Open in new window / Try shogun cloud
--- Log opened Mon Feb 11 00:00:45 2013
-!- FSCV [~FSCV@173.254.212.46] has joined #shogun00:02
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]00:46
-!- FSCV_ [~FSCV@173.254.212.46] has joined #shogun02:05
-!- FSCV [~FSCV@173.254.212.46] has quit [Ping timeout: 240 seconds]02:05
-!- blackburn [~blackburn@85.114.170.181] has quit [Quit: Leaving.]03:33
-!- FSCV_ [~FSCV@173.254.212.46] has quit [Quit: Leaving]04:26
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun07:22
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has quit [Quit: Page closed]07:54
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun07:56
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun08:38
-!- n4nd0_ [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun08:53
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has quit [Ping timeout: 245 seconds]08:54
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]10:36
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun10:52
-!- n4nd0_ [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]11:17
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has joined #shogun13:53
-!- FSCV [~FSCV@187.210.54.165] has joined #shogun14:13
-!- FSCV_ [~FSCV@173.254.212.46] has joined #shogun14:15
-!- FSCV [~FSCV@187.210.54.165] has quit [Ping timeout: 255 seconds]14:18
-!- FSCV [~FSCV@204.45.110.163] has joined #shogun14:31
-!- FSCV_ [~FSCV@173.254.212.46] has quit [Ping timeout: 244 seconds]14:31
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun15:04
-!- Hausas [~admin@unaffiliated/hausas] has left #shogun []15:04
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]15:27
-!- Koen_ [c372e982@gateway/web/freenode/ip.195.114.233.130] has joined #shogun15:30
-!- wiking_ [~wiking@info2k1.hu] has joined #shogun16:03
-!- zxtx_ [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun16:07
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 276 seconds]16:07
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 276 seconds]16:07
-!- n4nd0 [~nando@n158-p227.kthopen.kth.se] has quit [Ping timeout: 264 seconds]16:22
-!- Koen_ [c372e982@gateway/web/freenode/ip.195.114.233.130] has quit [Ping timeout: 245 seconds]16:29
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun16:59
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun17:06
-!- blackburn [~blackburn@188.168.14.94] has joined #shogun17:10
sumitblackburn,hi!have been busy working on some course projects for the past few days,shall tell you more on scopes for deep learning, :) did you go through the code..?i had mailed it on your other email address ,18:16
sumiti ll work on refining that code too!18:16
blackburnsumit: hey, yes, glanced over it18:16
sumitgood :)18:16
blackburnsumit: however if you want to integrate it you'd have to learn shogun structure and github development workflow18:17
sumityou might not have found it upto the mark though :P ,, or simply futile,but i promise i shall work on them all ,,Yes exactly,that was a big drawback,,I should better get well versed woth shogun and github developments :) my apologies for that18:18
blackburnwell nothing to apologize - just try shogun at some point18:19
sumitohk perfectly fine :) I will do the needful :) thanks ...18:20
sumitany news on gsoc 2013..?18:21
blackburnno, I think we have nothing yet18:22
sumitok :) thanks , anyways I shall get back to you soon :)18:23
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Quit: Page closed]18:45
-!- heiko1 [~heiko@nat-166-222.internal.eduroam.ucl.ac.uk] has joined #shogun19:32
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 264 seconds]19:57
@sonney2kblackburn, heiko1 wiking_ - so it is official20:29
@sonney2kgsoc 2013 is there20:29
heiko1yeah!20:29
@sonney2khttps://www.google-melange.com/gsoc/homepage/google/gsoc201320:30
@sonney2k1month till application deadline20:30
heiko1just read20:30
heiko1I will write a description soon20:30
-!- cavalier- [~cavalier-@70.118.110.73] has quit [Ping timeout: 264 seconds]20:30
-!- cavalier- [~cavalier-@70.118.110.73] has joined #shogun20:31
heiko1sonney2k, already thoughts on projects?20:33
@sonney2kblackburn, so put the ideas page up then we can continue20:33
@sonney2kheiko1, sure20:33
heiko1sonney2k, well the question was: which?20:33
@sonney2k1 cleanup project for supporting i/o for all the different features in all standard formats as usual :D20:33
@sonney2k1 project doing examples in $LANG + graphically / interactive via website w/ d3js20:34
heiko1awesome!20:34
heiko1like that20:34
@sonney2k1 project to get rid of the static interfaces mostly and merge them in elwms and move the elwms features to python modular too (being able to run R scripts from python_modular etc)20:35
heiko1btw sonney2k, you once sent me a link where some presentation was made with d3js and something else and all was moving, which one was that?20:35
@sonney2k1 project optimization stuff with vojtech maybe20:35
heiko1like that too, we need R support20:35
@sonney2k1 project hashing features with olivier20:36
heiko1oh, I could need a krylov method iterative linear solver for mine, maybe this could be combined somehow20:36
@sonney2k1 project fast kernel approximations with smola's gang ;)20:36
heiko1that also sounds cool20:36
@sonney2kheiko1, I cannot admin this year though but 1 student should be OK20:37
@sonney2kheiko1, yeah that was d3js20:39
heiko1but I remember it moving around, when I google d3js nothing moves20:40
@sonney2kheiko1, it does20:40
@sonney2kcrazily actually20:40
heiko1so where to find thos20:40
@sonney2kd3js20:40
heiko1ah, I guesss last time my browser wasnt working20:41
heiko1awesome!20:42
heiko1ah not all of the examples move20:42
@sonney2kheiko1, it is a matter of 1 line of js to make things morph in d320:47
@sonney2kbut it would be very cool to have such stuff on the website20:47
heiko1indeed20:47
heiko1especially with interactive examples :)20:47
heiko1gp heatmaps and stuff like that :)20:48
@sonney2kthe only thing that worries me is that we are late organizing this year20:48
@sonney2kit is all very doable still but we need a release this year before gsoc at all cost20:48
@sonney2kthat is within 1 month or we will screw up badly with tons of new changes coming in again20:49
@sonney2kheiko1, btw I didn't like your answer on the mailinglist20:49
heiko1sonney2k, which one?20:49
@sonney2kI am very much in favor of removing C1/C220:49
@sonney2kit is totally useless in practice20:50
heiko1sonney2k, well still doesnt hurt you20:50
@sonney2kthe most general solution would be to use SGMatrix<flaot64_t> C20:50
heiko1my point was more that its not high-priority (we talked about this 2 years ago)20:50
heiko1sonney2k, agreed20:50
@sonney2kheiko1, it hurts - takes way to long to do MS20:50
@sonney2kheiko1, but consider how many people are asking about this20:51
@sonney2kso yes it is needed20:51
heiko1sonney2k, true20:51
@sonney2kmuch more than C1/C220:51
@sonney2kquestion is if we would do SGMatrix C20:51
@sonney2kcan we do MS then?20:51
heiko1so sgmatrix C has a C for every *what*?20:51
@sonney2kor would we have to use some hack say - float C : if C == 0 -> c_matrix is used?20:52
@sonney2kheiko1, no c_matrix is like:20:52
@sonney2k1) 1 element == usual C20:52
@sonney2k2) 2 elements == like C1 / C220:52
@sonney2k3) a (2,n) matrix -> a C for each element20:53
heiko1ah, I see20:53
@sonney2kheiko1, but can we do MS with SGMatrices?20:53
heiko1thinking ...20:53
heiko1I mean, in principal, yes20:54
heiko1but this having multiple possibilities is a bit more difficult I guess20:55
heiko1but wait20:55
heiko1no it should be no problem20:55
heiko1since the SVM handles to case distinctions for the C-matrix having different structure right?20:55
@sonney2kheiko1, I am talking about the standard case mostly - would grid search work for a single element?20:55
heiko1I think yes, it might need a bit of  tuning (since not done yet)20:56
heiko1but it is possible20:56
@sonney2k(like everything :)20:56
heiko1I mean the model selection just iterates through the parameter tree combinations and sets their elements to the registered parameters20:56
heiko1and this works for matrices20:56
heiko1the parameter trees can also contain matrices20:57
heiko1the only thing is that we do not have a set-range for matrices20:57
heiko1but these can be set by hand for now20:57
heiko1quite some time ago, I cannot perfectly remember how this works :)20:58
heiko1sonney2k, I will go home now, still in the office and its definitely time for Feierabend! :)20:59
heiko1see you later20:59
@sonney2kcu20:59
heiko1btw I will have a look at the stuff soon and come up with a more precise "how difficult is that" answer21:00
heiko1I also recently started coding a bit again, so will try to fix/finish things for release, and I just stopped adding features to my stuff, so only testing etc,21:01
heiko1bye21:01
-!- heiko1 [~heiko@nat-166-222.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]21:01
blackburnalright sonney2k21:59
blackburnsonney2k: will make a draft of ideas page22:03
blackburntomorrow22:03
@sonney2kblackburn, ok - just copy the page from last year22:09
blackburnsonney2k: I wanted to fill in some ideas22:09
@sonney2ksure feel free to clean up the old page22:10
blackburnsonney2k: so you want 1) cleanup, i/o 2) examples 3) static stuff 4) vojtech 5) olivier 6) fast kernel22:14
blackburnif we get igor 7) some compressive sensing22:14
blackburnthen 8) heiko's krylov (??, I am not sure it is a summer-long project)22:14
blackburnand 9) I'd like to have someone on dimension reduction22:15
blackburnI am not that happy with 2) and 3) - it is way too difficult for me to define such projects22:16
blackburnin general this could be done meanwhile by most active students and mentors like you and me22:17
blackburnso lets negotiate that later22:17
blackburnsonney2k: I'd appreciate if you give me some information (just brief) who should I contact with (olivier, ...)22:20
@sonney2kblackburn, I will ask them23:00
@sonney2kblackburn, I would mentor 2) or 3) - but I guess 3 only if your roommate does it :D23:02
@sonney2kit is tough stuff23:02
@sonney2kbut 2) is really what we need for promotion one needs python, d3js, django, a little ML and artistics skills to do this tough ...23:05
@sonney2kthough23:05
--- Log closed Tue Feb 12 00:00:45 2013