Open in new window / Try shogun cloud
--- Log opened Mon May 26 00:00:04 2014
-!- khalednasr [~k.nasr92@41.69.198.3] has quit [Quit: Leaving]00:52
-!- khalednasr [~k.nasr92@41.69.198.3] has joined #shogun01:08
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Remote host closed the connection]01:16
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun01:18
-!- khalednasr [~k.nasr92@41.69.198.3] has quit [Quit: Leaving]01:40
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection]06:03
-!- Saurabh7 [~Saurabh7@59.88.32.255] has joined #shogun06:50
-!- Saurabh7 [~Saurabh7@59.88.32.255] has quit [Quit: Leaving]07:37
-!- parijat [75dc2dd5@gateway/web/freenode/ip.117.220.45.213] has joined #shogun07:56
-!- kislay [~Abhijeet@106.221.136.29] has joined #shogun08:01
-!- kislay [~Abhijeet@106.221.136.29] has quit [Quit: Leaving]08:13
-!- parijat [75dc2dd5@gateway/web/freenode/ip.117.220.45.213] has quit [Quit: Page closed]09:21
thoralfHey.09:48
-!- sonney2k_ [~shogun@7nn.de] has quit [Ping timeout: 276 seconds]09:51
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has joined #shogun09:54
-!- soumyaC [uid15286@gateway/web/irccloud.com/x-jwvcsdwkzaikrjjd] has joined #shogun09:56
-!- sonney2k [~shogun@7nn.de] has joined #shogun09:57
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has quit [Quit: PirosB3]10:19
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has joined #shogun10:21
@besser82thoralf, moin!  :D10:47
@besser82thoralf, do you have any news for me?10:48
-!- thijs__ [~quassel@nat-p-08.extern.umcn.nl] has joined #shogun11:22
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun11:38
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has quit [Quit: PirosB3]12:13
-!- khalednasr [~k.nasr92@196.221.66.173] has joined #shogun13:04
-!- khalednasr [~k.nasr92@196.221.66.173] has quit [Client Quit]13:06
-!- khalednasr [~k.nasr92@196.221.66.173] has joined #shogun13:43
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun13:51
shogun-notifier-shogun: Wu Lin :develop * 45da8f1 / src/shogun/machine/gp/ (2 files): https://github.com/shogun-toolbox/shogun/commit/45da8f1e79aad016c12f0e9f1be0df0cefecb00a13:51
shogun-notifier-shogun: the base class13:51
shogun-notifier-shogun: Heiko Strathmann :develop * f103880 / src/shogun/machine/gp/ (2 files): https://github.com/shogun-toolbox/shogun/commit/f1038804a2b8b7b39802b7239cecb8760db304d213:51
shogun-notifier-shogun: Merge pull request #2246 from yorkerlin/variational-likelihoods13:51
shogun-notifier-shogun:13:51
shogun-notifier-shogun: the base class (VariationalLikelihood)13:51
shogun-notifier-shogun: Wu Lin :develop * b7e98db / / (3 files): https://github.com/shogun-toolbox/shogun/commit/b7e98db219d909ee0c61b359ee83e1687a79368e13:54
shogun-notifier-shogun: bug fix13:54
shogun-notifier-shogun: Heiko Strathmann :develop * 7587570 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/758757094c30ae249f5ddc84f3cdc11b4b4203c413:54
shogun-notifier-shogun: Merge pull request #2248 from yorkerlin/lnorm_cdf13:54
shogun-notifier-shogun:13:54
shogun-notifier-shogun: bug fix for normal_cdf and lnormal_cdf13:54
shogun-notifier-shogun: Wu Lin :develop * 382e7fd / / (4 files): https://github.com/shogun-toolbox/shogun/commit/382e7fde54b29e5391a243d85f4d67dbd91142b113:59
shogun-notifier-shogun: bug fix13:59
shogun-notifier-shogun: Heiko Strathmann :develop * a4e7ce2 / / (2 files): https://github.com/shogun-toolbox/shogun/commit/a4e7ce28b8dde2d743759556aefbd5799cba757713:59
shogun-notifier-shogun: Merge pull request #2253 from yorkerlin/probitlikelihood13:59
shogun-notifier-shogun:13:59
shogun-notifier-shogun: bug fix (ProbitLikelihood)13:59
khalednasrlisitsyn, hey, do you have a moment?14:10
lisitsynkhalednasr: sure14:11
khalednasrlisitsyn, I'm thinking about using ViennaCL directly for both CPU and GPU, instead of extending linalg14:11
lisitsynkhalednasr: hmm you mean instead of this thing lambday works on?14:12
khalednasryeah14:12
khalednasrcause it would take too much time to add all the methods I need to the linalg library, and some of them will be too specialized, like: multiply-by-the deriviative-of-the-logistic-function14:12
-!- travis-ci [~travis-ci@ec2-54-227-86-191.compute-1.amazonaws.com] has joined #shogun14:14
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/2604262014:14
-!- travis-ci [~travis-ci@ec2-54-227-86-191.compute-1.amazonaws.com] has left #shogun []14:14
khalednasrlisitsyn, the downside of this is that neuralnets would require viennacl to compile, but it's a header only library so that shouldn't be a big problem14:15
lisitsynkhalednasr: we usually try to provide some fallback implementation14:17
lisitsynthat would be an issue14:17
khalednasrlisitsyn, I could try to add a naive implementation in case viennacl wasn't available. The way I'm doing right now with Eigen314:20
lisitsynkhalednasr: hmm well makes sense for now14:21
lisitsynkhalednasr: could you please catch up with lambday if you see him online14:22
lisitsynso you can discuss it as well14:22
khalednasrlisitsyn, yeah14:22
lisitsynkhalednasr: you could start with naive impls before I think14:23
lisitsyndoes that make sense?14:23
lisitsynthey should be easier to debug etc14:23
khalednasrlisitsyn, we already have one :)14:24
lisitsynyeah cool14:25
khalednasrlisitsyn, the problem is with datatypes, since viennacl needs to use it's own vector/matrix type and not SGVector/SGMatrix14:25
lisitsynyeah14:25
lisitsynthat's always an issue14:25
khalednasrlisitsyn, do you know if Theo is on the mailing list? I wanna send the weekly report14:29
lisitsynkhalednasr: please just CC him14:29
lisitsynthat'd be better14:29
khalednasrok14:30
thijs__Hi, I have an issue with the PCA and kernel PCA14:37
thijs__when I try to compile I get the following:14:37
thijs__/usr/include/shogun/mathematics/lapack.h:45:21: fatal error: clapack.h: No such file or directory14:37
thijs__ #include <clapack.h>14:37
thijs__I am running Kubuntu 14.04 and the Shogun version in the repositories (3.1.1 I think)14:38
thijs__(sorry if this is a stupid question)14:39
khalednasrthijs__, you need to install lapack on your system14:42
thijs__yes this is installed14:42
thijs__liblapack-dev:14:43
thijs__  Installed: 3.5.0-2ubuntu114:43
thijs__  Candidate: 3.5.0-2ubuntu114:43
thijs__  Version table:14:43
thijs__ *** 3.5.0-2ubuntu1 014:43
sonne|worklibatlas-dev: /usr/include/atlas/clapack.h14:43
thijs__        500 http://nl.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages14:43
thijs__        100 /var/lib/dpkg/status14:43
sonne|workthijs__: ^ you need this14:43
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has left #shogun []14:43
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun14:44
thijs__I seem to have this as well and it comes with shogun14:45
thijs__this is an output from apt-cache:14:45
thijs__libatlas-dev:14:45
thijs__  Installed: 3.10.1-414:45
thijs__  Candidate: 3.10.1-414:45
thijs__  Version table:14:45
thijs__ *** 3.10.1-4 014:45
thijs__        500 http://nl.archive.ubuntu.com/ubuntu/ trusty/universe amd64 Packages14:45
thijs__        100 /var/lib/dpkg/status14:45
thijs__should I link against this explicitly while compiling?14:46
sonne|workthijs__: please do dpkg -S clapack.h14:47
thijs__yeah it seems to be there: libatlas-dev: /usr/include/atlas/clapack.h14:48
sonne|workthijs__: well looks like you have to put /usr/include/atlas into INCLUDES14:48
thijs__could you give me a hint on where to do this14:58
thijs__I'm rather new to this, sorry14:58
-!- kislay [~Abhijeet@110.226.241.228] has joined #shogun15:38
-!- travis-ci [~travis-ci@ec2-54-227-86-191.compute-1.amazonaws.com] has joined #shogun15:38
travis-ci[travis-ci] it's lambday's turn to pay the next round of drinks for the massacre he caused in khalednasr/shogun: http://travis-ci.org/khalednasr/shogun/builds/2604809815:38
-!- travis-ci [~travis-ci@ec2-54-227-86-191.compute-1.amazonaws.com] has left #shogun []15:38
-!- travis-ci [~travis-ci@ec2-54-83-250-234.compute-1.amazonaws.com] has joined #shogun15:39
travis-ci[travis-ci] it's lambday's turn to pay the next round of drinks for the massacre he caused in khalednasr/shogun: http://travis-ci.org/khalednasr/shogun/builds/2604809815:39
-!- travis-ci [~travis-ci@ec2-54-83-250-234.compute-1.amazonaws.com] has left #shogun []15:39
-!- kislay [~Abhijeet@110.226.241.228] has quit [Read error: Connection reset by peer]15:39
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun15:59
-!- Saurabh7 [~Saurabh7@117.200.185.239] has joined #shogun16:09
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]16:28
-!- khalednasr [~k.nasr92@196.221.66.173] has quit [Quit: Leaving]16:29
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun16:30
-!- khalednasr [~k.nasr92@196.221.66.173] has joined #shogun16:42
shogun-buildbotbuild #317 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/317  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>16:43
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]16:59
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]17:06
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun17:23
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]17:28
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]17:36
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun17:40
-!- kislay [~Abhijeet@106.221.145.154] has joined #shogun17:42
-!- jiaolong [9e6d09e4@gateway/web/freenode/ip.158.109.9.228] has joined #shogun18:01
-!- thijs__ [~quassel@nat-p-08.extern.umcn.nl] has quit [Remote host closed the connection]18:21
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]18:24
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun18:35
shogun-notifier-shogun: jiaolong :develop * 3085a0f / / (7 files): https://github.com/shogun-toolbox/shogun/commit/3085a0fa2e6993b0f37b67cef8173c2080ae3a4c18:35
shogun-notifier-shogun: Fixed a bug in factor graph ipython notebook18:35
shogun-notifier-shogun: Fernando Iglesias :develop * 67d5555 / / (7 files): https://github.com/shogun-toolbox/shogun/commit/67d55553fa37ca5facc35ca5b0e0d6885e122fb418:35
shogun-notifier-shogun: Merge pull request #2189 from Jiaolong/FGM_notebook18:35
shogun-notifier-shogun:18:35
shogun-notifier-shogun: Fixed a bug in FGM notebook18:35
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun18:35
-!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has joined #shogun18:48
shogun-notifier-shogun: lambday :develop * 7c09faf / / (7 files): https://github.com/shogun-toolbox/shogun/commit/7c09fafcfdc8130f11ea37a190cf5f76ebac229518:57
shogun-notifier-shogun: Updated QuadraticTimeMMD with new formulas18:57
shogun-notifier-shogun:18:57
shogun-notifier-shogun: The previous implementation is also kept alongside with the new18:57
shogun-notifier-shogun: implementation as DEPRECATED versions.18:57
shogun-notifier-shogun:18:57
shogun-notifier-shogun: Heiko Strathmann :develop * eb1f080 / / (7 files): https://github.com/shogun-toolbox/shogun/commit/eb1f080e75db84baf6b1f134d714501b8b8c605618:57
shogun-notifier-shogun: Merge pull request #2256 from lambday/feature/quadratic_time_mmd18:57
shogun-notifier-shogun:18:57
shogun-notifier-shogun: Updated QuadraticTimeMMD with new formulas18:57
-!- HeikoS [~heiko@0542590b.skybroadband.com] has joined #shogun18:59
-!- mode/#shogun [+o HeikoS] by ChanServ18:59
@HeikoSSaurabh7: hi!19:01
@HeikoSSaurabh7: notebook is looking nice!19:01
Saurabh7HeikoS, hey!19:01
@HeikoSSaurabh7: we could also do some things for missing data maybe .....19:01
Saurabh7HeikoS, oh19:02
@HeikoSSaurabh7: like replace values with mean19:02
Saurabh7the pima thng19:02
Saurabh7yes nice idea19:02
@HeikoSWe could also ask parjiat to implement an incomplete svd19:02
@HeikoSthen we can do low-rank fill ins of missing values19:02
@HeikoSSaurabh7: could you suggest him and fernando to do this? with my best regards? Should be easy19:02
@HeikoSHere is some reference19:03
Saurabh7ok19:03
@HeikoSSaurabh7: http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/050512.zip19:03
@HeikoSSaurabh7: Actually, let me file an issue for this19:04
Saurabh7HeikoS, yes , i will have a look at it in meanwhile19:04
@HeikoSSaurabh7: also, dont forget the weekly update :)19:05
@HeikoSyou can send a link of the notebook, to get more feedback!19:05
Saurabh7HeikoS, i was working on the diagram too19:05
Saurabh7HeikoS, ok will do that19:05
@HeikoSSaurabh7: cool, any results to show so far?19:05
Saurabh7HeikoS, but dont know if the unfinished one is good enough19:05
Saurabh7HeikoS, i will put a link in the pr19:05
Saurabh7HeikoS, if it looks okay i will try to make it better19:06
khalednasrlambday, hey19:09
lambdaykhalednasr: hey19:09
khalednasrlambday, do you have a moment?19:09
lambdaykhalednasr: ah brb... on phone19:09
khalednasrlambday, tyt19:09
@HeikoSSaurabh7: https://github.com/shogun-toolbox/shogun/issues/225719:10
@HeikoSSaurabh7: well take your time to finish it19:10
Saurabh7HeikoS, woah you type fast!19:11
@HeikoSSaurabh7: btw, I will have a super busy week this week and next, I might not respond immediately, (deadlines) But please keep on pushing your things, take initiative - you can submit multiple PRs etc19:11
@HeikoSSaurabh7: hehe :) making lots of noise while doing this too! :)19:11
Saurabh7HeikoS, yes i figured it.. i am starting to look for examples for next nb19:11
@HeikoSlambday: hey! pls also dont forget the weekly report :)19:12
@HeikoSlambday: you have done so many things, that is an impressive list, hehe19:12
@HeikoSSaurabh7: cool!19:12
@HeikoSSaurabh7: you can also do multiple notebooks at once, and then update once you got feedback19:12
Saurabh7HeikoS, yes thats exaclty what i thought19:12
@HeikoSSaurabh7: also, you should ask our community for feedback on your notebooks, whenever you feel like on is in a good state, actively ask the mailing list for suggestions etwc19:13
-!- aboutGod [~aboutGod@static-72-66-66-50.washdc.fios.verizon.net] has joined #shogun19:13
Saurabh7HeikoS, sure will do it now19:13
@HeikoSSaurabh7: nice!19:14
-!- aboutGod [~aboutGod@static-72-66-66-50.washdc.fios.verizon.net] has left #shogun []19:23
shogun-buildbotbuild #318 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/318  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>19:24
lambdayre19:36
lambdayHeikoS: hehe.. shall I mention the linalg stuffs there? I was confused.. :D19:36
lambdaykhalednasr: hey man... sorry took a while... mom was on call :(19:36
@HeikoSlambday: sure, you worked on that last week right? :)19:37
lambdayHeikoS: yeah...19:37
khalednasrlambday, no problem :)19:37
lambdayHeikoS: will add vector sum methods this week and on streaming mmd stuffs19:37
lambdaykhalednasr: how's it going? something you wanted to discuss?19:37
khalednasrI want to use the linalg library in the neural net stuff19:37
lambdaykhalednasr: great! as of now we have dot products (works with eigen3 and viennacl) and matrix sum methods (works with eigen3)19:38
lambdaykhalednasr: but the problem is... its a feature branch of shogun and still isn't merged with develop19:38
khalednasrlambday, yeah I wanted to talk to you about that19:38
lambdaykhalednasr: so in case you use these methods, please send your PR to feature/linalg19:39
@HeikoSlambday: you think its worth it?19:39
lambdayHeikoS: vector sums?19:39
@HeikoSlambday: yeah19:39
lambdayHeikoS: doesn't harm, right? I am convinced that its going to be faster19:39
lambdayHeikoS: plus we already have sums for matrices.... why not for vectors19:40
@HeikoSlambday: yeah, in particular for permutation tests without precomputing kernel matrices19:40
@HeikoSyep19:40
@HeikoSagree19:40
khalednasrlambday, the thing is, the linalg branch is behind when it comes to the latest neuralnets code19:40
@HeikoSlambday: whats the state of the feature branch?19:40
lambdaykhalednasr: HeikoS: it works... but I'm not sure whether we should merge this with develop unless we hear from wiking19:41
lambdaykhalednasr: umm... may be its better if this gets merged with develop.. I'm not sure.. that's easier to avoid anomalies19:41
khalednasrlambday, until it gets merged, I'm gonna need to add some methods to linalg19:42
lambdaykhalednasr: alright... which ones do you need?19:42
khalednasrlambday, I could start with viennacl sums19:43
khalednasrthen matrix multiplication19:43
lambdaykhalednasr: we gotta figure out which module that method should go in - because we support module wise settings19:43
lambdaykhalednasr: ah great! viennacl sums for matrices?19:43
khalednasrlambday, yeah19:43
lambdaykhalednasr: cool man!19:43
lambdayso redux is the module19:43
lambdaymatrix multiplication... umm.. may be a new util module? or misc? :/19:44
khalednasrnot sure..19:44
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has joined #shogun19:44
lambdaykhalednasr: the code there is a bit tricky - but super flexible - we can directly pass SGVectors/Eigen3-vectors/viennacl-vectors in the same methods and it handles it all from within19:44
khalednasrit's an important method in a linear algebra library so it probably shouldn't be in a misc. module :)19:45
lambdayhaha... agree!19:45
khalednasryeah the code is great :)19:45
lambdaymay be we can steal some ideas from eigen3.. watching which module do they put matrix multiplication in19:45
lambdaykhalednasr: :)19:46
lambdaykhalednasr: also, please notice how things are done in dot product - its totally possible to pass an eigen3 vector to dot method but internally it uses viennacl19:46
lambdaythings should be similar for sums...19:46
lambdaytruly backend independent is the goal here19:47
khalednasryeah I'll keep that in mind19:47
@HeikoSlambday: we should definitely not wait for too long with this19:47
@HeikoSlambday: about the feedback you got from him, did you adress all the issue he pointed out? with cmake etc?19:48
lambdayHeikoS: yeah.. if khalednasr is on board, this would be finished within one week19:48
@HeikoSlambday: well thats good19:48
@HeikoSlambday: did you hear from yorkerlin?19:48
lambdayHeikoS: yes.. but didn't hear from him afterwards19:48
lambdayHeikoS: no for yorkerlin19:48
-!- vlado862 [~vlado@hgfw-01.soe.ucsc.edu] has joined #shogun19:48
lambdayHeikoS: I asked for besser82's help on cmake stuffs... but.. umm.. may be he forgot :P19:48
lambdaybesser82: there man?19:49
lambdayHeikoS: what do you feel about merging this with develop?19:49
lambdayHeikoS: I mean, which other functionalities would be there before it makes it to develop?19:49
lambdayby next week, sum and dot would be ready I hope19:49
@HeikoSlambday: if we merge, it has to be clear that the build not is not broken and that we dont get stuck with this19:50
@HeikoSbut that seems fine19:50
@HeikoSlambday: so I think we should do soon19:50
@HeikoSwiking: any thoughts?19:50
lambdayHeikoS: hmm... I am worried about cmake stuffs..19:50
@HeikoSlambday: the implemented methods do not need to be super complete (just some factorisations maybe)19:50
@HeikoSlambday: but the cmake stuff (in particular if we add viennacl) needs to be waterproof19:51
@HeikoSlambday: we need another build on the bot then too19:51
lambdayHeikoS: it detects viennacl perfectly on my machine...19:51
@HeikoSlambday: thats good19:51
lambdaybut not sure if its gonna work for other archs19:51
lambdaysay bsd or debian19:51
lambdayI use fedora19:51
@HeikoSlambday: I mean if you keep on putting effort into this, I dont see any problems19:51
@HeikoSlambday: I suggest you get besser82 to review the cmake things19:52
lambdayHeikoS: may be yorkerlin can help a bit there with factorization...19:52
@HeikoSand then we merge, if nobody objects19:52
lambdaycholesky and stuffs19:52
@HeikoSlambday: yeah19:52
lambdayHeikoS: I'll push besser82 a bit about this :)19:53
@HeikoSlambday: cool! :)19:53
lambdaybut factorization is tough to be done within next week on our own :( need yorkerlin on board19:53
@HeikoSlambday: the only reason to have some more functionality is that it will reveal problems19:53
@HeikoSlambday: ok, I will push him a bit to get involved there.19:53
lambdayumm yeah... problems with the framework you mean19:53
@HeikoSlambday: would you be able to guide him a bit on this?19:54
lambdayHeikoS: yeah I'll try my best :)19:54
@HeikoSlambday: yes, like things that dont work, its good to find them before merging :)19:54
lambdaytotally agree19:54
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has quit [Quit: PirosB3]19:54
khalednasrlambday, I'm also gonna need to add random number generator for the gpu19:55
@besser82HeikoS, lambday: I'm on it... but I'm currently a bit over occupied with things for my $money_jobs19:55
lambdaybesser82: no worries - whenever you have some free time19:55
lambdaykhalednasr: umm.. that I don't know anything about - but that should not go into linalg I think... right?19:56
khalednasrlambday, it's gonna generate random matrices. any suggestions on where to put it?19:57
lambdaykhalednasr: so, you mean... something like Eigen::MatrixBase::Random()19:58
lambdaykhalednasr: umm these can do inside linalg::Util module19:58
lambdaybut if you just need a rng in general,.. maths is the place19:59
khalednasryeah but the generation has to be done the GPU19:59
khalednasrto avoid moving tons of data across the PCI bus19:59
@HeikoSbesser82: just leave some basic comments on how things should be done, or give your ok20:00
lambdaykhalednasr: I meant to say, do you think that it can work if you put a rng for GPU in maths and then use that in linalg::Util::generate_random() or something20:00
@HeikoSlambday, khalednasr see wiking 's rng framework!20:00
@besser82HeikoS, will do so during tonite20:00
vlado862Hi there - I am trying to install the Shogun toolbox with python and R modular interfaces on a Centos 6.1 server. I am having trouble getting the modular interfaces to install - I can't seem to find the right development packages. I've been looking at the github and webpage documentation, but I can't find a list of the Centos packages I need. I'd greatly appreciate some tips on how to proceed. Thank you for your help.20:01
lambdaykhalednasr: yes... he added it all last year..20:01
-!- kislay [~Abhijeet@106.221.145.154] has quit [Ping timeout: 240 seconds]20:01
-!- Saurabh7 [~Saurabh7@117.200.185.239] has quit [Remote host closed the connection]20:01
khalednasrlambday, HeikoS yeah I saw it20:01
khalednasrI'm not sure I'll be able to port it to opencl though20:02
lambdayvlado862: which packages are you missing?20:02
vlado862Hi lambday - I am not sure - I just see the cmake output tells me none of the modular interfaces are ON20:04
khalednasrlambday, HeikoS I was thinking of doing something like this: http://cas.ee.ic.ac.uk/people/dt10/research/rngs-gpu-mwc64x.html20:04
@HeikoSkhalednasr: you should definitely talk to wiking about this!20:05
lambdayvlado862: you need to explicitly turn those on... like cmake -DPythonModular=ON -RModular=ON .. inside build dir20:05
@HeikoSkhalednasr: and to sergey !g20:05
lambdaykhalednasr: yep! I don't know much about this stuff20:05
khalednasrsure, I'll talk to them when I start working on it20:06
-!- lisitsyn [~qdrgsm@80.252.20.67] has joined #shogun20:07
vlado862lambday: this is the command I ran cmake -DCMAKE_INSTALL_PREFIX="$home_dir/shogun-3.2.0" -DCmdLineStatic=ON -DPythonModular=ON  -DMatlabStatic=ON -DRStatic=ON  ..20:08
vlado862lambday: the erro I get is R was not built as a library CMake Error at /inside/home/uzunangelov/cmake-2.8.12/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108 (message):   Could NOT find R (missing: R_LIBRARIES R_CXX_FLAGS) Call Stack (most recent call first):20:09
lambdayvlado862: I never installed R myself but be sure that R-dev pkg is intalled20:10
vlado862lambday: OK, great, will do. I actually started looking into that, but I don't have root privileges on our server, and the R-dev package has a bunch of dependencies that I basically need to install manually. I was wondering if there is a way I can save myself the grief20:12
lambdayvlado862: may be you can.. I'm not entirely sure.. may be besser82 and wiking can comment better on this...20:13
-!- lisitsyn [~qdrgsm@80.252.20.67] has quit [Read error: Connection reset by peer]20:15
vlado862lambday: got it - thanks for your help.20:16
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun20:37
-!- mode/#shogun [+o iglesiasg] by ChanServ20:37
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]20:38
@iglesiasgGood evening people20:40
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun20:42
-!- soumyaC [uid15286@gateway/web/irccloud.com/x-jwvcsdwkzaikrjjd] has quit [Quit: Connection closed for inactivity]20:46
-!- vlado862 [~vlado@hgfw-01.soe.ucsc.edu] has quit [Quit: irc2go]20:53
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]20:57
-!- pickle27f [~pickle27@216.191.231.218] has joined #shogun21:05
khalednasrlambday, I'm working on  implementing sum() with ViennaCL, but there might be a problem21:17
lambdaykhalednasr: what problem?21:18
khalednasrthe base sum struct expects the matrix to be like: template<class,Info...>21:18
khalednasrhowever, ViennaCL matrices are defined like: template <class SCALARTYPE, typename F, unsigned int ALIGNMENT>21:18
khalednasrthis F is making it fail to compile21:18
lambdaykhalednasr: what's the F here?21:19
khalednasrviennacl uses to specify row major or column major format21:19
lambdaykhalednasr: hmm.. I see..21:20
-!- HeikoS [~heiko@0542590b.skybroadband.com] has quit [Ping timeout: 264 seconds]21:22
lambdayumm... may be we have to change that a bit... so... class Other1, class Other2, template<class,Other1...,Other2...> class Matrix, class T.... somethign like that21:22
lambdayEigen3 matrix has int... params for their matrix21:22
lambdayFor viennacl vector it was template <class SCALARTYPE, unsigned int SOMETHING> ... now for matrix 2nd arg is problematic21:23
khalednasryeah21:23
lambdayso we gotta change the base to handle two different types of Info... there...21:23
lambdaythen specializations would take care of the particulars21:23
khalednasralright, I'll give it a try and get back to you21:24
lambdaykhalednasr: sounds good :) let me know if I can help with something :)21:24
khalednasrlambday, you're already helping :)21:25
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]21:57
-!- HeikoS [~heiko@0542590b.skybroadband.com] has joined #shogun21:59
-!- mode/#shogun [+o HeikoS] by ChanServ21:59
-!- parijat [75c7d551@gateway/web/freenode/ip.117.199.213.81] has joined #shogun21:59
parijatiglesiasg: hi!21:59
parijatiglesiasg: I'm here finally!! :-)22:00
@iglesiasghey parijat!22:08
@iglesiasggreat :)22:08
@iglesiasgparijat, so I am basically revising the code. I have no good understanding of the theory of the cost-complexity pruning, but I trust you22:09
@iglesiasgparijat, I see your trees are on lsd :P (L302 of CART impl)22:11
parijatiglesiasg: ya, they are indeed22:12
parijatiglesiasg: :p22:12
@iglesiasgparijat, so a couple of things I would like to discuss22:13
parijatiglesiasg: sure please go ahead22:13
@iglesiasgparijat, 1) the unit tests are getting overly large due to the data definition22:13
@iglesiasgparijat, it is probably not a big deal, but there's maybe a solution?22:14
parijatiglesiasg: I agree.. wherever data are same I can use a common method to create data.. but in many cases the data are not exactly same22:15
@iglesiasgparijat, that's a way of going with it, yes22:16
parijatiglesiasg: let me try shotening with that.22:17
@iglesiasgparijat, don't spend much time with it though, I prefer CART to be finished first22:17
@iglesiasgparijat, this is something that could be done afterwards once the more important stuff is done22:18
parijatiglesiasg: ok ya sure! CART is almost done.. I am finishing things today... probably I will send in PR after this current one is merged22:19
shogun-buildbotbuild #319 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/319  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com>, jiaolong <jiaolongxu@gmail.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>22:20
@iglesiasgparijat, I will merge this PR then22:22
@iglesiasgparijat, well, as soon as you have addressed the current comments22:23
parijatiglesiasg: there are a few things in the current PR which I have changed later on to fit the cross validation part...22:24
@iglesiasgparijat, hmm that doesn't sound good22:24
parijatiglesiasg: I can send in a big PR together then and close this..22:25
@iglesiasgparijat, no please -- this PR is already too large22:27
parijatiglesiasg: ya I know :-)22:28
@iglesiasgparijat, the prune alone first and the missing values afterwards would have been better22:28
@iglesiasgparijat, focus on addressing the comments, then I will merge22:28
@iglesiasgremember to write a comment when you update it so I get notified by mail22:28
parijatiglesiasg: ok!22:29
-!- lisitsyn1 [~qdrgsm@80.252.20.67] has joined #shogun22:36
lisitsyn1khalednasr: lambday: so what's up? :)22:39
lisitsyn1what you have decided?22:39
khalednasrlisitsyn1, I'm gonna add some basic methods to linalg and use them in neuralnets22:39
lisitsyn1cool22:39
lisitsyn1khalednasr: lambday convinced you it is good? ;)22:39
khalednasrthings like matrix multiplication and sums and so on..22:39
khalednasrlisitsyn1, I thought about it and  I guess it would be better to use the library22:40
lisitsyn1yeah maybe22:40
khalednasrlisitsyn1, I'm gonna use for the main things only, specialized things like logistic and softmax will be just handled in neuralnets22:41
khalednasrlisitsyn1, what do you think?22:43
lisitsyn1khalednasr: I am good with that so far22:43
lisitsyn1but lets see PRs22:43
lisitsyn1we can refine it later then I guess22:43
khalednasrlisitsyn1, cool. I'm working on adding viennacl sums to linalg right now22:44
-!- PirosB3 [~pirosb3@host80-223-dynamic.246-95-r.retail.telecomitalia.it] has joined #shogun22:58
parijatiglesiasg: in case you are just waiting for my updated PR, please don't do so. I will send in the update quite late since I am in the middle of debugging something... And plan on finishing that before addressing your comments..23:05
-!- parijat [75c7d551@gateway/web/freenode/ip.117.199.213.81] has quit [Quit: Page closed]23:16
@iglesiasgparijat, thanks for letting me know23:36
khalednasrlambday, still there?23:42
-!- pickle27f [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep]23:47
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Quit: Leaving]23:58
--- Log closed Tue May 27 00:00:06 2014