IRC logs of #shogun for Friday, 2013-06-14

--- Log opened Fri Jun 14 00:00:44 2013
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has joined #shogun00:08
pickle27hey guys00:10
pickle27sorry I've been so mia California turned out busier than expected00:10
pickle27I'll be around more for the next few days before I am back for good on the 18th00:11
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has quit [Ping timeout: 276 seconds]00:31
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has joined #shogun01:54
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has quit [Quit: Leaving]02:01
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 256 seconds]02:15
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has left #shogun ["PART #gsoc-gr :QUIT :Leaving."]02:27
-!- nube1 [~rho@49.244.57.13] has quit [Quit: Leaving.]04:56
-!- nube [~rho@116.90.239.3] has joined #shogun05:51
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has joined #shogun06:42
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun08:20
-!- gsomix_ [~gsomix@95.67.173.36] has joined #shogun08:26
-!- gsomix [~gsomix@88.200.242.102] has quit [Ping timeout: 252 seconds]08:30
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds]10:59
-!- nube [~rho@116.90.239.3] has joined #shogun11:06
-!- HeikoS [~heiko@nat-171-129.internal.eduroam.ucl.ac.uk] has joined #shogun11:51
-!- mode/#shogun [+o HeikoS] by ChanServ11:51
@HeikoSsonney2k: around?12:00
@HeikoSsonne|work:12:00
sonne|work?12:01
lisitsynsonne|work: I stumbled upon a service that provides 8*2Ghz + 16 gb for 53$/month12:02
sonne|workHeikoS: yes?12:03
sonne|worklisitsyn: link?12:03
@HeikoSsonne|work: just wondering when the next gsoc irc meeting is12:03
@HeikoScouldnt find the mail/log12:03
lisitsynsonne|work: http://le.lc/12:03
lisitsyn:D12:03
lisitsynsonne|work: 30 rub = 1 $12:05
sonne|workvirtual server12:05
lisitsynsonne|work: is that bad?12:06
lisitsyndedicated are available too but cost a bit more12:06
@HeikoSlisitsyn: do you know the meeting date we agreed upon last time? DId we set one?12:10
lisitsynHeikoS: uh12:10
@HeikoSjust asking since gsoc starts next monday12:10
lisitsynHeikoS: yes I can't remember12:11
lisitsynHeikoS: we had meeting 31 may12:12
lisitsynor?12:13
@HeikoSdont know12:13
lisitsyn3012:13
lisitsynyes 3012:14
@HeikoS30 may12:14
sonne|workwell check the doodle!12:14
sonne|workiglesiasg send this around12:14
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.]12:14
@HeikoSsonne|work: thanks that was what I was looking for12:16
@wikingmmm lot of comments :)12:17
@wikinglisitsyn: have u had time to fidle with the static inferfaces with cmake?12:18
lisitsynwiking: will fiddle with it tonight12:18
@wikingkk12:19
@wikingthnx!12:19
lisitsynpretending I am working now :D12:19
@wikinggood good12:19
@wikingsomebody has to do that :P12:19
lisitsynwiking: uh I should get some matlab I guess12:19
lisitsynto test it properly12:19
@wikingyes :(12:19
lisitsynwiking: I have some cracked^W matlab around12:19
lisitsynlol12:19
@wikingheheh watch out12:20
@wikingthis is a logged irc channel :P12:20
@wikingbsa might come for u12:20
lisitsynwiking: commies will defend me ;)12:20
@wikinghehehe comrad putyin is now more busy with a girl i guess ;P12:21
lisitsynwiking: no why12:21
lisitsynhe don't have to find a girl12:21
lisitsynthat's time consuming so he has one already12:21
@wiking:D12:21
lisitsynI mean we all knew he is having fun with kabayeva instead of wife wiking12:22
lisitsyn:D12:22
lisitsynwiking: I am just curious will it compile on windows afterwards12:24
@wikingyou need python and shit12:25
lisitsynwiking: it'd definitely nice to get it to work with matlab under windows12:26
lisitsynI guess it is an often configuration12:26
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun12:37
@HeikoSwiking:  how is the unit-test stuff going?12:43
@HeikoSvotjakovr: hi!12:43
votjakovrHeikoS: hi, i'm finally back12:43
@HeikoSvotjakovr: good to see you!12:44
@HeikoShow is it going?12:44
votjakovrHeikoS: good to see you too :)12:45
votjakovrHeikoS: I'm refactoring GPs12:46
@HeikoSvotjakovr: what exactly?12:46
votjakovrHeikoS: inference methods, i'm thinking about good alternative for check_members() method12:49
@HeikoSvotjakovr: you wanted to do that in the constructors/setters right?12:49
@HeikoSsetters can be used from constructors to have the checks only once12:50
votjakovrHeikoS: yep, but we need to check all members in all setters, otherwise we can't call update methods there12:53
@HeikoSvotjakovr: I see, well why is the check_members then a bad thing to have?12:53
@HeikoSit does exactly this12:53
@wikingHeikoS: yeah i had it almost yesterday12:57
votjakovrHeikoS: yep, it's doing that, but user doesn't know in which method generates an exception. Maybe replace check_members() with check_members(char* location) ?12:57
@wikingi'm just a bit comatised... slept about 4 hours12:57
@HeikoSwiking: haha, man dont do this12:58
@wikingHeikoS: wasn't my option :)12:58
@HeikoSvotjakovr: from which methods can check_members fail?12:58
votjakovrHeikoS: yep12:59
@wikingHeikoS: the landlord though that it's a good idea to have a quick chat at 7am :P12:59
@HeikoSwiking: I see ;)12:59
@HeikoSvotjakovr: are there many user-generated problems that can appear12:59
@HeikoSlet me check the method ...12:59
@HeikoSvotjakovr: all these errors are inference method related13:00
@HeikoSso its fine if one doesnt know where they are coming from13:00
@HeikoSmostly NULL checks13:00
@HeikoSvotjakovr: So I dont really see why this should be changed?13:04
votjakovrHeikoS: Ok13:04
votjakovrHeikoS: but i think it's better to replace some SG_ERROR with REQIURE13:05
@HeikoSyep totally, but note that this results in the same code for the c++ compiler13:05
@HeikoSits just a macro that makes code easier to read13:05
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun13:07
shogun-notifier-shogun: Viktor Gal :feature/RandomForest * e4f8384 / src/shogun/machine/BaggingMachine.cpp,src/shogun/machine/BaggingMachine.h: https://github.com/shogun-toolbox/shogun/commit/e4f8384e1cf60871b368229a2bc7b53fb4b68e1013:07
shogun-notifier-shogun: Fix BaggingMachine based on @karlnapf feedback13:07
votjakovrHeikoS: i know.  I'll do that quickly13:09
@HeikoSvotjakovr: cool13:10
@HeikoSvotjakovr: whats next?13:10
votjakovrHeikoS: and solve problem with cholesky in parallel13:11
@HeikoSvotjakovr: remind me what that was?13:11
votjakovrHeikoS: it's comuted twice: when one called constructor and get_cholesky() method13:13
@HeikoSvotjakovr: I see, well that should be a quick one right?13:13
votjakovrHeikoS: yep, just add update_parameter_hash() call13:14
votjakovrHeikoS: Also i'd like to move all parameters from log-space13:14
@HeikoSvotjakovr: I thought about this, I think we should do it in the same way as the GPML toolbox dont you think?13:15
@HeikoSto avoid that people are confused when using both of them13:15
votjakovrHeikoS: yep, it's good, but then we need to move parameters to all parameters to log-space, for example width of gaussian kernel, is it good for other part of shogun? Or is it better to leave things as they are?13:22
@HeikoSvotjakovr: good question13:24
@HeikoSI think its unfeasible to change this13:24
@HeikoStoo much development work, and too many users not used to this13:24
@HeikoSbut in fact, you are right, log space is better (for positive parameters)13:25
@HeikoSvotjakovr: no thats impossible to do now13:25
@HeikoSlets keep them in std space13:25
@HeikoSlisitsyn: comments?13:26
votjakovrHeikoS: all of them?13:26
lisitsynoop13:26
lisitsynI wasn't following13:26
@HeikoSvotjakovr: what is the current state on this again?13:26
@HeikoSlisitsyn: positive parameters are better stored in log-space13:27
lisitsynHeikoS: always?13:27
@HeikoSlisitsyn: positive ones, yes thats good in general13:27
@HeikoSvotjakovr: maybe leave them as they are for now13:27
lisitsynHeikoS: sounds like a big piece of work13:29
@HeikoSlisitsyn: yeah not really possible13:29
@HeikoSalso shogun users might revolt ;)13:29
lisitsynHeikoS: where are these guys13:30
lisitsyn:D13:30
@HeikoSlisitsyn: users? :D13:30
@HeikoShaha13:30
@HeikoSusing scipy13:30
lisitsynahh13:30
lisitsynthey don't yet know13:30
lisitsynthat they are shogun users13:30
lisitsyn:D13:30
@HeikoSvotjakovr: rather lets focus on other things for now: for example pulling the GP class out of regression and draw class diagrams for classification13:34
votjakovrHeikoS: Ok13:38
thoralfDid anyone use shogun for multilabel classification?15:53
thoralfi.e. examples, that belong to a subset of classes {1..n} instead of exactly one class15:54
lisitsynthoralf: I added a class for that iirc15:58
thoralflisitsyn: Didn't see it.  How is it called?15:58
lisitsynthoralf: MulticlassMultipleOutputLabels15:59
thoralfOh, thanks.15:59
lisitsynthoralf: I am not sure it works :D15:59
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has quit [Ping timeout: 245 seconds]15:59
thoralflisitsyn: I'll tell you ;)15:59
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has joined #shogun16:00
lisitsynthoralf: oh yes it works16:00
lisitsynthoralf: outputs top of K16:00
lisitsynthoralf: examples/undocumented/python_modular/classifier_multiclassmultipleoutputliblinear_modular.py16:01
lisitsynthoralf: here we predict top 2 classes16:01
thoralflisitsyn: Having n classes, I have to fix how many (k) classes I want?16:02
lisitsynthoralf: yes16:02
lisitsynin this example16:02
lisitsynn = 316:02
lisitsynk = 216:02
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has left #shogun ["ERC Version 5.3 (IRC client for Emacs)"]16:02
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]16:07
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Quit: Leaving.]17:12
-!- nube [~rho@49.244.25.67] has joined #shogun17:31
-!- lambday [67157e4f@gateway/web/cgi-irc/kiwiirc.com/session] has joined #shogun17:36
lambdayHeikoS: there??17:36
@HeikoSlambday: just a sec, tel17:37
lambdayHeikoS: okie17:37
@HeikoSlambday: ok, done, how is it going?17:43
lambdayHeikoS: hi.. coding.. fixing bugs... leaks... please have a look https://github.com/lambday/shogun/commit/7c6c9ba934d5a77b95c54d8b505a6a7f5bc1536e17:44
lambdayHeikoS: tonight I'll write the normal sampler17:44
lambdayHeikoS: and finish up the CLogDetEstimator by tomorrow probably..17:45
@HeikoSchecking ...17:45
lambdayHeikoS: this code currently doesn't leak/give any invalid read/write errors17:46
lambday3 unit-tests added17:46
lambdaylots of debug msgs :( will remove once the whole is tested17:47
@sonney2khttp://stackoverflow.com/questions/12515866/atomic-compare-exchange-vs-mutex17:48
@sonney2khaha17:48
@sonney2kanswer 2, point 3.17:49
@HeikoSlambday: I will get back to you in a minute ...17:55
lambdayHeikoS: okay17:55
thoralfsonney2k: Haha.17:57
thoralfsonney2k: That's why we're coding. :)17:57
lambdayargh.. more bugs!18:15
@HeikoSlambday: haha, thats the way to go :D18:19
@HeikoSlambday: will start commenting soon, currently a bit busy18:19
lambdayHeikoS: alright..18:20
lambdayHeikoS: you're the man! for suggesting me to start with exact log by the way...18:20
-!- lambday [67157e4f@gateway/web/cgi-irc/kiwiirc.com/session] has quit [Changing host]18:26
-!- lambday [67157e4f@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.126.79] has joined #shogun18:26
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!]18:28
lambdayHeikoS: there?19:12
@HeikoSlambday:  yep19:12
lambdayjust saw your comments19:12
lambdayI'll try to fix those first, before going into the rest of the things19:13
@HeikoSlambday: its a lot of stuff19:13
lambdayHeikoS: yeah :(19:13
lambdaymy head is spinning :(19:13
lambdayspecially when I check the debug msgs19:13
@HeikoSonce its working locally for you, please send PRs of classes one by one (in such way that things dont break)19:13
@HeikoSthen we can go through them seperately19:13
@HeikoSotherwise its more likely to not properly read everything19:13
@HeikoSlambday: you are doing a great job dont worry19:14
lambdayHeikoS: thanks.. one commit per class you mean19:14
@HeikoSlambday: no one class per PR19:14
lambdayokay19:14
@HeikoSstarting with the very basic ones19:14
@HeikoSthen the ones that depend on the first ones19:14
lambdayyeah19:15
lambdaywhen I do, please suggest more and more unit-tests19:15
@HeikoSthats mainly what I  had in  mind19:15
@HeikoSits way easier to see what to test when there is only one class19:16
lambdayyes19:16
lambdaynow I am getting what you were saying about going for the full thing at once19:16
* lambday bows19:16
@HeikoSlambday: haha imagine ;)19:16
lambdaycan't even if I try to :D19:16
@HeikoSits already a lot19:17
@HeikoSconsider sending a few things as PR (the abstract base classes for example)19:17
@HeikoSthen things will get a little smaller19:17
lambdayHeikoS: alright.. I'll make the changes and send those first19:18
lambdayHeikoS: I'll be back after dinner19:19
@HeikoSlambday: alright!19:19
@HeikoSI might be gone then, but I will check things tomorrow19:19
lambdayHeikoS: alright.. :)19:20
@HeikoSlambday: btw I will be away for a few days end of next week19:20
@HeikoScycling trip, I will take my tablet with me so can answer some emails and maybe chat19:20
@HeikoSbut might not be around all the time, rather checking things in the evening (probably this current time every day or so)19:20
lambdaywow man!! :-o I've been thinking of doing that since ages!19:20
@HeikoSlambday: you should! :)19:21
lambdayHeikoS: its alright :) you have a great time :)19:21
@HeikoSstill here until next wednesday, but thanks19:21
lambdayI'll keep doing things as planned :)19:21
lambday:)19:21
@HeikoSenjoy your dinner, see you later19:21
lambdayHeikoS: ciao :)19:21
@sonney2kRe19:44
lisitsynRe19:45
lisitsyn:D19:45
lisitsynsonney2k: we are back at the same time19:45
lisitsynsonney2k: do you know any details on the mentors waiting list?19:48
lisitsynI mean carol said it will be announced closer to the event but may be you have seen some date19:49
lisitsyn(for the summit)19:49
@sonney2klisitsyn, no idea19:50
lisitsynrighto19:51
lisitsynsonney2k: what's with sgstring -> sgvector?19:51
lisitsynI don't know what is the status19:56
@sonney2klisitsyn, it is in everything is b0rken status20:00
lisitsynnice20:00
@sonney2klisitsyn, all the streaming features are totally broken now20:00
lisitsynwere they working before?20:00
@sonney2kand I am at a point where I don't know how to fix it currently20:01
lisitsynthey have been unhealthy since the first gsoc20:01
@sonney2kno20:01
@sonney2kthey were working just fine20:01
lisitsynnot broken but not really in healthy status20:01
@sonney2kuntil we introduced SGVector20:01
@sonney2kproblem is that using just pointers has a benefit when you work with external libraries20:01
@sonney2kso we have a problem with getting this to work with VW now20:02
lisitsynsonney2k: our vw is also a bit outdated20:02
lisitsynI want to cry when I see JL's code20:03
@sonney2klisitsyn, but that doesn't count. You always cry :P20:04
lisitsynsonney2k: no that's not how should code look like20:05
lisitsynno no no20:05
@wikinglisitsyn: come on it's cool he uses v_array instead of std::vector20:09
lisitsynwiking: I can live with that..20:09
lisitsynbut there are a lot of other things20:10
@wikinglike20:10
@wiking?20:10
lisitsyna spaghetti of pointers20:10
lisitsynsome magic numbers20:10
@wikingsum_abs_x += fabsf(elem->x);20:10
@wiking:)20:10
lisitsynre-implemented MAX20:10
@wikingi guess CMath did not exist that time :P20:10
lisitsynwiking: what cmath it is in std..20:11
lisitsynwiking: no that's pure horror - I tried to patch covertree three times20:11
lisitsynI always end up with segfaults I can't debug20:11
lisitsynso even though his code works in his hands it is totally impossible to re-use it20:12
@sonney2kcode is not only about beauty.20:14
lisitsynsonney2k: it is not about beauty it is about reusability20:15
lisitsynif I can't debug it to run under windows it is a very unhealthy sign20:16
lisitsynsonney2k: you know that some researchers avoid using functions because 'they are slow'20:18
@sonney2klisitsyn, real researchers would not even be able to compile shogun nor use C++ code20:19
@sonney2kor write it20:19
lisitsynsonney2k: I don't know what they are doing then20:20
lisitsynwhat can you do without experiments20:20
lisitsynwriting equations is ok but you still have to test it somehow20:20
@sonney2klisitsyn, they write matlab code20:21
@sonney2kspaghetty throw away code20:21
lisitsyna lot of people are using C++ and I have seen pure horror20:21
lisitsynsonney2k: no it is better than when they write C++20:21
lisitsyn    for (int kkk=0;kkk<M;kkk++)20:22
lisitsyn      input[n][kkk]=(complex<double>)line[kkk];20:22
-!- HeikoS [~heiko@nat-171-129.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]20:23
lisitsynthat's two lines from code of one teaching assistant in my university20:23
lisitsynhe probably knows a way in his freaking optics :D20:23
lisitsynbut this "loadsave.cpp" is crazy20:24
-!- nube [~rho@49.244.25.67] has quit [Quit: Leaving.]20:29
-!- iglesiasg [d58f3210@gateway/web/freenode/ip.213.143.50.16] has joined #shogun20:30
-!- mode/#shogun [+o iglesiasg] by ChanServ20:30
@iglesiasgI finally graduated in Spain people!!20:36
@iglesiasgBTW they asked for shogun during the presentation and I made advertisement :)20:36
@iglesiasgnow it is time for LMNN-ing20:37
lisitsyniglesiasg: ohh congrats!20:38
lisitsyniglesiasg: so how to call you now?20:39
lisitsynengineer iglesias? ;)20:39
@iglesiasgfernando please :)20:39
lisitsyniglesiasg: like in cz they are Ing.20:39
@iglesiasglisitsyn: aham! I've never seen that20:39
@iglesiasgjust for Prof. and Dr.20:39
lisitsyniglesiasg: here in russia we have both engineers and two stage thing like in the other world20:40
@iglesiasglisitsyn: two state thing?20:40
@iglesiasgwhat is it?20:40
lisitsynI chose to be a bachelor then masters20:40
@iglesiasgaham got it20:40
lisitsynI mean bachelor and masters20:40
lisitsynso I am kinda bachelor already20:40
lisitsyn:D20:40
lisitsyniglesiasg: when is sweden thing then?20:41
@iglesiasglisitsyn: not sure yet20:42
@iglesiasgbut not within the next few months20:42
lisitsynoh! I see20:42
@iglesiasgI would like to start my phd by october-november20:42
@iglesiasgso just earlier than that20:42
@iglesiasgbut in Sweden I finish it and present at any moment of the year20:42
@iglesiasgit is not like Spain, so that gives more freedom20:43
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has joined #shogun21:00
lisitsynpickle27: hey there21:04
pickle27hey!21:08
@iglesiasggtg now, see you later people21:10
-!- iglesiasg [d58f3210@gateway/web/freenode/ip.213.143.50.16] has quit [Quit: Page closed]21:11
pickle27lisitsyn: how are things?21:14
lisitsynpickle27: missing you ;)21:26
lisitsynhow is your conference trip?21:26
-!- hushell [~hushell@c-24-21-141-32.hsd1.or.comcast.net] has quit [Ping timeout: 246 seconds]21:27
pickle27lisitsyn, good, I'm glad to be done with presentations for now though!21:28
pickle27still not back in my timezone yet. Im visiting family for a bit but spending more time by the computer again so I'll be around21:29
lisitsynI see, nice21:30
lisitsynpickle27: so when you are finally back?21:30
pickle27Im actually back a day later now because I am flying to Montreal for a site interview with Google for next year21:32
pickle27so I am flying back the 18th but I am interviewing all day on the 19th21:32
pickle27then the 20th I'll finally resume my regular life finally21:32
lisitsynpickle27: going to be a googler?21:37
pickle27lisitsyn, perhaps!21:37
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has joined #shogun21:42
lisitsynpickle27: but that's a second interview right?21:42
pickle27yeah this is my site interview so I'll know whether or not they are making an offer after this one21:43
lisitsynpickle27: I see - good luck on that then!21:45
pickle27thanks!21:46
pickle27I'm spending this morning on my thesis but I'll probably starting working out those first test cases soon21:46
lisitsynpickle27: just take your time for now, interview is quite important :)21:47
lisitsynsonney2k: I am quite satisfied with cmaking wiking does21:54
lisitsyneverything is much cleaner now21:55
-!- pickle27 [~kevin@S0106002191dec7e8.cg.shawcable.net] has quit [Quit: Leaving]22:19
-!- hoijui [~hoijui@dslb-092-078-039-106.pools.arcor-ip.net] has joined #shogun22:19
-!- pickle27 [~Kevin@S0106002191dec7e8.cg.shawcable.net] has joined #shogun22:25
-!- hoijui [~hoijui@dslb-092-078-039-106.pools.arcor-ip.net] has quit [Quit: Leaving]22:25
-!- pickle27 [~Kevin@S0106002191dec7e8.cg.shawcable.net] has quit [Client Quit]22:28
-!- pickle27 [~Kevin@S0106002191dec7e8.cg.shawcable.net] has joined #shogun22:28
lisitsynwiking: tough shit!22:29
-!- lambday [67157e4f@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.126.79] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]23:30
-!- pickle27 [~Kevin@S0106002191dec7e8.cg.shawcable.net] has quit [Quit: Leaving]23:33
--- Log closed Sat Jun 15 00:00:46 2013

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!