to write that manually (there are lot of those variables around)12:42
lambdayn4nd0: then I should already have HAVE_EIGEN3 defined, and hence the includes in eigen3.h (which includes Eigen/Dense) should work, no?12:42
n4nd0so you should use the Makefiles, as blackburn suggested, not compile on your own I think12:42
lambdayn4nd0: ohkay... got it12:43
lambdayheiko, n4nd0, blackburn thanks a lot :)12:43
n4nd0;)12:43
n4nd0sonne|work: so what do you think about Patrick's project, does it look enough for you? I think it is.12:48
-!- kakashi_ [~kakashi_@nltk/kakashi] has joined #shogun12:50
-!- FSCV [~FSCV@204.45.132.11] has joined #shogun14:32
sonne|workn4nd0: yes looks good!14:40
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has joined #shogun14:46
-!- SCV__ [~FSCV@76.73.18.131] has joined #shogun14:47
-!- FSCV [~FSCV@204.45.132.11] has quit [Ping timeout: 264 seconds]14:48
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has quit [Ping timeout: 245 seconds]14:51
-!- FSCV_ [~FSCV@65.19.131.247] has joined #shogun14:56
-!- SCV__ [~FSCV@76.73.18.131] has quit [Ping timeout: 240 seconds]14:57
heikoblackburn15:16
heikohow to sum the elements in a matrix row-wise in eigen?15:17
heikoah15:18
heikofound it15:18
-!- SCV__ [~FSCV@216-230-229-167-colo.oplink.net] has joined #shogun15:19
heikoA.colwise().sum()15:20
heikonice!15:20
n4nd0heiko: eigen is awesome :)15:21
n4nd0It is like coding matlab/python in C++15:21
heikon4nd0: indeed :)15:21
-!- FSCV_ [~FSCV@65.19.131.247] has quit [Ping timeout: 246 seconds]15:21
heikoyes, I like that a lot15:21
n4nd0and it has soooo much stuff15:21
heikowe could actually remove half of the shogun math code :D15:21
n4nd0I am not even aware of the 10% probably15:22
n4nd0hehe15:22
heikon4nd0: This col/row-wise mean in CStatistics for example15:22
heikon4nd0: same here!15:22
heikojust discovering ...15:22
heikon4nd0: we have to be careful though, we are hyping it currently :D15:22
n4nd0hehe yeah, maybe :D15:23
blackburnheiko: sorry didn't notice15:23
heikoblackburn: no worries15:23
heikothe tutorial is very nice, answers many questions15:23
blackburnI am thinking about helping them somehow15:23
heikoblackburn: I am pushing this GPU cholesky guy to implement it in eigen :)15:24
blackburnheiko: quite fair hype though15:24
lambdayheiko: I added a log_det method in the CStatistics class... pasting the code ... Statistics.h [http://pastebin.com/H24CAPec] (see line#456-463) and Statistics.cpp [http://pastebin.com/taWLjYFR] (See line#25-28, 1962,1978)15:25
lambdaydid a make and it compiles15:25
blackburnlambday: you should rather use github stuff for that :)15:25
lambdayblackburn: as in commit?15:26
blackburnlambday: as pull request15:26
heikolambday: blackburn is right, commit it to your local repo, push, and send a pull request15:26
heikomuch easier15:26
heikolambday: but the code looks good!15:26
lambdayheiko: thanks.. but haven't done testing yet... just did the coding15:27
heikolambday: write a unit test which covers a few cases (small and large)15:27
heikolambday: you can find inspiration in the tests/unit folder, probably have to add a new file (cannot recall if there already is one for CStatistics)15:28
heikolambday: then send the PR and I will comment more, will be back later today15:28
lambdayheiko: okay.. thanks man :)... I am checking the tests/unit dir15:29
heikolambday: nice! :)15:29
lambdayheiko: I know I am sound like a total idiot but could you please tell me how do I send the pull request? I am new to git :(15:41
lambdayI already committed using this ---- git commit -a -m "Added log_det method in CStatistics class"15:41
n4nd0hey lambday, you are not an idiot, it is normal to struggle with git a bit at the beginning15:42
lambdayn4nd0: :(15:43
n4nd0so when you do a commit your changes have only been stored in your computer, locally15:43
lambdayn4nd0: how do I do it?15:43
lambdayn4nd0: hmmm...15:43
n4nd0then you need to publish your changes into your repository. This is call push in git15:43
n4nd0something like git push origin master should do it15:43
n4nd0although it is not the best way to develop directly in your master branch...15:44
lambdayokie15:44
n4nd0but let's better learn git step by step, so for the moment it can be ok15:44
lambdayso while developing something, one usually work on a forked branch or something?15:44
lambdayn4nd0: okay..15:45
n4nd0I normally use the term fork to refer to whole repositories, not only a branch15:45
n4nd0for example, now you are working in a fork of shogun15:45
n4nd0but your fork has several branches (master, among others)15:46
n4nd0once you have push your changes to your repository15:46
n4nd0you can issue a pull request (PR) using the github web interface15:46
n4nd0the PR means that you want to introduce the changes you have made in your shogun fork into the main shogun repository15:47
n4nd0but please update your branch before doing the PR, you should actually do this after committing and before pushing15:50
lambdayn4nd0: update branch?15:51
n4nd0introduce in you fork the changes that have taken place in the main shogun repo15:52
n4nd0you can do this with git pull upstream master or, even better, git pull --rebase upstream master15:52
n4nd0upstream is a remote pointing to the main shogun repo that I think you will need to define15:53
lambdayn4nd0: okay...15:53
-!- FSCV_ [~FSCV@206.225.135.24] has joined #shogun15:54
-!- SCV__ [~FSCV@216-230-229-167-colo.oplink.net] has quit [Ping timeout: 245 seconds]15:55
lambdayn4nd0: I need to define upstream.. as in "origin" points to "https://github.com/shogun-toolbox/shogun.git" in github, something like that?15:55
n4nd0lambday: mmm origin should point to your fork15:56
lambdayohkay..15:56
heikolambday: upstream for the shogun repo15:59
lambdayshould "git push" ask for github username and password? in my system its asking when I try to do git push origin master...16:06
n4nd0lambday: yeah, you can avoid that though introducing your ssh public key in your github account16:07
lambdayn4nd0: hmm... I cannot do git push origin master, when I try it shows "403 while accessing https://github.com/shogun-toolbox/shogun.git/info/refs"... I am not getting if origin points to my forked repo, why would it try to access remote shogun repo?16:15
n4nd0lambday: is it possible that you cloned shogun-toolbox/shogun.git instead of lambday/shogun.git?16:16
n4nd0i.e. the shogun repo instead of your fork16:17
lambdayoh shit!16:17
lambdaythat is why!16:17
lambdayn4nd0: thanks :(16:17
n4nd0lambday: no problem16:17
n4nd0don't be sad, it is not a big deal ;)16:18
n4nd0just clone your fork16:18
n4nd0copy the changes you made there (the ones in CStatistics.[h|cpp] I believe)16:18
n4nd0commit and push, and then PR from github.com16:18
lambdayn4nd0: okay... my fork [https://github.com/lambday/shogun.git] also contains two more files that are not in shogun, which I added durng an assignment few weeks back.. would it create problem?16:20
n4nd0if they are new files just copy them16:21
n4nd0if they are files that were already in shogun that you modified16:21
n4nd0then make a diff and introduce your changes16:21
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun16:21
lambdayn4nd0: okay... those were new files..16:22
n4nd0copying them directly in this second scenario could be a bit bad in case those files tracked in the repo had been modified by someone else, and those changes would be overwritten when you copy the files16:22
lambdayn4nd0: hmmm...16:23
n4nd0but nvm, just copy or move them if they were new16:23
lambdayn4nd0: alright16:24
n4nd0ok I am off now for a while16:28
n4nd0see you later16:28
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]16:28
lambdayn4nd0: thanks man.. see you :)16:29
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Ping timeout: 240 seconds]16:30
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:32
-!- hoijui [~hoijui@141.23.67.236] has joined #shogun16:35
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun16:45
-!- hoijui [~hoijui@141.23.67.236] has quit [Ping timeout: 264 seconds]16:54
-!- hoijui [~hoijui@141.23.67.236] has joined #shogun16:59
lambdayheiko, n4nd0 I have submitted the pull request. Will see about the test cases.17:00
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun17:09
shogun-notifier-shogun: Soeren Sonnenburg :master * 0574d56 / / (9 files): https://github.com/shogun-toolbox/shogun/commit/0574d5669bd561e4f4b65256f56b53a7c7c4ff7517:09
shogun-notifier-shogun: Don't use streaming features for simple SGMatrix/SGVector i/o17:09
shogun-notifier-shogun:17:09
shogun-notifier-shogun: Instead use their load routines. Require splitted up 7 class data17:09
shogun-notifier-shogun: example (into data and labels) and adjust news accordingly. Drop17:09
shogun-notifier-shogun: unnecessary includes in examples (don't!)17:09
-!- hoijui [~hoijui@141.23.67.236] has quit [Read error: Operation timed out]17:11
-!- blackburn [~blackburn@188.168.14.165] has joined #shogun17:14
@sonney2kblackburn, have you seen? a first PR...17:15
@sonney2kalready now17:15
blackburnsonney2k: yeah17:15
blackburnsonney2k: quite fast yeah17:16
@sonney2kshogun-buildbot, force build "deb3 - modular_interfaces"17:18
shogun-buildbotbuild forced [ETA 47m08s]17:18
shogun-buildbotI'll give a shout when the build finishes17:18
@sonney2kshogun-buildbot, force build "rpm1 - libshogun"17:19
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Ping timeout: 245 seconds]17:19
shogun-buildbotbuild forced [ETA 4m50s]17:19
shogun-buildbotI'll give a shout when the build finishes17:19
shogun-buildbotbuild #378 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/37817:25
shogun-buildbotbuild #961 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/96117:46
heikosonney2k, blackburn I will take care of the PR17:49
heikolambday: nice, will look at it later, currently busy17:50
heikosonney2k: can I trust travis?18:06
heikoif it sais build failed?18:06
heikoblackburn, sonney2k: are these kind of calls still needed? ((CStringFeatures<char>*) lhs)->free_feature_vector(avec, idx_a, free_avec);18:12
blackburnheiko: uglyy18:28
heikoright, I thought so, rather use SGVector nowadays right?18:30
blackburnheiko: I don't know exactly18:32
blackburnheiko: what I actually would like to see is18:34
blackburneach kernel should implement some kernel(SGVector, SGVector) or so18:34
blackburnother parts like get/free should be generic18:34
blackburnheiko: when did you contact sonney2k for the very first time?18:54
heikoblackburn: 15 months before the gsoc 201118:54
blackburnheiko: whoop18:54
heiko2010 shogun got rejected, was before that18:54
blackburnheiko: ohh so you were around before dinosaurs18:55
heikono that was 2011 , even before :)18:55
blackburnheiko: 27 march, 2011 is the day I joined #shogun :D18:55
blackburncelebrating! :D18:55
heikoha18:56
heikotmorow! :)18:56
blackburnheiko: did you want to participate in 2010?18:58
heikoblackburn: yes18:59
blackburnohh I see18:59
heikofound the modelselection idea18:59
heikowas the time when I wrote my Bsc18:59
heikowhich was on modelselection among other things18:59
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has joined #shogun19:00
-!- lambday [b613104d@gateway/web/freenode/ip.182.19.16.77] has quit [Ping timeout: 245 seconds]20:02
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]20:09
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has joined #shogun20:13
lambdayheiko: hi... got your messages, will be updating a description of the dense matrix log_det computation soon..20:15
lambdaythe ssk string kernel part was done a few weeks back by me for one of our class assignments... will make the necessary changes as you suggested..20:15
heikolambday: cool!20:15
heikolambday: remove the string kernel for now20:15
heikolets merge it after the chol stuff20:16
lambdayheiko: that was already a part of my forked repo... so got committed during the pull req... but if its possible, for now I'd only like to work on the large scale gaussian project20:16
heikootherwise things will be messy20:16
lambdayheiko: alright... removing the ssk kernel20:16
heikolambday: just temprarily remove and store somewhere20:16
heikolambday: we can add this later, but needs to be cleaned up a bit, since based on very old shogun code :)20:17
lambdayheiko: okay... :) I'm updating20:17
heikolambday:  the large scale Gaussian is the cool stuff ;)20:17
heikolambday: I will head home now, might not be online today, but tomorrow for sure, see you!20:18
lambdayheiko: yeah it is... I was reading the pdf that you posted on the gsoc ideas page... really cool stuff... although I feel I need to understand the maths a bit more deeper :-/20:18
lambdayheiko: alright man... its almost 1 am here in india.. I'll also get some sleep... see you tomorrow :)20:19
heikolambday: good night!20:19
lambdayheiko: good night :)20:20
-!- heiko [~heiko@nat-172-192.internal.eduroam.ucl.ac.uk] has left #shogun []20:23
-!- lambday [3ba2174d@gateway/web/freenode/ip.59.162.23.77] has quit [Quit: Page closed]20:35
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun21:16
@sonney2kblackburn, does classifier_svmlight_string_features_precomputed_kernel work on your machine?21:28
blackburnsonney2k: let me check21:28
@sonney2k(it is here plus it is valgrind clean)21:29
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun21:32
shogun-notifier-shogun: Soeren Sonnenburg :master * c24c395 / examples/undocumented/libshogun/classifier_svmlight_string_features_precomputed_kernel.cpp: https://github.com/shogun-toolbox/shogun/commit/c24c3954550f6cf489107d2310c9ef1ad442cf6e21:32
shogun-notifier-shogun: seed random number generator21:32
@sonney2kblackburn, could you please make your examples not fail when eigen3 is not available?21:36
@sonney2khttp://shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/378/steps/test/logs/stdio21:36
blackburnsonney2k: ohh21:36
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Remote host closed the connection]21:37
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun21:37
shogun-buildbotbuild #379 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/379  blamelist: Soeren Sonnenburg <sonne@debian.org>21:48
blackburnsonney2k: will fix that tom morning21:49
shogun-buildbotbuild #380 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/380  blamelist: Soeren Sonnenburg <sonne@debian.org>21:53
blackburnsonney2k: yes that example works smooth21:54
shogun-buildbotbuild #963 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/963  blamelist: Soeren Sonnenburg <sonne@debian.org>22:21
shogun-buildbotbuild #962 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/96222:22
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]22:54
-!- hoijui [~hoijui@dslb-088-075-046-171.pools.arcor-ip.net] has quit [Ping timeout: 264 seconds]23:57
--- Log closed Wed Mar 27 00:00:16 2013