Open in new window / Try shogun cloud
--- Log opened Fri Apr 26 00:00:29 2013
-!- shogun-notifier- [] has quit [Quit: transmission timeout]00:14
-!- Netsplit *.net <-> *.split quits: @HeikoS00:15
-!- Netsplit over, joins: @HeikoS00:15
-!- iglesiasg [] has quit [Quit: Lost terminal]00:17
-!- lisitsyn [~blackburn@] has quit [Ping timeout: 245 seconds]00:27
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Quit: Page closed]00:44
-!- makokal [] has quit [Quit: makokal]00:45
-!- sijin [~smuxi@] has joined #shogun02:22
-!- van51 [] has left #shogun ["PING 1366939949"]03:32
-!- sids_aquarius [] has quit [Remote host closed the connection]03:44
-!- sids_aquarius [] has joined #shogun03:52
-!- Yanglittle [deb20af0@gateway/web/freenode/ip.] has joined #shogun06:07
Yanglittlehey, does any examples output the testdat's posterior probability in python_modular?06:13
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]06:58
-!- van51 [] has joined #shogun08:23
-!- van51 [] has quit [Quit: Leaving.]08:24
-!- iglesiasg [] has joined #shogun08:28
-!- mode/#shogun [+o iglesiasg] by ChanServ08:28
-!- lisitsyn [~blackburn@] has joined #shogun08:32
-!- van51 [] has joined #shogun08:36
-!- makokal [] has joined #shogun08:44
-!- foulwall [] has joined #shogun08:58
foulwallHi sonney2k08:58
foulwallI saw the idea " and integration" is also available. I think I can do some work on that after I finish the demos.09:02
sonne|workfoulwall: I think demos is more work than you expect :)09:06
sonne|workBut sure we could also allow uploads of *small* data sets for the demo09:07
-!- lisitsyn [~blackburn@] has quit [Quit: Leaving.]09:08
foulwallsonne|work: Maybe after the GSoC I'll help that.09:08
foulwallsonne|work: yeah, some demos will need bigger data set09:09
sonne|workfoulwall: cheng and me would be happy if you do...09:10
foulwallsonne|work: I love shogun, so just at you service, meanwhile I can get myself trained in developing them.09:13
-!- foulwall [] has left #shogun []09:19
-!- erlenda [] has joined #shogun09:40
Yanglittlehey, LibSVM can  map the distance-from-margin into probalilities, does shogun do ?09:52
-!- makokal [] has quit [Quit: makokal]09:58
-!- van51 [] has left #shogun ["PING 1366963177"]09:59
-!- lisitsyn [] has joined #shogun10:09
-!- mode/#shogun [+o lisitsyn] by ChanServ10:09
@HeikoSYanglittle, hi, yes we can do this10:51
@HeikoScheck scores_to_probabilities of BinaryLabels10:52
@HeikoSthere is also an example in libshogun10:52
@wikingHeikoS: what bug?10:57
@iglesiasgwiking, HeikoS : hey guys, one thing that I don't quite understand regarding the randomness issue with the tests11:01
@iglesiasgwiking: it fails across platforms/archs because we are relying too much on the seed right?11:02
@wikingnot about seed11:02
@wikingit's about RNG11:02
@iglesiasgwiking: then?11:02
@wikingit's not the same accross platforms11:02
@iglesiasgwiking: ok but say11:02
@wikinghence seeded RNG won't be the same either11:02
@iglesiasgif it is because we are relying on the seed I understand11:03
@iglesiasgotherwise, I don't11:03
@iglesiasgbecause if we are generating numbers from say an interval [0,1]11:03
@HeikoSwiking: ?11:03
@iglesiasgwhy does it make a difference how are they generated as long as they are in that interval?11:03
@wikingiglesiasg: because they are not the same exact number? :)11:04
@HeikoSiglesiasg: we save results to files, then compare against them11:04
@HeikoSto any random stuff will kill us11:04
@iglesiasgwiking: then, we are relying too much on the seed11:04
@HeikoSiglesiasg: some methods are random11:04
@wikingiglesiasg: with seed(x) in the code u suppose that the random will be always the same11:04
@iglesiasgwiking: i.e. we are generating "random numbers" that are not random actually11:04
@HeikoSno way around that11:04
@iglesiasgwiking: ok, I understand11:04
@HeikoSdata doesnt have to be, but other things have to be11:04
@wikingwhich is true until u r using the same system11:04
@HeikoSwiking: what was with the bug?11:04
@wikingHeikoS: nada realised11:04
@iglesiasgwiking: then it doesn't make sense to me to face problems for that11:04
@HeikoSwhat if we have our own random machine in shogun?11:05
@HeikoSwould that solve it?11:05
@iglesiasgit is just that the tests don't work on random data11:05
@HeikoSwhat about python?11:05
@HeikoSiglesiasg: its not only that data11:05
@wikingHeikoS: python uses system rng afaik11:05
@iglesiasgHeikoS: yeah, it is also internally like in CV you said11:05
@HeikoSwiking: so even python is not reproducible?11:05
@wikingHeikoS: and anyhow now we are failing exactly with python_modular ;P11:05
@HeikoSiglesiasg: exactly11:05
@HeikoSwiking: so we need a random machine in shogun that can produce everything11:06
@HeikoSas far as i know we mainly use normal and uniform samples from python11:06
@HeikoSwe have both in shogun11:06
@HeikoSjust need to use them instead of the numpy methods then11:07
@HeikoSor we change our approach11:07
@lisitsynHeikoS: wiking: sonne|work: iglesiasg: please check mail11:07
@HeikoSthat is, only use fixed data11:07
@HeikoSdont test random methods but whether results make sense11:07
@HeikoSI did that a while ago but its also problematic11:07
@HeikoSlike x-val results11:07
@HeikoSyou can define an interval of what you expect11:08
@HeikoSbut no guarantee11:08
@iglesiasglisitsyn: it is looking good :)11:09
@HeikoSlisitsyn: dito11:09
@lisitsynHeikoS: is it bad or good?11:10
@iglesiasgHeikoS: rather than using our random machine11:11
@iglesiasgHeikoS: I would enable the possibility of setting randomly generated values as the deafult option, but in tests we set that from outsite11:11
@iglesiasgthat would be for the parts that use random numbers internally, e.g. CV11:12
@lisitsynwiking: I like your font11:12
sonne|workHeikoS, wiking, iglesiasg - we need a portable random number generator11:12
@wikinglisitsyn: which? :))))11:12
@wikinglisitsyn: i think i've fixed it11:12
@lisitsynwiking: in the issue you just created11:12
@lisitsynwiking: I just read a mail and whoa!11:13
@iglesiasgsonne|work: I think the solution can be a bigger problem than the original one :)11:13
@wikinglisitsyn: hahahah yeah11:13
@lisitsynsonne|work: eigen3 doesn't have great random11:13
@wikinglisitsyn: github took it's work seriously with highlighting :))11:13
@wikingsonne|work: afaik openssl/ssh will have one ;)11:13
@lisitsynsonne|work: C++11 has portable random stuff11:14
@wikingsonne|work: and those ones are for sure portable11:14
@wikingsonne|work: it's now matter of teste11:14
@wikingto depend shogun on openssl11:14
@wikingor just cut out the rng from openssl11:15
@lisitsynare we going to use C++11 btw?11:15
@wikingi'd rather go with simple dependency, as nowadays a system w/o openssl is pretty weird... unless it's not connected to the network :P11:17
@HeikoStrue ssl is easy to depend on11:21
@HeikoShow about c++11?11:21
@HeikoSwhat would that do to us?11:21
@lisitsynHeikoS: I am not sure it is really portable but I just ask if we are going to use new standard11:22
@HeikoSlisitsyn: different question, lets solve the other one first :)11:23
@HeikoSif it *is* portable, we could do it11:23
@HeikoSbut what would it do to us?11:23
@HeikoSwhats needed to use it?11:23
@lisitsynHeikoS: --std=c++0x IIRC11:24
@lisitsynHeikoS: starting from gcc 4.6 or 4.711:24
@HeikoSlisitsyn: have you tried what happens then?11:24
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun11:29
-!- iglesiasg [] has quit [Quit: leaving]11:29
-!- makokal [] has joined #shogun11:30
@lisitsynHeikoS: happens like?11:31
@lisitsynHeikoS: once you use newer standard you can use new headers11:32
@lisitsynlike shared_ptr or these rng things11:32
@lisitsynHeikoS: ah and std collections use move semantics11:33
@HeikoSlisitsyn: I mean, does it compile?11:33
@wikingHeikoS: it does11:33
@lisitsynHeikoS: yes of course, back compatibility is a must11:33
@HeikoSthen why not use it?11:33
@wikingat least with clang yes11:33
@lisitsynHeikoS: it can get into troubles though11:34
@lisitsynHeikoS: as I know shared libraries are not compatible11:34
@HeikoSlisitsyn: I see11:35
@wikinglisitsyn: do you know an easy way to generate a skeleton of a polygon?11:35
@lisitsynwiking: generate?11:35
@lisitsynwiking: like that?11:36
@wikinglisitsyn: yeah similar but my polygon is a simple 2D one11:37
@lisitsynI don't know the algorithm11:38
@wikingyeah i'm trying to find one as well atm...11:41
-!- gsomix [] has joined #shogun11:43
gsomixsonne|work: sonney2k: there?11:43
-!- saket [0e8b5206@gateway/web/freenode/ip.] has joined #shogun11:48
@wikinglisitsyn: anybody is working on RandomizedPCa and fast k-Means?11:51
@lisitsynwiking: no, probably no11:51
@lisitsynwiking: I wanted to but swamped11:51
@wikingi took random forest11:52
@lisitsynwiking: want to participate in gsoc? ;)11:52
@wikinglisitsyn: heheh can i ?11:52
@lisitsynwiking: are you eligible at all?11:53
@wikinglisitsyn: i've realised that yes :)11:53
-!- saket [0e8b5206@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:54
@wikingbtw: for CRF we could use
@wikingit's bsd license11:54
* wiking just looking through the development issues on github11:55
@wikingbtw isn't #943 fixed?11:57
@wikingjson serialization ?11:57
-!- iglesiasg [] has joined #shogun12:24
-!- mode/#shogun [+o iglesiasg] by ChanServ12:24
-!- gsomix [] has quit [Read error: Connection reset by peer]12:56
@wikingbtw opencv has RNG as well
-!- gsomix [] has joined #shogun13:04
-!- pranjaldaga [~pranjalda@] has joined #shogun13:13
-!- gsomix [] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]13:18
-!- gsomix [] has joined #shogun13:21
-!- sids_aquarius [] has quit [Ping timeout: 264 seconds]13:23
-!- pranjaldaga [~pranjalda@] has quit [Ping timeout: 260 seconds]14:02
-!- iglesiasg [] has quit [Quit: leaving]15:13
-!- iglesiasg [] has joined #shogun15:30
-!- mode/#shogun [+o iglesiasg] by ChanServ15:30
-!- kabujinuo [ca288b6b@gateway/web/freenode/ip.] has joined #shogun15:39
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]15:47
-!- erlenda [] has quit [Quit: Leaving]16:00
-!- iglesiasg [] has quit [Quit: leaving]16:06
-!- sonne|work [] has quit [Quit: Leaving.]16:28
-!- lisitsyn [] has quit [Quit: Leaving.]16:28
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun16:48
-!- kaushalb09 [] has joined #shogun16:57
-!- kaushalb09 [] has quit [Client Quit]16:58
-!- gsomix [] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]17:07
-!- makokal [] has quit [Quit: makokal]17:20
-!- pranjaldaga [~pranjalda@] has joined #shogun17:25
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]17:38
-!- kabujinuo [ca288b6b@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:48
-!- lisitsyn [~blackburn@] has joined #shogun17:51
-!- Yanglittle [deb20af0@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]17:51
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun17:57
-!- gsomix [~Miranda@] has joined #shogun18:01
-!- lisitsyn [~blackburn@] has quit [Quit: Leaving.]18:13
-!- makokal [~bokal@] has joined #shogun18:39
-!- mikhailBelous [~towelenee@] has joined #shogun18:44
-!- iglesiasg [] has joined #shogun18:54
-!- mode/#shogun [+o iglesiasg] by ChanServ18:54
-!- sids_aquarius [] has joined #shogun19:15
-!- makokal [~bokal@] has quit [Quit: makokal]19:19
-!- van51 [] has joined #shogun19:24
-!- iglesiasg [] has quit [Quit: Lost terminal]19:37
-!- iglesiasg [] has joined #shogun19:39
-!- mode/#shogun [+o iglesiasg] by ChanServ19:39
-!- pranjaldaga [~pranjalda@] has quit [Ping timeout: 260 seconds]19:51
-!- pranjaldaga [~pranjalda@] has joined #shogun19:52
-!- naywhaya1e is now known as naywhayare20:14
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Quit: Page closed]20:20
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has joined #shogun20:20
-!- van51 [] has left #shogun ["QUIT :Leaving."]20:28
gsomixsonney2k: are you there?20:32
-!- lisitsyn [] has joined #shogun21:12
-!- mode/#shogun [+o lisitsyn] by ChanServ21:12
-!- pranjaldaga [~pranjalda@] has quit [Ping timeout: 258 seconds]21:13
-!- gsomix [~Miranda@] has quit [Read error: Connection reset by peer]21:53
-!- vgorbati [d4029f22@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]22:07
-!- gsomix [~Miranda@] has joined #shogun22:09
-!- poojits [75d35d4a@gateway/web/freenode/ip.] has joined #shogun22:26
poojitsI am interested in "Interface and develop the general quadratic solver library libqp" project for GSOC, but there is no preliminary task for this in the issue list on github. Should I get in touch with mentor Vojtech Franc directly?22:28
@lisitsynpoojits: yes22:29 is his university email address. Is this okay?22:30
@lisitsynpoojits: yes totally22:30
poojitsthanks alot.22:31
-!- gsomix [~Miranda@] has quit [Read error: Connection reset by peer]23:15
-!- gsomix [~Miranda@] has joined #shogun23:26
-!- HeikoS [] has left #shogun []23:46
-!- iglesiasg [] has quit [Quit: leaving]23:49
-!- mikhailBelous [~towelenee@] has quit [Ping timeout: 264 seconds]23:55
--- Log closed Sat Apr 27 00:00:29 2013