Open in new window / Try shogun cloud
--- Log opened Sun Oct 13 00:00:39 2013
-!- van51 [] has joined #shogun01:18
-!- van51 [] has quit [Read error: No route to host]01:31
shogun-buildbotbuild #576 of nightly_default is complete: Failure [failed test]  Build details are at
-!- new_lido_afk [~walid@] has quit [Ping timeout: 245 seconds]07:04
-!- new_lido_afk [~walid@] has joined #shogun07:18
-!- new_lido_afk is now known as new_lido07:25
-!- hushell [] has quit [Quit: WeeChat 0.3.7]08:26
-!- new_lido [~walid@] has quit [Quit: Leaving]09:23
-!- iglesiasg [] has joined #shogun10:36
-!- mode/#shogun [+o iglesiasg] by ChanServ10:36
@iglesiasghey guys, good morning10:36
-!- shogun-notifier- [] has joined #shogun10:44
shogun-notifier-shogun: hushell :develop * 9d942b1 / / (6 files):
shogun-notifier-shogun: FGM notebook update: using subset of training data10:44
shogun-notifier-shogun: Fernando Iglesias :develop * 4117c9d / / (6 files):
shogun-notifier-shogun: Merge pull request #1719 from hushell/develop10:44
shogun-notifier-shogun: FGM notebook update: using subset of training data10:44
-!- van51 [] has joined #shogun10:59
shogun-buildbotbuild #1960 of deb3 - modular_interfaces is complete: Failure [failed test python modular]  Build details are at  blamelist: hushell <>, Fernando Iglesias <>11:22
@sonney2kshogun-buildbot, force build --branch=develop nightly_default11:39
shogun-buildbotbuild #577 forced11:39
shogun-buildbotI'll give a shout when the build finishes11:39
van51sonney2k: I have a couple more ready11:41
van51sonney2k: shall I make a PR for those or all of them together?11:42
-!- van51 [] has quit [Read error: No route to host]11:54
-!- travis-ci [] has joined #shogun11:57
travis-ci[travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []11:57
-!- van51 [] has joined #shogun11:58
-!- votjakovr [] has joined #shogun12:12
shogun-buildbotbuild #577 of nightly_default is complete: Failure [failed test]  Build details are at
@iglesiasgsonney2k, ping?13:10
van51hey iglesiasg13:24
@iglesiasghi van5113:24
van51iglesiasg: do you know anything about where the web demos will be hosted and/or when?13:25
@iglesiasgvan51, no sorry, no idea13:25
van51iglesiasg: there is a data mining course starting here in my uni and I know the TA and I was thinking of asking them to show some algorithms from shogun13:25
@iglesiasgvan51, that is an awesome idea13:26
van51I remember us viewing kmeans for instance from a java applet13:26
@iglesiasgvan51, so what do you think it is the state of the demos?13:26
@iglesiasgvan51, I poked a bit around the ones in the PR you just sent13:26
van51iglesiasg: most of them are pretty good13:27
van51iglesiasg: I don't like the toy generator when there are more than 1 classes13:27
@iglesiasgvan51, I found some glitches though :S13:27
@iglesiasgvan51, that was one of them13:27
van51iglesiasg: what else?13:27
@iglesiasgvan51, in KMeans the circunferences are pretty weird13:28
@iglesiasgvan51, they become way too large13:28
@iglesiasgI just tried a couple of examples13:28
van51iglesiasg: yeah indeed13:28
@iglesiasgand in KMeans also13:28
@iglesiasgif you click with the right button13:28
@iglesiasgit creates points of another class13:28
@iglesiasgthat shouldn't be allowed here I guess13:28
-!- iglesiasg [] has left #shogun ["Leaving"]13:29
-!- iglesiasg [] has joined #shogun13:29
-!- mode/#shogun [+o iglesiasg] by ChanServ13:29
@iglesiasgups.. wrong hotkey and close the chat window heh13:29
van51iglesiasg: so about the radius it comes down to what is calculated in the KMeans class as radius13:29
van51iglesiasg: and the code mentions something about variance and not radius13:29
@iglesiasgvan51, mm I see13:29
van51iglesiasg: so we may have to calculate the max distance from the points and draw that13:29
@iglesiasgvan51, but hold on a second13:30
@iglesiasgvan51, I used that KMeans not so long ago for my thesis13:30
@iglesiasgand the circunferences were fine IIRC13:30
@iglesiasglet me double check13:30
@iglesiasgin case I was using another method13:30
@iglesiasgkMeans indeed13:31
van51iglesiasg: get_radius()?13:31
@iglesiasglet me check the code13:31
@iglesiasgvan51, the figure generated in on page 4 here
van51iglesiasg: hmm I see13:33
van51iglesiasg: I will have another look then13:33
van51iglesiasg: could be I overlooked something13:33
van51iglesiasg: thx for showing me that13:33
@iglesiasgvan51, hehe no problem13:34
@iglesiasgI make publicity of the things I do this way :P13:34
van51iglesiasg: hah true ;)13:34
van51iglesiasg: as for the secondary class, it plays no role idd, butI think the label coloring may good for demonstration purposes13:35
van51may be*13:35
@iglesiasgvan51, but, how are the blue points used in the clustering demo?13:35
van51iglesiasg: they are all treated the same way13:36
@iglesiasgvan51, are they ignored? are they considered to be of the same class as the red ones?13:36
van51iglesiasg: the latter13:37
@iglesiasgvan51, I see13:37
van51iglesiasg: from what I see in the demo code, it parses the labels but never passes them to the kmeans object13:38
@iglesiasgvan51, I see13:38
van51as expected thatis13:38
@iglesiasgvan51, what about the toy data in the classification demos?13:39
@iglesiasgvan51, because the generator only creates examples of one class, right?13:39
van51iglesiasg: yeah13:39
van51iglesiasg: I was thinking of extending it to allow assigning more than 1 labels13:40
van51iglesiasg: but maybe it needs a new toy generator so that the classes are pretty much seperated13:40
@iglesiasgvan51, aham13:41
@iglesiasgvan51, what do you think is best to do then?13:41
@iglesiasgvan51, I don't think it is a good idea to keep the generator that just generates one-class data for classification demos13:41
van51iglesiasg: yeah I agree. I also prefer the second approach13:42
lisitsynhello guyz13:42
van51lisitsyn: hey13:43
van51iglesiasg: I'll finish with the descriptions and start after that on a new generator13:43
lisitsynwho's got time? ;)13:43
van51iglesiasg: if you spot anything else let me know :)13:43
@iglesiasgvan51, ok, that sounds great13:43
van51lisitsyn: me if I can help13:43
-!- shogun-notifier- [] has quit [Quit: transmission timeout]13:44
lisitsynvan51: okay let me check what can you help us with13:44
van51lisitsyn: ah, if it's not urgent then I'll start something else after the demos13:45
lisitsynvan51: argh - I don't know13:48
lisitsynplease check and see if you can resolve any of these13:48
lisitsynthere are a lot of infrastructure issues unfortunately13:48
lisitsynlike cmake etc13:48
van51lisitsyn: not sure how much I can help with cmake, but I will have a look later or tonight13:50
lisitsynvan51: may be this one?13:50
van51lisitsyn: yeah that sound better13:52
-!- iglesiasg [] has quit [Quit: Leaving]14:03
-!- votjakovr [] has quit [Quit: WeeChat 0.4.0]14:52
-!- lisitsyn [] has left #shogun []15:04
van51sonney2k: can you give me some info on the application/arts and application/asp demos?15:15
-!- iglesiasg [] has joined #shogun15:31
-!- mode/#shogun [+o iglesiasg] by ChanServ15:31
-!- van51 [] has quit [Read error: Connection reset by peer]16:03
-!- iglesiasg [] has quit [Read error: Operation timed out]16:55
-!- iglesiasg [] has joined #shogun19:02
-!- lisitsyn [] has joined #shogun19:05
-!- lisitsyn [] has quit [Ping timeout: 240 seconds]19:22
-!- iglesiasg [] has quit [Quit: Ex-Chat]19:27
-!- lisitsyn [] has joined #shogun19:32
-!- shogun-notifier- [] has joined #shogun20:31
shogun-notifier-shogun-demo: van51 :master * c679642 / / (6 files):
shogun-notifier-shogun-demo: Updated description in more examples and fixed a couple errors20:31
shogun-notifier-shogun-demo: Soeren Sonnenburg :master * e09d850 / / (6 files):
shogun-notifier-shogun-demo: Merge pull request #27 from van51/feature/docs20:31
shogun-notifier-shogun-demo: Updated description in more examples and fixed a couple errors20:31
-!- iglesiasg [] has joined #shogun20:59
-!- mode/#shogun [+o iglesiasg] by ChanServ20:59
@iglesiasgsonney2k, hey20:59
@sonney2kiglesiasg, hi21:03
@iglesiasgsonney2k, writing a mail to you about it right now...21:05
@iglesiasgsonney2k, sent!21:05
@sonney2kiglesiasg, well links on the info page are broken for many refs - you could correct them though with proper ones pointing to the refernces page21:11
@iglesiasgsonney2k, ok, but it is not just the links21:11
@iglesiasgI mean21:11
@iglesiasgsonney2k, [1] in info page21:11
@iglesiasgdoes not reference to [1] in the reference pages I think21:12
@iglesiasgsonney2k, do you see what I mean?21:12
@iglesiasgsonney2k, and still, even if they were correct, I'd rather see the references after the description in the info page too21:14
@iglesiasgI don't see the point of having this References page honestly21:14
-!- iglesiasg [] has quit [Quit: Ex-Chat]22:08
-!- shogun-notifier- [] has quit [Quit: transmission timeout]23:31
--- Log closed Mon Oct 14 00:00:40 2013