Open in new window / Try shogun cloud
--- Log opened Fri Jul 11 00:00:10 2014
-!- PirosB3 [] has quit [Read error: Connection reset by peer]00:06
-!- PirosB3 [] has joined #shogun00:06
-!- PirosB3 [] has quit [Quit: PirosB3]00:24
-!- Netsplit *.net <-> *.split quits: shogun-buildbot00:33
-!- Netsplit over, joins: shogun-buildbot00:33
-!- lisitsyn [~qdrgsm@] has quit [Ping timeout: 240 seconds]01:25
shogun-buildbotbuild #744 of nightly_all is complete: Failure [failed compile test]  Build details are at  blamelist: abinashpanda <>, tklein23 <>03:05
shogun-buildbotbuild #848 of nightly_default is complete: Failure [failed compile]  Build details are at  blamelist: abinashpanda <>, tklein23 <>03:07
-!- Netsplit *.net <-> *.split quits: naywhayare, shogun-buildbot, khalednasr, zxtx, chakri03:27
-!- Netsplit *.net <-> *.split quits: @ChanServ, @sonney2k, @wiking, soumyaC, @besser|fresh03:28
-!- Netsplit over, joins: chakri03:53
-!- besser|fresh [quassel@fedora/besser82] has joined #shogun03:55
-!- ServerMode/#shogun [+o besser|fresh] by barjavel.freenode.net03:55
-!- sonney2k [] has joined #shogun03:57
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun03:57
-!- ServerMode/#shogun [+oo sonney2k wiking] by barjavel.freenode.net03:57
-!- shogun-buildbot [] has joined #shogun03:59
-!- zxtx [] has joined #shogun03:59
-!- naywhayare [~ryan@] has joined #shogun04:07
-!- ChanServ [ChanServ@services.] has joined #shogun04:20
-!- ServerMode/#shogun [+o ChanServ] by barjavel.freenode.net04:20
-!- rajul [~rajul@] has joined #shogun07:49
-!- rajul [~rajul@] has quit [Quit: Leaving]08:12
-!- zxtx_ [~zv@] has joined #shogun09:12
-!- zxtx_ [~zv@] has quit [Ping timeout: 245 seconds]09:21
-!- zxtx_ [~zv@] has joined #shogun09:33
-!- zxtx_ [~zv@] has quit [Ping timeout: 240 seconds]10:10
-!- PirosB3 [~pirosb3@] has joined #shogun10:27
-!- HeikoS [] has joined #shogun12:34
-!- mode/#shogun [+o HeikoS] by ChanServ12:34
-!- HeikoS [] has quit [Client Quit]12:36
-!- HeikoS [] has joined #shogun12:39
-!- mode/#shogun [+o HeikoS] by ChanServ12:39
-!- PirosB3 [~pirosb3@] has quit [Quit: PirosB3]12:41
-!- HeikoS1 [] has joined #shogun12:43
-!- HeikoS [] has quit [Ping timeout: 240 seconds]12:45
-!- shogun-notifier- [] has joined #shogun12:46
shogun-notifier-shogun: Saurabh :develop * ab2aa11 / doc/ipython-notebooks/classification/SupportVectorMachines.ipynb:
shogun-notifier-shogun: extend svm nb12:46
shogun-notifier-shogun: Saurabh :develop * a213d63 / doc/ipython-notebooks/classification/SupportVectorMachines.ipynb:
shogun-notifier-shogun: revamp structure, add math12:46
shogun-notifier-shogun: Saurabh :develop * 29d3f46 / doc/ipython-notebooks/classification/SupportVectorMachines.ipynb:
shogun-notifier-shogun: cleanups12:46
shogun-notifier-shogun: Saurabh :develop * 3960adf / doc/ipython-notebooks/classification/SupportVectorMachines.ipynb:
shogun-notifier-shogun: add example12:46
shogun-notifier-shogun: Heiko Strathmann :develop * 9ccd6ca / doc/ipython-notebooks/classification/SupportVectorMachines.ipynb:
shogun-notifier-shogun: Merge pull request #2318 from Saurabh7/svmnb12:46
shogun-notifier-shogun: extend svm nb12:46
shogun-notifier-shogun-demo: Saurabh :master * 5312856 / util/
shogun-notifier-shogun-demo: update dataset importer12:49
shogun-notifier-shogun-demo: Heiko Strathmann :master * 1f25516 / util/
shogun-notifier-shogun-demo: Merge pull request #52 from Saurabh7/master12:49
shogun-notifier-shogun-demo: update dataset importer12:49
-!- HeikoS1 [] has quit [Ping timeout: 240 seconds]12:56
-!- HeikoS [] has joined #shogun13:01
-!- mode/#shogun [+o HeikoS] by ChanServ13:01
-!- PirosB3 [~pirosb3@] has joined #shogun13:02
-!- HeikoS [] has quit [Remote host closed the connection]13:03
-!- HeikoS [] has joined #shogun13:10
-!- mode/#shogun [+o HeikoS] by ChanServ13:10
-!- HeikoS [] has quit [Client Quit]13:14
shogun-buildbotbuild #3084 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Saurabh <>14:19
shogun-buildbotbuild #3085 of deb1 - libshogun is complete: Failure [failed compile test]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh <>14:36
-!- shogun-notifier- [] has quit [Quit: transmission timeout]15:49
-!- pickle27 [~pickle27@] has joined #shogun16:07
-!- zxtx_ [~zv@] has joined #shogun16:26
-!- pickle27 [~pickle27@] has quit [Quit: This computer has gone to sleep]16:26
-!- zxtx_ [~zv@] has quit [Ping timeout: 240 seconds]16:33
-!- pickle27 [~pickle27@] has joined #shogun16:38
-!- PirosB3 [~pirosb3@] has quit [Read error: Connection reset by peer]18:10
-!- pickle27 [~pickle27@] has quit [Quit: This computer has gone to sleep]18:16
-!- PirosB3 [~pirosb3@] has joined #shogun18:17
-!- rajul [~rajul@] has joined #shogun18:30
-!- PirosB3 [~pirosb3@] has quit [Quit: PirosB3]18:35
-!- zxtx_ [] has joined #shogun18:35
-!- rajul [~rajul@] has quit [Remote host closed the connection]18:40
-!- soumyaC [uid15286@gateway/web/] has joined #shogun18:59
-!- PirosB3 [] has joined #shogun19:04
-!- PirosB3 [] has quit [Read error: No route to host]19:12
-!- PirosB3 [~pirosb3@] has joined #shogun19:13
-!- PirosB3 [~pirosb3@] has quit [Ping timeout: 240 seconds]19:19
-!- rajul [~rajul@] has joined #shogun19:43
-!- pickle27 [~pickle27@] has joined #shogun20:27
-!- zxtx_ [] has quit [Ping timeout: 240 seconds]20:30
-!- rajul [~rajul@] has quit [Read error: Connection reset by peer]20:38
-!- khalednasr [~k.nasr92@] has joined #shogun20:49
-!- pickle27_ [~pickle27@] has joined #shogun21:04
-!- pickle27 [~pickle27@] has quit [Ping timeout: 240 seconds]21:04
-!- pickle27_ [~pickle27@] has quit [Read error: Connection reset by peer]21:25
-!- rajul [~rajul@] has joined #shogun21:26
-!- pickle27 [~pickle27@] has joined #shogun21:26
-!- rajul [~rajul@] has quit [Ping timeout: 240 seconds]21:33
-!- rajul [~rajul@] has joined #shogun21:47
-!- lisitsyn [~qdrgsm@] has joined #shogun21:50
lisitsynkhalednasr: hey21:52
-!- rajul [~rajul@] has quit [Ping timeout: 240 seconds]21:52
khalednasrlisitsyn, hey21:52
lisitsynkhalednasr: I am just recovering from my job exit beer and will be soon ready to talk about gpu stuff :D21:53
lisitsynkhalednasr: will you be there still in a half of an hour?21:53
khalednasrlisitsyn, haha, yeah :D21:54
-!- rajul [~rajul@] has joined #shogun22:05
-!- pickle27 [~pickle27@] has quit [Quit: This computer has gone to sleep]22:16
-!- lisitsyn [~qdrgsm@] has quit [Ping timeout: 240 seconds]22:18
-!- lisitsyn [~qdrgsm@] has joined #shogun22:23
lisitsynkhalednasr: ok let me see what guys think22:24
khalednasrlisitsyn, ok22:24
lisitsynkhalednasr: it is not very clear for me whether lambday is ok to merge :D22:26
khalednasrlisitsyn, the PR has to be modified before merging22:27
khalednasrto address the issues we discussed22:27
lisitsynkhalednasr: how do we address 1), 2), 3)&22:28
khalednasrlisitsyn, for 1, we could modify the functions so that the result vector/matrix is an argument to the function22:29
lisitsynso it would be22:29
-!- rajul [~rajul@] has quit [Read error: Connection reset by peer]22:29
khalednasryeah, for column/row wise sums22:29
khalednasrregular sums don't really need that, since they just return a number22:30
lisitsynkhalednasr: is my understanding correct22:30
lisitsyneach time we do this operation22:31
lisitsynwe create some colwise_sum object?22:31
khalednasrlisitsyn, do you mean an instance of the colwise_sum struct?22:32
lisitsynkhalednasr: yes22:32
khalednasrlisitsyn, nope, all the methods in the struct are static, so we don't have to create an instance22:32
lisitsynkhalednasr: ah ok22:32
lisitsynkhalednasr: ok got it, we just solve partial specialization problem this way22:33
khalednasrlisitsyn, yeah22:33
lisitsynkhalednasr: alright so 1) is quite ok22:33
lisitsynwe don't really change anything but add one more parameter22:33
khalednasrlisitsyn, yup22:34
lisitsynkhalednasr: what about 2)?22:34
khalednasrlisitsyn, for 2, we could remove all the function overloads that explicitly take an eigen3 matrix22:35
-!- jiaolong [9e6d1f01@gateway/web/freenode/ip.] has joined #shogun22:35
khalednasrand let eigen3 matrices be handles implicitly using the SGMatrix cast operators22:35
lisitsynah so we have22:36
lisitsynwhich is called there22:36
lisitsynthat sounds cool22:36
lisitsynif we go for AnyKindOfOtherLibrary22:36
lisitsynwe just add it to sgmatrix22:36
khalednasryeah, that'll make our lives easier :)22:38
lisitsynkhalednasr: and 3)?22:38
khalednasrlisitsyn, we could the backend-dependent matrix/vector class we talk about earlier22:39
lisitsynkhalednasr: it looks quite useful22:40
khalednasrso developers who want to allow their algorithms to run on either GPU/CPU would just use that class22:40
lisitsynkhalednasr: do you have any idea what time would it take to address these things?22:40
khalednasrlisitsyn, not much, I'll probably finish them tomorrow22:41
lisitsynkhalednasr: that's cool22:41
lisitsynkhalednasr: what about having something to make neural nets gpu-able?22:42
lisitsynkhalednasr: I am afraid I need some explanation what has to be done22:42
khalednasrlisitsyn, first we need to add ViennaCL backend to the sum functions22:43
khalednasrlisitsyn, and then add more methods that are necessary for neural nets22:43
khalednasrmatrix multiplication, addition, and so on22:43
lisitsynkhalednasr: do I get it right you'd need to add these things for neural layers?22:44
-!- rajul [~rajul@] has joined #shogun22:44
khalednasrlisitsyn, I'm not sure I get what you mean22:44
lisitsynkhalednasr: I mean are specific neural layers the only parts that need to be changed?22:45
khalednasrlisitsyn, unfortunately no22:45
jiaolongHi guys, what is the best way to do vector<vector<T>>  in shougun with SGVector?22:46
lisitsynahh yeah I see we have some linalng in RBMs as well22:46
khalednasrlisitsyn, there're everywhere :D22:46
jiaolongI don't like SGVector<T*>22:46
lisitsynjiaolong: hey, you mean std::vector?22:46
lisitsynuhmm do you really need one?22:46
lisitsynkhalednasr: yeah sure I just mean if you have dot product22:47
jiaolonglisitsyn: I don't know if there is some alternative way in shogun22:47
khalednasrlisitsyn, anything that modifies the main matrices/vectors used in the algorithm will have to be through the linalg library22:47
lisitsynjiaolong: I mean you'd need vector<vector<>> if you have something dynamic22:48
lisitsynis that the case?22:48
jiaolonglisitsyn: yes, I need22:48
jiaolonglisitsyn: yes, it is the case22:48
lisitsynhmm I see22:48
lisitsynjiaolong: do you need to create it once and treat as sgvector later?22:50
lisitsynor is that something that is dynamic all the time?22:50
lisitsynkhalednasr: it looks like gpu stuff is more tough :D22:50
jiaolonglisitsyn: dynamic all the time22:50
lisitsynjiaolong: hmm then sgvector would be very tough to use22:51
jiaolonglisitsyn: yes22:51
lisitsynjiaolong: I don't see anything from shogun to cope with that22:52
lisitsynyou'd either stick with vector<vector<<> or add some class for that :)22:52
jiaolonglisitsyn: OK, thanks, then I will try to think about other data structure  :)22:53
khalednasrjiaolong, lisitsyn maybe CDynamicArray would work?22:53
lisitsynkhalednasr: yeah but it is 1d22:54
lisitsynit is not22:54
khalednasrI mean CDynamicArray<SGVector<T>>22:54
lisitsynjiaolong: you need cdynamicarray22:54
jiaolongkhalednasr: yeah, it can work, but I just need vector <vector<int>>, not ashogun object22:54
lisitsynit is 1d and 2d and 3d22:54
lisitsynbut why?22:55
jiaolongkhalednasr: lisitsyn CDynamicArray<SGVector<T>> looks good22:55
lisitsynkhalednasr: am I right it would be cool to start with layers?22:55
khalednasrlisitsyn, yeah22:56
khalednasrlisitsyn, we'll probably need to add some sort of a SpecialPurpose module to the library, for all the algorithm specific operations22:57
lisitsynyou'd need some opencl kernels to do some stuff, right?22:57
khalednasrlisitsyn, yup22:57
khalednasrlisitsyn, but they'll mostly be simple kernel, since most of the operations are elementwise22:58
lisitsynkhalednasr: ok then just let me know when you're done with PR changes22:58
lisitsynand we can push viennacl stuff a bit more22:59
khalednasrlisitsyn, ok cool22:59
lisitsynI wrote some opencl code so I would be able to help you22:59
lisitsynI'd suggest to gpu-ize one neural net22:59
lisitsynand then I can also help you to rewrite some code with gpu22:59
lisitsynI mean if it is quite mechanical why not to do that in parallel :)23:00
khalednasrlisitsyn, in parallel, pun intended? :D23:01
lisitsynkhalednasr: haha23:01
khalednasrlisitsyn, cool :)23:02
lisitsynkhalednasr: ok gonna sleep now, could you please mail me once you're done?23:02
khalednasrlisitsyn, sure23:02
lisitsynwith pr thing23:02
lisitsynthen see you tomorrow23:02
khalednasrlisitsyn, goodnight23:02
-!- zxtx_ [~zv@] has joined #shogun23:12
-!- rajul [~rajul@] has quit [Ping timeout: 240 seconds]23:28
-!- rajul [~rajul@] has joined #shogun23:42
--- Log closed Sat Jul 12 00:00:11 2014