Open in new window / Try shogun cloud
--- Log opened Tue Apr 17 00:00:09 2012
--- Day changed Tue Apr 17 2012
emreceliktenand get accepted to a conference?00:00
emreceliktenNo guidance from anyone whatsoever00:00
-!- blackburn [~qdrgsm@] has quit [Ping timeout: 248 seconds]00:01
emreceliktenApparently no :P00:46
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 276 seconds]01:04
-!- av3ngr [av3ngr@nat/redhat/x-ucigzyknjltzoyko] has joined #shogun01:48
-!- wiking [] has joined #shogun02:01
-!- wiking [] has quit [Changing host]02:01
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun02:01
-!- emrecelikten [~Anubis@] has quit [Quit: Leaving.]02:03
-!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection]02:45
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun02:45
-!- vikram360 [~vikram360@] has quit [Ping timeout: 272 seconds]06:13
-!- vikram360 [~vikram360@] has joined #shogun08:13
-!- vikram360 [~vikram360@] has quit [Read error: Connection timed out]08:44
-!- vikram360 [~vikram360@] has joined #shogun08:45
-!- gsomix [~gsomix@] has quit [Ping timeout: 240 seconds]08:59
CIA-64shogun: Josh Klontz master * rb8bec85 / (5 files in 5 dirs): added basic CMake project files -
CIA-64shogun: Soeren Sonnenburg master * rce82cc1 / (5 files in 5 dirs):09:06
CIA-64shogun: Merge pull request #461 from jklontz/master09:06
CIA-64shogun: Remove some deprecated function calls and add cmake project files. -
-!- uricamic [9320543b@gateway/web/freenode/ip.] has joined #shogun09:13
-!- vikram360 [~vikram360@] has quit [Ping timeout: 252 seconds]09:18
-!- pluskid [] has joined #shogun09:20
sonne|workpluskid: what is the lasso parameter doing in CLARS?09:24
pluskidsonne|work: you mean the regularization coefficient?09:25
sonne|workthe parameter in the constructor09:25
sonne|workit can be true or false09:25
pluskidoops, cat on keyboard09:25
pluskidwhen it is true, it runs the LASSO09:25
pluskidwhen false, it runs LARS09:26
pluskidthe LARS algorithm has a modification, which could give us a LASSO solution09:26
pluskidwhen passed true in the constructor, this modification is turned on09:27
pluskidI think the most case we use LARS is to get the LASSO solution (instead of the original LARS), so I made that parameter default to true09:27
pluskidsonne|work: am I clear?09:27
sonne|workthat was just missing from the doc09:30
CIA-64shogun: Soeren Sonnenburg master * r879ba9e / src/shogun/regression/LARS.h : fix doxygen warning in lars -
sonne|workhowever I still don't know what the difference between LASSO / LARS is09:30
pluskidsonne|work: thanks for the fix of doxygen09:31
pluskidin my understanding, LARS is a stagewise variable selection algorithm09:32
pluskidLASSO in its general meaning corresponds the the linear regression with L1-regularization09:33
pluskidthe original LARS algorithm doesn't corresponds to any objective function though09:33
pluskidbut when the *lasso modification* is enabled in the LARS algorithm, it will provide the solution for the LASSO problem09:34
pluskidwhen referring LASSO, we generally mean the problem09:34
pluskidand there are many different algorithms for solving LASSO09:34
pluskidsonne|work: that's my understanding, if you have interest, you may refer to the paper for more details :)
-!- av3ngr [av3ngr@nat/redhat/x-ucigzyknjltzoyko] has quit [Remote host closed the connection]09:40
-!- av3ngr [av3ngr@nat/redhat/x-utoswxpenjswqpzs] has joined #shogun09:41
-!- av3ngr [av3ngr@nat/redhat/x-utoswxpenjswqpzs] has quit [Client Quit]09:45
sonne|workpluskid: ok I will have to read that09:47
sonne|workpluskid: I understand the problem with multiclass/kernel machine09:47
sonne|workone cannot set a multiclass machine as constraint generator09:49
sonne|workthere is a separate multiclass mkl for that09:49
sonne|workso you should be fine09:49
pluskidsonne|work: that's nice, so we would be safe to simply remove the (currently) commented out code in GUIClassifier09:54
uricamichi there10:26
uricamicI have some problems with git push10:26
uricamicI have committed changes to my branch, followed instructions from and when I try to git push origin my_branch it results in this error: failed to push some refs to ''10:28
pluskiduricamic: what is the detailed output of git?10:40
uricamicpluskid: you mean after that push?10:40
uricamicTo  ! [rejected]        GSKRR -> GSKRR (non-fast-forward) error: failed to push some refs to '' To prevent you from losing history, non-fast-forward updates were rejected Merge the remote changes (e.g. 'git pull') before pushing again.  See the 'Note about fast-forwards' section of 'git push --help' for details.10:41
pluskiduricamic: it seems you have rebased your local branch, isn't it?10:42
pluskidone simple solution is to first delete your remote branch, and then push again10:42
pluskidsomething like10:42
pluskidgit push origin :branch-name10:42
pluskidand then10:42
uricamicpluskid: yes, I have followed this link:
pluskidgit push origin branch-name10:42
uricamicdelete my remote branch?10:43
pluskidyes, try the command I mentioned above10:43
uricamichow can I delete my remote branch?10:44
pluskidgit push origin :branch-name10:44
pluskidnote the colon (:) before the branch name10:44
uricamicok, I will try this10:44
uricamicpluskid: thanks, it seems that this worked10:45
pluskidalthough it seems you can also solve this problem by using git pull, but that will introduce an empty commit for merging, so I think the method here is more cleaner10:45
shogun-buildbotbuild #477 of ruby_modular is complete: Success [build successful]  Build details are at
pluskiduricamic: then you should be able to push your branch to origin now :)10:46
-!- n4nd0 [] has joined #shogun10:48
uricamicpluskid: ok, I think the pushed worked, but now when I have already some opened pull request for this, should I close it and do a pull request again?10:48
pluskiduricamic: you have to create multiple branches to make multiple pull request10:49
pluskidI think the pull request are to be closed by the maintainers when they merged your commits to the official shogon repo10:50
uricamicpluskid: yes, I already know this :) the question is, if I have to do the pull request again, or if the changes which I have pushed right now will appear in the original pull request10:50
pluskiduricamic: OK, I see, as long as the branch name does not change, your new commits will be automatically "merged" into the pull request10:51
pluskiduricamic: you can look at the pull request you originally opened, navigate to "commits" tab, and see your new commits there10:51
uricamicpluskid: ok, thanks again :)10:52
pluskiduricamic: u r wel~ :)10:53
n4nd0hey guys10:55
n4nd0so now you were talking about the topic10:55
n4nd0if a pull request is already opened10:55
n4nd0is there any danger to rebase with the changes added to remote master since that pull request was opened?10:56
pluskidn4nd0: I don't think so10:57
pluskidactually, rebasing is always encouraged , I think10:58
n4nd0ok, so I can just go to my master10:58
n4nd0rebase upstream/master10:58
n4nd0push origin master10:58
n4nd0go to my branch10:58
n4nd0rebase master10:58
n4nd0and later commit and push the new changes to the branch to update the pull request10:59
pluskidwhat github do is simply keeping a reference to the branch name you are using for that pull request10:59
pluskidwhen your branch changes (i.e. new commits are added), the pull request updates automatically10:59
n4nd0but I am afraid that the changes that enter in my branch when doing rebasing10:59
n4nd0the ones that are not mine10:59
n4nd0are introduced in the pull request as mine11:00
pluskidrebasing is inserting changes *before* your branch commits11:00
pluskidso when you run git rebase master11:00
pluskidit takes those commits in master that happens *after* you create the branch,11:01
n4nd0but now we are talking about changes that happened *after* the first changes in that pull request11:01
pluskidthose commits are inserted *before* the first commit in your branch11:01
n4nd0that makes sense11:02
pluskidhowever, we do have problem pushing to remote branch when we do rebasing11:02
pluskidif we have already pushed parts of our branch to origin11:03
pluskidwe get non-fast-forward error11:03
pluskidyou have encountered this error before, IIRC :)11:03
pluskidyou solved it with git pull11:03
pluskidwhich produce an empty commit of merging11:03
pluskidMy solution is to delete the remote branch first and push again11:04
pluskidwhich would be cleaner in my view11:04
n4nd0yeah, I think I had an error related to this before11:04
n4nd0but I think it was because I didn't push to origin master after rebasing11:05
pluskidoh, maybe11:05
n4nd0I will try it soon and let you know11:05
n4nd0I agree, to delete the remote branch looks cleaner11:05
-!- vikram360 [~vikram360@] has joined #shogun11:41
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun11:53
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds]11:56
-!- wiking_ is now known as wiking11:56
-!- nickon [] has joined #shogun11:58
-!- Marina [8d175a52@gateway/web/freenode/ip.] has joined #shogun12:08
-!- vikram360 [~vikram360@] has quit [Ping timeout: 272 seconds]12:09
-!- vikram360 [~vikram360@] has joined #shogun12:11
MarinaHi, I have the same problem with poims like Ralf Anders:                     can anyone help me12:13
n4nd0Marina: I have no experience in that part of the project but I can take it a look12:15
Marinathat would be great12:16
n4nd0Marina: so you get an error with a python script right?12:18
-!- Marty28 [~Marty@] has joined #shogun12:18
n4nd0Marina: in that case, can you please tell me the name of that file if included in shogun or give me a pointer to it?12:18
Marty28hi marina12:19
Marty28Marina: I guess I just wrote you an email, right?12:20
MarinaI got your email.12:21
Marty28fine :)12:22
Marinathe name of the file is poim.py12:26
-!- pluskid [] has quit [Quit: ChatZilla [Firefox 11.0/20120312181643]]12:27
Marty28Easysvm has some POIM.12:28
n4nd0Marty28, Marina: I don't know if you already solved this issue or not12:30
-!- pluskid [~pluskid@] has joined #shogun12:30
-!- pluskid [~pluskid@] has left #shogun []12:31
Marinano i didn't12:37
MarinaI made poims with half a year before and now it doesn't  work anymore....12:38
n4nd0Marina: all right12:38
n4nd0Marina: so where is this
Marty28It is in the easysvm package. It uses shogun-content.12:40
n4nd0ok found it12:40
Marinacan I send it to you per email?12:40
Marinaa yes that's it12:40
n4nd0Marina: ok, so can you tell me exactly how to reproduce the error?12:41
* n4nd0 is afk12:42
* n4nd0 is back12:44
n4nd0wiking: around?12:45
Marty28My code here:12:46
Marty28    print "Loading kernel"12:46
Marty28    k1 = rdata.dataio.load_kernel(basename + ".kernel.dat",True)12:46
Marty28    print "Loading SVM"12:46
Marty28    svm1 = rdata.dataio.loadbz2(basename + ".svm.bz2")12:46
Marty28    # POIM stuff12:47
Marty28    print "Computing POIMs"12:47
Marty28    from poim import compute_poims12:47
Marty28    poimdegree = 512:47
Marty28    max_len = 212:47
Marty28    #12:47
Marty28    print "Trying to POIM..."12:47
Marty28    poim, max_poim, diff_poim, poim_totalmass = compute_poims(svm1, k1, poimdegree, max_len)12:47
* Marty28 tries to remember IRC...12:47
wikingn4nd0: sup?12:48
n4nd0wiking: git support ... as usual :P12:48
n4nd0wiking: I don't know if you read my conv with pluskid before, it is about that12:48
n4nd0wiking: I can make a summary otherwise12:48
wikingsorry was away :(12:48
n4nd0wiking: ok so I openend a pull request12:49
n4nd0didn't work on it for a couple of days12:49
n4nd0now I am back12:49
n4nd0I rebased my origin master with upstream master, push to origin master12:49
n4nd0back in my branch where I do the work12:49
n4nd0I do some commits12:49
n4nd0rebase master12:50
n4nd0and when I try to push, the usual error12:50
n4nd0git push origin jlcovertree12:50
n4nd0To ! [rejected]        jlcovertree -> jlcovertree (non-fast-forward)12:50
n4nd0error: failed to push some refs to ''12:50
Marinaand if you use exactly the code from you get the error:  TypeError: WeightedDegreePositionStringKernel_prepare_POIM2() takes exactly 4 arguments (2 given)12:50
n4nd0To prevent you from losing history, non-fast-forward updates were rejected12:50
n4nd0Merge the remote changes (e.g. 'git pull') before pushing again.  See the12:50
n4nd0'Note about fast-forwards' section of 'git push --help' for details.12:50
Marty28compute_poims does     distr = ones((max_len,4))/4 ;12:50
Marty28then kernel.prepare_POIM2(distr, poimdegree, max_len)12:50
wikingn4nd0: and after doing a git pull?12:50
n4nd0wiking: I didn't pull yet because as far as I remember12:51
n4nd0wiking: last time I did that12:51
Marty28The error is only about the distr (numpy.ones generated it)12:51
n4nd0wiking: and my commits entered twice in the branch12:51
MarinaSo I have write kernel.prepare_POIM2(distr) to kernel.prepare_POIM2(distr, poimdegree, max_len) like you12:51
Marinaand than I get the same error12:51
wikingn4nd0: mmm that's interesting12:51
Marina TypeError: in method 'WeightedDegreePositionStringKernel_prepare_POIM2', argument 2 of type 'float64_t *'12:51
Marty28The error is only about the type of dist.12:52
wikingafaik a git pull would try to do a merge basically12:52
n4nd0wiking: I will pull and push and see what happens12:52
wikingpull will do you a merge thing12:52
wikingor at least will try12:52
n4nd0wiking: in any case, pluskid suggested to delete the remote branch doing git push :branch12:52
wikingsince what it'll try to do12:52
wikingis to sync your remote repo with your local changes12:52
wikingthat's another option of course12:53
n4nd0wiking: if I remove the remote branch, the pull request will disappear as well or not?12:53
wikingit's more brute force solution12:53
wikingbut it'll solve your thing12:53
Marty28So probably either WeightedDegreePositionStringKernel.prepare_POIM2 was changed or numpy.ones (improbable)12:53
wikingsince that commit only exist in that branch12:53
wikingbut i'm not 100% sure of github internals12:54
wikingso maybe they do save somewhere the pull requested commits12:54
wikingi wouldn't do but who knows12:54
MarinaMarty28: yes I think so too12:54
wikingmaybe for some caching purposes or something they'd do it12:54
n4nd0Marina, Marty28: mm I can execute without problems, maybe I am trying exactly what fails for you12:55
n4nd0maybe I am not, sorry12:55
n4nd0wiking: I am going to pull and push12:55
Marty28numpy.ones gives <type 'numpy.ndarray'>12:55
n4nd0wiking: hopefully the commits won't be duplicated12:55
Marty28prepare_POIM2 needs float64_t *12:55
wikingwell that happens when rebase/merge fails12:56
wikingbut it's not a real duplication...12:56
n4nd0wiking: git pull origin branch right?12:56
wikingyep the remote branch...12:56
n4nd0wiking: mmm, I got some conflicts :(12:57
wikingso the merge was not possible12:57
wikingi mean automatically12:57
wikingwhat are you conflict files?12:57
Marty28n4nd0: Of what type is your "dist" (1st parameter of prepare_POIM2)?12:57
n4nd0wiking: is there a way to indicate to accept my local version?12:58
wikingn4nd0: you'll have to fix it by yourself by hand12:58
n4nd0wiking: all right12:58
wikingn4nd0: check the git status12:58
wikingand you'll see the conflicted files12:58
wikingthere go in edit them as you please12:58
wikingand then do a git commit -a12:58
wikingthus you'll have the merge done12:58
n4nd0Marty28: I am taking a closer look to poim.py12:59
n4nd0Marty28: I can just see method definitions there13:00
Marty28    print "TYPE(distr): " + str(type(distr)) gives <type 'numpy.ndarray'> for me13:00
Marinayes right13:00
Marty28prepare_POIM2 needs 'float64_t *'13:01
n4nd0Marty28: yeah but what do you want me to execute?13:03
n4nd0Marty28: I mean, there must be another script making use of the functions in poim.py13:04
n4nd0Marty28: is it what you pasted above?13:04
Marinabut if I make print distr.dtype than the answer is  float6413:04
Marty28Our own?13:05
Marinano I've found it on
Marty28n4nd0: poim.compute_poims is used13:06
n4nd0Marty28: from where?13:06
Marty28From my script13:08
Marty28n4nd0: poim, max_poim, diff_poim, poim_totalmass = compute_poims(svm1, k1, poimdegree, max_len)13:08
n4nd0Marty28: yeah but ... how I can execute your script?13:08
Marty28Of course the script needs an svm and a kernel.13:09
n4nd0Marty28: so do you have all the code for me to reproduce the error in one file or in several files?13:09
Marty28I would have to prepare that, mom13:09
n4nd0you can paste your code there if so13:10
n4nd0wiking: :O it went pretty bad13:25
n4nd0wiking: I pull, solved the conflict13:25
n4nd0wiking: commit the merge13:25
n4nd0wiking: push all the changes and look13:26
n4nd0in my pull request the commits appear twice and I have even got commits not done by me13:26
sonne|workn4nd0: you forgot a rebase13:27
n4nd0sonne|work: really? in my branch or where?13:29
sonne|workn4nd0: at some point you did a git pull13:30
sonne|workand then a git push13:30
n4nd0summary of what I did13:30
sonne|workwhich will make all the commits you did before the pull not to appear on top13:30
sonne|workbut before that13:30
sonne|workso things look cluttered13:30
n4nd0the pull request was opened from the other day13:31
n4nd0git checkout master13:31
n4nd0git fetch upstream13:31
n4nd0git rebase upstream/master13:31
n4nd0git push origin master13:31
n4nd0git checkout <my branch>13:31
n4nd0git rebase master13:31
n4nd0some new commits13:32
n4nd0I couldn't push for the error ^13:32
n4nd0git pull13:32
n4nd0fix conflicts13:32
n4nd0git commit (commit for the merge)13:32
n4nd0git push origin <my branch>13:32
n4nd0what should I have done?13:32
sonne|workthe last git pull should have been a git pull --rebase13:32
n4nd0it was git pull origin <my branch>13:33
n4nd0I should have done then git pull --rebase origin < my branch >13:33
wikingmmm we don't have a structural solver out of box right? :)13:33
sonne|workwiking, after the end of gsoc13:33
wikinghehehe who's doing it?13:33
sonne|workwe will see :)13:33
wikingah ok shit yeah13:34
wikingno decision yet13:34
wikinganyhow i'll put one now in13:34
wikingsome hackstuff based on libqp13:34
sonne|workn4nd0: I don't know how you work but I ususally keep master in sync with origin/master13:34
sonne|workn4nd0: so I would have done git checkout master; git pull --rebase ; git checkout <my branch> ; git rebase master ; ...13:35
n4nd0how would you have solved my situation then?13:36
n4nd0there was a branch13:36
-!- rico [2e1fd566@gateway/web/freenode/ip.] has joined #shogun13:36
n4nd0with a pull request issued13:37
n4nd0where I wanted to make new changes13:37
n4nd0but also to bring in there the changes from master13:37
Marty28kernel_weighted_degree_position_string_modular is from shogun examples13:37
Marty28Data is shogun data13:38
ricohi there13:38
n4nd0Marty28: ok, let me try to make it work :)13:38
sonne|workn4nd0: via git checkout master; git pull --rebase ; git checkout <my branch> ; git rebase master ; git push13:38
Marty28Just a type error, it says.13:38
n4nd0sonne|work: and after that start the new work13:40
n4nd0sonne|work: maybe it was that, I should have pushed changes from master before commiting my own changes13:40
ricoI am working on action recognition using opencv (mostly in c++)13:40
ricoI am looking for code example which shows use of libshogun on opencv data structure.13:40
sonne|workI don't understand - I thought the new work was already committed?13:40
ricomaybe a way to convert SGMatrix to opencv matrix ?13:40
Marty28n4nd0: thanks :)13:41
sonne|workrico: SGMatrix is just a ptr to some values, i.e. double*13:41
n4nd0sonne|work: no, the new work was still to be done13:41
Marinan4nd0: many thanks :)13:41
sonne|workn4nd0: then branch off from master, do the new work and push13:42
n4nd0sonne|work: mmm I think there's a missunderstanding here13:44
n4nd0sonne|work: the branch already existed, I did some work, commit, ..., pull request and fine13:44
n4nd0sonne|work: after a couple of days, I want to continue with my work in that branch13:45
n4nd0sonne|work: but sync it with master too13:45
rico sonne|work: float64_t don't match with opencv's double, is there any issue already reported ?13:46
n4nd0that branch had a pull request open13:46
-!- gsomix [~gsomix@] has joined #shogun13:59
sonne|workn4nd0: yes I understand that is why I am saying that you should do git checkout master; git pull --rebase ; git checkout <my branch> ; git rebase master  to be in sync again14:04
sonne|workrico - it is just a double14:04
n4nd0sonne|work: ok, for the moment I will delete the pull request since it has got quite dirty14:05
gsomixit's seems I will be an optician :)14:13
-!- vikram360 [~vikram360@] has quit [Ping timeout: 276 seconds]14:19
Marty28sonne: Hi! Do you know who takes care of WeightedDegreePositionStringKernel.compute_POIM2 etc. in the shogun team? Is this the most modern method for POIMs?14:34
sonne|workI should but don't have time currently14:43
n4nd0sonne|work: it looks like it is a problem with the typemaps14:44
n4nd0sonne|work: my guess is that to change the type of the argument to an SGVector should make it work14:45
n4nd0sonne|work: what do you think?14:45
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun14:45
sonne|workn4nd0: sounds likely14:45
n4nd0sonne|work: I am working on it14:46
wikinghaha anybody up for extending the HKM?
blackburnhey there14:48
n4nd0blackburn: hey!14:48
blackburnwiking: I told you they had rbf stuff!14:49
wikingthey even have this now:
blackburnwiking: before :D14:49
wikingah  ok14:49
blackburnhmm last one is quite interesting14:50
sonne|workwhat? how do they do it for gaussian kernels?14:53
-!- vikram360 [~vikram360@] has joined #shogun14:55
blackburnsonne|work: magic14:55
sonne|workblack or white?14:56
blackburnsonne|work: just some fourier stuff14:57
blackburnas in HIK and JS14:57
sonne|workso gaussians become polynomials?14:57
blackburnsonne|work: I am not sure what you mean14:58
blackburnusing bochner's theorem they obtain spectrum of kernel usually14:58
blackburnso they just construct features from spectrum14:59
Marinan4nd0 : have you found the "poim - problem?15:24
n4nd0Marina: it looks like there is class that is not up to date15:25
n4nd0Marina: I am working on it15:25
Marinathank you!!!!!15:26
blackburnMarina: are you able to validate the fix?15:28
blackburnwe have not much expertise in that stuff but there are some pitfalls I'd like to avoid15:30
-!- harshit_ [~harshit@] has joined #shogun15:31
-!- harshit_ [~harshit@] has quit [Remote host closed the connection]15:37
n4nd0Marty28: about the parameter poimdegree15:38
n4nd0Marty28: you gave it the value of 5 in your example15:38
n4nd0Marty28: I think that makes the program break a bit later15:39
Marty28I just used some number, AFAIK.15:39
n4nd0aham, all right15:39
n4nd0it looks like it cannot be any number15:39
Marty28use 115:40
n4nd0I mean, it breaks too15:40
Marty28 poim 10 6 wd 10 2 arff testmotif1.arff poims.png dna R15:41
Marty28 poim 10 6 wd 10 2 fasta testmotifpos.fasta testmotifneg.fasta poims.png dna R15:41
Marty28This is an application of POIM functionality.15:41
n4nd0what you tell me the values then?15:41
n4nd0for poimdegree and max_len15:42
n4nd010 and 6?15:42
Marty28There svm_poim15:43
Marty28so it is the 6 in the example15:43
Marty28usage: %s poim C poimdegree wd [kernelparameters] [arff|fasta] inputfiles  poim.png [dna|protein] non(nucleotide|amino)converter15:44
n4nd0what about max_len?15:44
Marty28I guess it is about the motifs.15:44
Marty28Maybe start with 5.15:44
n4nd06 for poimdegree and 5 for max_len fails15:46
Marty28max_len 2?15:46
n4nd06 and 23 made it work15:47
n4nd0yeah ...15:47
n4nd0but examples was?15:47
n4nd0anyway, that is solved now15:48
n4nd0so now you script executes without problems15:48
Marty28I guess 92, but I do not know exactly.15:48
Marty28fine, thanks!15:48
n4nd0blackburn: should we modify the whole class first?15:49
n4nd0all the float64_t* to SG_Vector or SGMatrix?15:50
blackburnn4nd0: yes float64_t* won't work at all15:50
n4nd0blackburn: ok, I will do that and prepare a pull request so you can check it15:51
blackburnn4nd0: I am worried they do some tricks there (like different real and efficient shapes of matrices)15:51
Marty28(if you succeed please also think about WeightedDegreeStringKernel, which has similar functionality. s?ren ought to know.)15:53
sonne|workn4nd0: that is legacy code - the remainders of the double* -> SGVector migration we started changing last gsoc15:59
n4nd0sonne|work: yeah, it looks like that15:59
blackburnlegacy shit16:00
Marty28sonne: So is there a more modern version for WD and POIMs? Or any string kernels which are even better for the purpose?16:02
-!- rico [2e1fd566@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:04
-!- tom__ [2e1fd566@gateway/web/freenode/ip.] has joined #shogun16:06
Marty28Still changing the data type of a simple 1d-array ought to be easy, I suppose16:07
tom__Is there an implementation of substring kernels in libshogun ?16:07
n4nd0Marty28: yeah, it is probably easy, I am a slow guy though :)16:08
Marty28Learning curves are exponential :D16:08
blackburnMarty28: you are always welcome to fix things by yourself :P16:09
Marty28Just a computational biologist. I can invent new applications and promote your stuff, though :)16:12
tom__Is there an implementation of substring kernels in libshogun ? Thanks.16:13
n4nd0tom__: I don't really know if we have what you are looking for16:15
n4nd0tom__: you can take a look here and check if any of these would serve you16:15
n4nd0yeah, better try the second link, documentation may not be updated16:16
n4nd0blackburn: thanks ;)16:16
sonne|workn4nd0: actually no new string kernel for a few years now - I guess we have them all :D16:23
tom__sonne|work: n4nd0: thx, great works btw.16:25
MarinaI get the same float64 error if I used : maxlen = 23 and poimdegree = 6... whats my mistake?16:26
Marty28Marina: I guess it only works for a new version he made.16:26
blackburnMarina: the problem is in old typemaps we have abandoned16:27
blackburnit was (float64_t*, int, int) or so16:27
blackburnand now it should be SGMatrix16:27
blackburnthe only issue with it is a proper shapes handling16:28
n4nd0Marty28, Marina: the thing is that not only the couple of changes that I made to make your script work are necessary to fix all the class16:29
n4nd0and it will take some time to have it ready16:29
Marinaah ok...16:30
Marinabut how coud I used the SGMatrix?16:30
blackburnMarina: you don't have to it is just dev issue16:30
Marty28n4nd0: Ok. Yet I think it will be worth it, the functionality sounds interesting :)16:33
n4nd0I have to go now16:45
n4nd0I will continue to work on this later16:45
n4nd0hopefully it will be ready soon16:45
-!- n4nd0 [] has quit [Quit: leaving]16:46
Marty28thanks :)16:47
-!- tom__ [2e1fd566@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:49
-!- pluskid [~pluskid@] has joined #shogun16:49
pluskidn4nd0, deleting the remote branch won't make your pull request disappear16:51
pluskidas long as you push again the branch to remote16:51
pluskidI have done this multiple times16:52
blackburnhmm nice idea16:52
pluskidn4nd0's solution: git pull and merge, will cause duplicated commits as he said16:53
pluskidso this would be a better solution16:53
pluskidif there's any wiki for shogun, I think we can write something about our git work flow16:54
blackburnpluskid: README.developer?16:54
blackburnthere is some mentioning of git workflow16:55
blackburnplease add16:55
blackburnpluskid: btw if you want to update chinese doc - feel free :) however I bet you won't like it (I wouldn't)16:59
pluskidcoz you don't know Chinese :p17:00
pluskidBTW: why are there English and Chinese doc in the homepage?17:00
pluskidI mean, no other languages?17:00
blackburnpluskid: soeren wanted me to do russian but I was lazy and didn't17:01
blackburnbut actually does it make sense?17:01
-!- flxb [] has joined #shogun17:01
blackburnI see no reason to use any other lang17:01
pluskidwhat make sense?17:01
-!- uricamic [9320543b@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:01
pluskidhmm, English is almost universal17:01
pluskidI'm just curious to see only Chinese there :p17:02
blackburnit is obvious that machine learners can read and write in english17:02
pluskidyes, that's almost true17:02
pluskidand some can read english but might prefer other (native) language when available17:02
blackburnit is a nightmare to support17:03
pluskidbut sync is a headache, when one has multiple translation17:03
pluskidI agree with you on this point17:03
blackburnonce some X language speaker leaves17:03
Marinacan I use an older shogun version to make poims work now?17:03
blackburnyou have nothing to do with that translation17:03
blackburnMarina: what stops you?17:04
-!- harshit_ [~harshit@] has joined #shogun17:04
-!- emrecelikten [~Anubis@] has joined #shogun17:04
pluskidbtw: I'll travel with the family tomorrow and be back the day after tomorrow17:04
pluskidnot sure whether there will be internet access during the trip17:05
blackburnthat's ok and thanks for notifying17:05
pluskidhope I won't miss some important thing17:05
blackburnno most important is 23, April17:05
blackburnpluskid: I guess it would be kind of morning for you when gsocers are announced?17:06
pluskidnot sure about the exact time17:06
blackburn21 UTC afaik17:06
pluskidnot good at counting numbers :p, but most of the time when friends in the US send me emails, I got them in the morning17:08
blackburnhard to believeheh17:08
pluskid21 + 8 % 24 = 5, seems so17:08
pluskidearly in the morning if I'm not counting wrong :p17:08
Marinai get still the error  TypeError: in method 'WeightedDegreePositionStringKernel_prepare_POIM2', argument 2 of type 'float64_t *'17:08
blackburnpluskid: however that could be a great moment and you shall be here :)17:09
blackburnMarina: still?17:09
@sonney2kblackburn, did anyone merge a fix?17:09
blackburnsonney2k: no, fix is not ready yet17:09
blackburnMarina: what do you mean saying still?17:09
pluskidblackburn, I'll try to be there, if necessary :)17:10
@sonney2kthen I don't understand17:10
@sonney2kwhy try again if there is no fix yet?17:10
blackburnsonney2k: no idea17:10
blackburnsonney2k: please let n4ndo know about shapes of matrices there if would you have time17:12
Marinaignore still ... I get the error and don't know what to do17:12
blackburnnon-trivial sometimes17:12
blackburnMarina: do you want all the fixes get merged in a minute? :)17:12
Marina:) sometimes17:14
blackburnMarina: I am not able to fix things now so you have to wait a little this time17:14
Marinaok, thanks!17:15
blackburnthis could be a day though ;)17:15
* sonney2k attempts to do a quick fix17:15
@sonney2k(within the next 2 train stations)17:16
blackburnsonney2k: good luck ;)17:17
blackburnsee you all17:17
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:17
CIA-64shogun: Soeren Sonnenburg master * r8436b0a / (2 files): convert get_POIM2 to user SGVector -
@sonney2kMarina, try again17:19
-!- vikram360 [~vikram360@] has quit [Ping timeout: 245 seconds]17:20
-!- n4nd0 [] has joined #shogun17:22
n4nd0sonney2k: I can do a pull request for the changes I have done so far17:22
n4nd0sonney2k: it might be ok as quick fix17:22
n4nd0right now it compiles correctly and the script Marty28 sent me works well17:23
shogun-buildbotbuild #725 of libshogun is complete: Failure [failed compile]  Build details are at  blamelist: sonne@debian.org17:25
pluskidupdated README.developer for some notes on git17:33
pluskidand opened a new pull request17:34
pluskidbtw, n4nd0, have you got our previous conversation just now? you can safely delete your remote branch without affecting your pull request17:34
n4nd0oh yes pluskid I just read it in the logs :)17:35
n4nd0it is a good plan then17:35
n4nd0so I just have to delete the branch and it later push again?17:35
pluskidI think better delete *the moment* before you would push again17:36
pluskidnot sure what github will show when someone go and look at the pull request, and find there's no such branch :p17:36
pluskidhowever, there's a "force" option for pushing, which could ignore the non-fast-forward error, but I haven't tried this17:37
-!- harshit_ [~harshit@] has quit [Ping timeout: 260 seconds]17:37
n4nd0wiking suggested me once not to force the push17:37
pluskidyes, that's sounds horrible17:38
pluskidI'd rather not try17:38
pluskidthough I guess the underlying thing is the same as deleting and pushing again17:39
Marinasonney2k: I reinstall shogun now, right?17:39
pluskidMarina, maybe have to reinstall from the latest git repo17:40
Marinathanks i do so17:41
-!- n4nd0 [] has quit [Quit: leaving]17:41
MarinaI get the error:   WeightedDegreePositionStringKernel.cpp:1861: error: prototype for ?shogun::SGMatrix<double> shogun::CWeightedDegreePositionStringKernel::get_POIM2()? does not match any in class ?shogun::CWeightedDegreePositionStringKernel17:42
-!- n4nd0 [] has joined #shogun17:43
n4nd0pluskid: I think that my problem is a bit worse though17:44
n4nd0the commit history in my branch screwed17:44
n4nd0I have duplicated commits :S17:44
-!- pluskid [~pluskid@] has quit [Ping timeout: 264 seconds]17:47
-!- pluskid [~pluskid@] has joined #shogun17:47
pluskidn4nd0, so you have merged locally?17:48
pluskidthat means those duplicated commits exists in your local branch17:49
n4nd0pluskid: I think so, since I couldn't push, I did git pull and solved conflicts17:49
pluskidI have a suggestion17:49
pluskidyou might create a new branch17:49
n4nd0and move the files there right?17:49
pluskidand make your changes of the current branch w.r.t. master as a patch17:50
pluskidand apply the patch in the new branch17:50
pluskidpush the new branch to origin17:50
pluskidupdate your pull request (you can edit the commit range, where you can change the branch name)17:50
pluskidthe pro is that you can avoid those duplicated commits17:51
pluskidthe cons is that you "lose" your individual commits, and merge them into one big commit17:51
pluskidI think there's a command git-format-patch17:51
pluskidfor convenient making patches17:51
n4nd0so I should create the new branch17:52
n4nd0and move there the files that I have been working on17:52
pluskidfirst goto the master branch17:53
pluskidthen create a new branch (based on master)17:53
pluskidthen goto jlcovertree branch17:53
pluskidrun git diff master > something.patch17:54
pluskidand then goto the new branch17:54
pluskidapply the patch17:54
n4nd0I didn't know about this patches17:54
n4nd0can I apply the changes directly?17:55
pluskidyes, I guess so17:55
pluskidsomething like17:55
pluskidpatch -p0 < the_path_file, IIRC17:55
n4nd0I will try it17:56
pluskidtry it, your work will be safe in the jlcovertree branch as long as you don't delete it17:56
n4nd0I need a rest but I will try it later17:57
n4nd0pluskid: thank you!17:57
pluskidbtw, I'm looking at the doc of git-format-patch17:58
pluskidwhich might give you a better solution17:58
pluskidwhich retains your individual commits17:58
-!- Marina [8d175a52@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:00
-!- pluskid [~pluskid@] has quit [Read error: Connection reset by peer]18:01
-!- Marty28 [~Marty@] has quit [Quit: ChatZilla [Firefox 11.0/20120310010446]]18:04
-!- vikram360 [~vikram360@] has joined #shogun18:06
-!- pluskid [~pluskid@] has joined #shogun18:08
pluskidgit-format-patch master18:08
pluskidwill give you an individual patch for each of your commits18:08
pluskidselect those good ones and discard duplicated ones18:08
pluskidand then use git-am to apply those patches18:08
pluskidin that order18:09
pluskidI think this seems to be a better solution, that retains your commit history18:09
pluskidn4nd0, see above18:09
pluskidand good night :)18:09
pluskidhope your problem could be resolved18:09
-!- pluskid [~pluskid@] has quit [Client Quit]18:10
-!- Marty28 [~chatzilla@] has joined #shogun18:21
-!- gsomix [~gsomix@] has quit [Quit: ????? ? ?? ??? (xchat 2.4.5 ??? ??????)]18:23
-!- flxb [] has quit [Quit: flxb]18:33
CIA-64shogun: Soeren Sonnenburg master * r2649a77 / (4 files in 2 dirs): SGVector fix for WD-kernel constructor -
-!- vikram360 [~vikram360@] has quit [Ping timeout: 264 seconds]18:45
-!- vikram360 [~vikram360@] has joined #shogun18:46
Marty28sounds good18:46
shogun-buildbotbuild #726 of libshogun is complete: Success [build successful]  Build details are at
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking]18:51
-!- vikram360 [~vikram360@] has quit [Ping timeout: 264 seconds]19:03
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun19:20
-!- vikram360 [~vikram360@] has joined #shogun19:42
@sonney2kn4nd0, pluskid yeah you should really use git format-patch - otherwise you loose all your commit messages19:43
n4nd0sonney2k: ok, I am watching now the video pluskid sent in the ML and I will try to fix it later in a new branch19:44
n4nd0sonney2k: but I have a doubt, if apply these patches in a new branch, the pull request will die right?19:44
n4nd0I mean that I have to do a new one19:46
-!- Marty28 [~chatzilla@] has quit [Ping timeout: 246 seconds]19:50
-!- Marty28 [~chatzilla@] has joined #shogun19:51
@sonney2kn4nd0, yeah you have to do a new one19:56
@sonney2kexcept if you use push --force19:56
-!- cronor [] has joined #shogun19:57
n4nd0sonney2k: I will do a new one then, if that is not a problem for you20:03
-!- cronor [] has quit [Read error: Connection reset by peer]20:18
-!- cronor [] has joined #shogun20:18
-!- cronor_ [] has joined #shogun20:24
-!- cronor [] has quit [Ping timeout: 260 seconds]20:24
-!- cronor_ is now known as cronor20:24
n4nd0sonney2k: about the WD-kernel, are you taking care of that then? I have seen your commits20:30
-!- n4nd0_ [] has joined #shogun20:33
-!- n4nd0_ is now known as fjig20:33
-!- fjig [] has left #shogun []20:33
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has joined #shogun20:40
-!- gsomix [~gsomix@] has joined #shogun20:50
-!- blackburn [5bde8018@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:53
-!- Marty28 [~chatzilla@] has left #shogun []20:56
gsomixI need more blackburn.21:02
-!- harshit_ [~harshit@] has joined #shogun21:14
-!- Marty28 [~Marty@] has joined #shogun21:17
-!- cronor_ [] has joined #shogun21:23
-!- Marty28 [~Marty@] has quit [Quit: ChatZilla [Firefox 11.0/20120310010446]]21:24
-!- cronor [] has quit [Read error: Connection reset by peer]21:24
-!- cronor_ is now known as cronor21:24
-!- n4nd0 [] has quit [Quit: leaving]22:42
-!- blackburn [~qdrgsm@] has joined #shogun23:05
-!- nickon [] has quit [Quit: ( :: NoNameScript 4.22 :: )]23:09
-!- cronor_ [] has joined #shogun23:28
-!- cronor [] has quit [Read error: Connection reset by peer]23:28
-!- cronor_ is now known as cronor23:28
CIA-64shogun: Sergey Lisitsyn master * rb45f523 / src/shogun/multiclass/MulticlassLibLinear.cpp : SG_ADD for MC liblinear -
CIA-64shogun: Sergey Lisitsyn master * r4997875 / src/NEWS : Updated news -
@sonney2kgsomix, for some gay games or what?23:32
blackburnsonney2k: phew23:33
@sonney2kn4nd0, yeah WD kernel stuff should work?23:33
@sonney2kgsomix, how is it going - happily SGVectoring?23:33
blackburnsonney2k: you are talking with sleeping deads probably23:33
blackburnn4ndo is not present at all and gsomix is sleeping afaik23:34
@sonney2kblackburn, the good thing when doing that is that they cannot prove me wrong - I am always right then :)23:35
blackburnsonney2k: yeah you may claim whatever :)23:35
-!- sonney2k is now known as sonne|god23:36
* sonne|god == god23:36
* sonne|god shogun is my mission 23:37
* sonne|god obey!23:37
blackburnsonne|god: I do23:37
@sonne|godblackburn, go review pluskids patch :)23:37
-!- sonne|god is now known as sonne|fallengod23:38
@sonne|fallengodblackburn, or did you do it already23:38
blackburnsonne|fallengod: I'll merge his readme stuff for now23:39
@sonne|fallengodblackburn, I mean the multiclass stuff?23:39
blackburnsonne|fallengod: yes I'll get back to it once I am finished with some shity issue ;)23:40
CIA-64shogun: Chiyuan Zhang master * r86dce0b / src/README.developer : Added some notes of git work flow for developers. -
CIA-64shogun: Sergey Lisitsyn master * r4287096 / src/README.developer : Merge branch 'git-workflow-doc' of git:// -
-!- sonne|fallengod is now known as sonney2k23:43
-!- cronor [] has quit [Read error: Connection reset by peer]23:47
-!- cronor [] has joined #shogun23:47
@sonney2kblackburn, I am looking at his MC stuff now23:47
@sonney2kblackburn, this patch is huuuuge!23:53
blackburnsonney2k: yes exactly23:53
@sonney2kblackburn, problem is (as he pointed out) that we have to paths23:53
@sonney2kone machine -> kernelmachine -> svm23:53
@sonney2kthe other: machine -> multiclassmachine -> mulitclasskernelmachine -> multiclasssvm23:54
@sonney2kso a multiclass svm cannot be an svm any longer23:54
blackburnsonney2k: I prefer second23:54
@sonney2kwe have both!23:54
@sonney2kit is only that multiclasssvm can no longer be an svm23:55
@sonney2kand this breaks a couple of things23:55
blackburnsonney2k: there are a couple of reasons why multiclasssvm is not an svm23:56
@sonney2kblackburn, yeah I am ok with that23:57
blackburnsonney2k: C1/C2 and a lot of other things23:57
@sonney2kblackburn, yeah but for multiclass svm he now basically needs to add all the svm relvant functions23:58
@sonney2klike C23:58
blackburnyes we have to..23:58
@sonney2kit is what he did23:58
blackburnyes I know23:58
@sonney2kI don't understand why he made multiclassvm a friend of svm though?23:58
@sonney2kdo you?23:58
blackburnhmm let me check23:58
--- Log closed Wed Apr 18 00:00:19 2012