Open in new window / Try shogun cloud
--- Log opened Sat Feb 22 00:00:09 2014
-!- new_lido [~walid@] has joined #shogun00:05
-!- HeikoS [] has joined #shogun01:05
-!- mode/#shogun [+o HeikoS] by ChanServ01:05
-!- zxtx [~zv@2607:ff38:3:811c:f4fd:b74f:f882:a9e6] has joined #shogun01:34
-!- new_lido [~walid@] has quit [Ping timeout: 252 seconds]01:50
-!- new_lido [~walid@] has joined #shogun02:02
-!- HeikoS [] has quit [Quit: Leaving.]02:18
-!- zxtx [~zv@2607:ff38:3:811c:f4fd:b74f:f882:a9e6] has quit [Ping timeout: 265 seconds]03:03
-!- new_lido [~walid@] has quit [Ping timeout: 252 seconds]03:33
-!- bumblebee [~quassel@] has joined #shogun05:17
-!- bumblebee is now known as Guest4497305:18
-!- Guest44973 [~quassel@] has quit [Remote host closed the connection]05:18
-!- new_lido [~walid@] has joined #shogun06:26
-!- lambday [67157d4f@gateway/web/freenode/ip.] has joined #shogun09:49
-!- iglesiasg [] has joined #shogun09:55
-!- mode/#shogun [+o iglesiasg] by ChanServ09:55
-!- new_lido [~walid@] has quit [Quit: Leaving]10:16
-!- iglesiasg [] has quit [Quit: Leaving]11:52
-!- parijat [671b082a@gateway/web/freenode/ip.] has joined #shogun12:27
-!- parijat [671b082a@gateway/web/freenode/ip.] has quit [Client Quit]12:29
-!- parijat [671b082a@gateway/web/freenode/ip.] has joined #shogun12:57
-!- parijat [671b082a@gateway/web/freenode/ip.] has quit [Client Quit]12:59
-!- iglesiasg [] has joined #shogun13:01
-!- mode/#shogun [+o iglesiasg] by ChanServ13:01
-!- Saurabh7 [~Saurabh7@] has joined #shogun13:36
Saurabh7hi, I was planing to apply for gsoc with shogun... for the fundamental ml project ,i was trying to work on
Saurabh7so is this a good idea? and i wanted to know what changes are expected there?14:18
@iglesiasghi Saurabh714:20
@iglesiasgSaurabh7, thank you for your interest in the project14:20
@iglesiasgSaurabh7, working on that issue is indeed a good idea, as it is mentioned as a GSoC entrance task for the fundamental ML project14:21
@iglesiasgSaurabh7, actually, it is very suitable since none is working yet on the decision trees part14:22
@iglesiasgSaurabh7, have you read the comments made by wiking in the pull request?14:22
Saurabh7iglesiasg, yup i read the comments14:23
Saurabh7it was suggested that shogun/multiclass/tree/TreeMachine.h should be involved14:24
@iglesiasgSaurabh7, and also14:26
@iglesiasg"a classifier and it's implementation should be under src/shogun/ and not part of an example. Examples are strictly for small examples where you 'show' how one can use a given module of shogun"14:26
@iglesiasgSaurabh7, maybe you can have a look at TreeMachine.h and think about the ID3 implementation of the PR can be done following what is in TreeMachine14:27
Saurabh7iglesiasg, yes sure, the PR defines its own nodes and stuff, instead will have to use the treemachine ones14:29
@iglesiasgI see that there is also some use of STL in library code that may be visible from the interfaces14:29
@iglesiasgonce has to fix that14:29
@iglesiasgSaurabh7, just go ahead then if you have the idea clear :)14:29
Saurabh7iglesiasg, ok sure I will start working,will ask again if i get stuck or something ....thanks14:34
@iglesiasgSaurabh7, sure, no problem! Feel free to discuss a little bit more in detail the design once you have thought about it14:37
@iglesiasgSaurabh7, wiking probably had something already in mind so his feedback will be useful14:37
Saurabh7ok, will get back soon then14:42
-!- Saurabh7 [~Saurabh7@] has quit [Quit: Leaving]14:46
-!- snik [] has quit [Ping timeout: 265 seconds]15:26
-!- shogun-toolbox [] has quit [Ping timeout: 265 seconds]15:26
--- Log closed Sat Feb 22 15:26:03 2014
--- Log opened Sat Feb 22 15:26:13 2014
-!- shogun-t1olbox [] has joined #shogun15:26
-!- Irssi: #shogun: Total of 16 nicks [3 ops, 0 halfops, 0 voices, 13 normal]15:26
-!- Irssi: Join to #shogun was synced in 9 secs15:26
-!- sonne|work1 [] has joined #shogun15:29
-!- huska [] has quit [Ping timeout: 272 seconds]15:31
-!- Netsplit *.net <-> *.split quits: snik, Guest71164, shogun-toolbox15:32
-!- Netsplit *.net <-> *.split quits: naywhayare15:35
-!- Netsplit *.net <-> *.split quits: shogun-buildbot, sonne|work15:50
-!- Netsplit over, joins: shogun-buildbot15:52
-!- shogun-buildbot [] has quit [Ping timeout: 252 seconds]16:02
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 248 seconds]16:03
-!- shogun-buildbot [] has joined #shogun16:04
-!- wiking [] has joined #shogun16:09
-!- besser82 [quassel@gateway/shell/bnc4free/session] has joined #shogun16:09
-!- naywhayare [] has joined #shogun16:09
-!- besser82 is now known as Guest3353116:10
-!- Netsplit *.net <-> *.split quits: @sonney2k, adrin16:14
-!- Guest33531 [quassel@gateway/shell/bnc4free/session] has quit [Changing host]16:14
-!- Guest33531 [quassel@gateway/shell/bnc4free/x-qadivtdumrsjwbwz] has joined #shogun16:14
-!- wiking_ [] has joined #shogun16:17
-!- wiking [] has quit [Read error: Connection reset by peer]16:18
-!- Netsplit over, joins: adrin, @sonney2k16:29
-!- shogun-buildbot [] has quit [Ping timeout: 260 seconds]16:29
-!- lambday [67157d4f@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]16:36
-!- Netsplit *.net <-> *.split quits: wiking_16:55
-!- wiking [] has joined #shogun17:09
-!- shogun-buildbot [] has joined #shogun17:11
-!- Netsplit *.net <-> *.split quits: wiking17:19
-!- sonney2k [] has quit [Ping timeout: 272 seconds]17:19
-!- shogun-buildbot [] has quit [Ping timeout: 272 seconds]17:21
-!- Netsplit *.net <-> *.split quits: Guest33531, naywhayare17:24
-!- shogun-buildbot [~shogun-bu@] has joined #shogun17:25
-!- adrin [] has quit [Ping timeout: 272 seconds]17:44
-!- sonney2k [~shogun@] has joined #shogun17:44
-!- shogun-buildbot [~shogun-bu@] has quit [Ping timeout: 272 seconds]17:44
-!- parijat [671b082a@gateway/web/freenode/session] has joined #shogun17:48
parijatiglesiasg: hi'17:51
parijatiglesiasg: i saw the chat logs about id3 implementation with saurabh17:52
parijatiglesiasg: i got a chance t briefly see monica's PR, i think in id3 one needs many children of a node (not just 2)17:53
-!- sonney2k [~shogun@] has quit [Ping timeout: 272 seconds]17:54
-!- parijat [671b082a@gateway/web/freenode/session] has quit [Quit: Page closed]17:54
-!- parijat [671b082a@gateway/web/freenode/session] has joined #shogun17:54
-!- parijat_ [671b082a@gateway/web/freenode/session] has joined #shogun17:56
-!- parijat [671b082a@gateway/web/freenode/session] has quit [Client Quit]17:56
-!- wiking [~wiking@] has joined #shogun18:00
-!- Netsplit *.net <-> *.split quits: wiking18:13
-!- adrin [] has joined #shogun18:16
-!- iglesiasg [] has quit [Quit: Leaving]18:23
-!- shogun-buildbot [~shogun-bu@] has joined #shogun18:26
-!- sonney2k [] has joined #shogun18:28
-!- iglesiasg [] has joined #shogun18:28
-!- mode/#shogun [+o iglesiasg] by ChanServ18:28
-!- shogun-buildbot [~shogun-bu@] has quit [Ping timeout: 272 seconds]18:36
-!- shogun-buildbot [~shogun-bu@] has joined #shogun18:38
-!- Guest33531 [quassel@gateway/shell/bnc4free/x-qadivtdumrsjwbwz] has joined #shogun18:38
-!- naywhayare [] has joined #shogun18:38
-!- parijat_ [671b082a@gateway/web/freenode/session] has quit [Quit: Page closed]18:52
-!- sonney2k [] has quit [Ping timeout: 272 seconds]18:53
-!- Netsplit *.net <-> *.split quits: shogun-buildbot, @iglesiasg18:56
-!- Netsplit over, joins: shogun-buildbot, @iglesiasg19:01
-!- sonney2k [~shogun@] has joined #shogun19:04
-!- Netsplit *.net <-> *.split quits: adrin19:07
-!- Netsplit *.net <-> *.split quits: Guest33531, naywhayare19:25
-!- new_lido [~walid@] has joined #shogun19:58
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun20:02
-!- mode/#shogun [+o wiking] by ChanServ20:03
@iglesiasgwiking, to who?20:15
@iglesiasgwhole freenode?20:16
sonney2kwiking, over?20:33
sonney2kwiking, do you have matlab?20:33
-!- Guest33531 [quassel@gateway/shell/bnc4free/x-qadivtdumrsjwbwz] has joined #shogun20:53
-!- HeikoS [~heiko@] has joined #shogun20:54
-!- naywhayare [] has joined #shogun20:54
-!- adrin [] has joined #shogun20:54
-!- iglesiasg [] has quit [Quit: Leaving]21:28
-!- Netsplit *.net <-> *.split quits: Guest3353121:35
-!- HeikoS1 [~heiko@] has joined #shogun21:35
@wikingsonney2k: well i've managed to install today one but it depends on JVM 1.621:36
@wikingwhich i soooo don't want ot install :)21:36
-!- pickle27 [c0008876@gateway/web/freenode/session] has joined #shogun21:39
pickle27sonney2k: around?21:39
sonney2kpickle27, hey!21:42
sonney2kwiking, you don't need java for matlab... you can call it with -nojvm21:42
pickle27can you send me a small dump of the irc logs from the website?21:43
pickle27I want to make sure that page looks right21:43
pickle27Im aiming to make a PR today then we can start to deploy21:43
-!- Netsplit *.net <-> *.split quits: adrin, naywhayare, HeikoS21:43
sonney2kpickle27, demo first ;)21:43
sonney2kpickle27, I am still searching for the script...21:48
sonney2kpickle27, okido you got mail21:54
-!- Netsplit over, joins: naywhayare, adrin21:54
pickle27awesome thanks!21:54
sonney2kpickle27, does it work?21:56
pickle27I haven't checked yet, got a few other things to fix up first21:57
HeikoS1sonney2k: hi!22:02
HeikoS1sonney2k:  did you see the mail from ryan?22:02
HeikoS1sonney2k: I think I will add a project description on this now22:02
HeikoS1if there would be two people working on this cross-testing stuff, that would be amazing22:02
sonney2kHeikoS1, *overload*22:03
HeikoS1sonney2k: could you send me my login for the webpage?22:03
HeikoS1sonney2k: not good?22:03
HeikoS1we could just attract someone and let him be mentored by ryan :)22:03
HeikoS1sonney2k: how do you like the example project ?22:04
* sonney2k in overload mode22:04
HeikoS1haha ;)22:04
sonney2kHeikoS1, barely read stuff in the ideas list22:04
HeikoS1sonney2k: ok, dont worry then22:04
HeikoS1I will write something up anyways, might be useful to have that description22:05
HeikoS1attract more people :)22:05
HeikoS1just got another co-mentor yesterday btw22:05
sonney2kHeikoS1, excellent22:06
sonney2kHeikoS1, for some reason shogun-python is stuck in debian new...22:06
HeikoS1sonney2k:  stuck means what?22:07
sonney2ksitting there for 2 weeks without anything to happen22:07
HeikoS1sonney2k: what would be a good mentor for this mlkpack collarboration?22:08
HeikoS1sonney2k: mmh, dont know what could cause that ....22:08
sonney2kHeikoS1, well lack of manpower / interest as usual22:08
sonney2kwe need an ftpmaster to review the pkg22:08
HeikoS1sonney2k: I see22:08
HeikoS1what about the guy we met at the summit?22:08
sonney2kHeikoS1, I asked already22:10
sonney2ktrying agin22:10
HeikoS1sonney2k: could you send me a login for the website?22:10
HeikoS1you can ecrypt to me adress22:11
sonney2kHeikoS1, anyone can but sure22:11
sonney2kHeikoS1, well you password is heiko now22:12
sonney2kjust log in and change it immediately22:12
HeikoS1sonney2k: thanks, done22:13
HeikoS1You don't have permission to edit anything.22:13
sonney2kHeikoS1, you are superuser now22:14
HeikoS1sonney2k: thanks, that worked22:15
HeikoS1sonney2k: whats really important next ?22:16
HeikoS1apart from all those bugs and more entrance tasks22:17
-!- Netsplit *.net <-> *.split quits: adrin, naywhayare22:17
sonney2kHeikoS1, for me debian packages22:18
sonney2kHeikoS1, and did somebody put up the task with the packages?22:18
sonney2kHeikoS1, we really need that22:19
sonney2kHeikoS1, and shogun and teaching / demos / notebooks - that needs to be put up too22:19
HeikoS1sonney2k: ah yeah22:19
sonney2kHeikoS1, then also workshop organizing / announcing22:19
sonney2kHeikoS1, and foundation22:19
HeikoS1sonney2k: I would do that, but I dont know anything about packaging22:19
sonney2kHeikoS1, btw is the money there yet?22:20
HeikoS1so no idea what to write apart from blabla intro22:20
sonney2kHeikoS1, well I want to - we need that22:20
sonney2kHeikoS1, ahh no gunnar sent some text around back then22:20
HeikoS1ok, ah yeah22:20
HeikoS1could you go over it once and then send my via email? I can make it final22:20
HeikoS1sonney2k: ah yeah notebooks and teaching/demos .....22:21
sonney2kHeikoS1, I guess me22:21
HeikoS110 projects ;)22:21
sonney2kHeikoS1, what?22:21
HeikoS1quite some stuff is there now22:21
HeikoS1I have a good feeling about this gsoc22:21
HeikoS1lots of very useful things this year22:21
sonney2kHeikoS1, lets hope we get accepted22:21
HeikoS1rather than 10 new algorithms22:21
HeikoS1sonney2k: why shouldnt we?22:21
HeikoS1more stuff than previous years22:22
sonney2kHeikoS1, well there are no guarantees22:22
HeikoS1+workshop + foundation22:22
HeikoS1I know22:22
HeikoS1sonney2k: mentor for packaging? you and viktor?22:26
sonney2kHeikoS1, yeah22:26
sonney2kHeikoS1, btw vojtech sent me an updated description22:26
sonney2kcould you put it live22:26
HeikoS1sonney2k: yeah forward22:26
sonney2kHeikoS1, he had also a couple more (cool) ideas - let me forward his email22:26
HeikoS1sonney2k: cool22:27
HeikoS1its always good to have more, less work next year :)22:27
HeikoS1sonney2k: btw whats this nonsense with realloc changing the pointer?22:27
HeikoS1why cant I cut off some part of a block?22:27
HeikoS1that seems weird to me22:27
sonney2kHeikoS1, not nonsense22:28
sonney2kHeikoS1, makes a lot of sense to me22:28
HeikoS1I mean yeah there are reasons for doing this, but I should also be allowed to say "no" to this in my eyes22:28
sonney2kHeikoS1, well some history about this hack(!)22:29
sonney2kHeikoS1, back then I had sth like a 3GB matrix (around year 2000)22:29
sonney2kand that was barely fitting into memory22:29
sonney2kprocessing time was too slow22:30
HeikoS1yeah I totally see the point of this in shogun22:30
HeikoS1but not in c++22:30
sonney2kand lots of noise in the data22:30
sonney2kso I used a double* data22:30
sonney2kand did modify the data *inplace*22:30
sonney2kand then just set different #features, #vectors22:30
HeikoS1yeah I see22:30
sonney2kkeeping the whole memory block allocated!22:30
-!- new_lido [~walid@] has quit [Ping timeout: 252 seconds]22:31
HeikoS1I told the guy to add an option either like you just said or copy it to a smaller one and free the old22:31
sonney2kHeikoS1, I mean even allocating anything on top was not possible22:31
sonney2kHeikoS1, yes seen that22:31
sonney2kmakes sense22:31
HeikoS1yes, but best thing would be to do in place and then cut off the last part22:31
sonney2kHeikoS1, that is surely not possible22:31
-!- new_lido [~walid@] has joined #shogun22:31
HeikoS1I dont understand why22:32
sonney2krealloc might do that or might not22:32
sonney2kwhy not?22:32
HeikoS1realloc doesnt offer an option to keep the pointer const and just change size22:32
HeikoS1(if it is made smaller)22:32
sonney2kyou cannot assume the memory management will just allow you to create huge holes22:33
-!- zxtx [~zv@2607:ff38:3:811c:f4fd:b74f:f882:a9e6] has joined #shogun22:33
sonney2kHeikoS1, so it might just put the tiny memory you request for alloc someplace where there is just some tiny hole22:33
sonney2kfreeing a huuge block22:33
sonney2kHeikoS1, did you get vojtechs email?22:34
HeikoS1sonney2k: I see that, keeps memory defragged, *but* it would be nice to have an option22:34
HeikoS1but whatever, cannot change that22:34
sonney2kHeikoS1, I think he is right - we should just demo how to use shogun w/ kaggle / common tasks22:34
HeikoS1finally the bug is sorted out ...22:35
sonney2kHeikoS1, I knew that bug but didn't find the time to reply...22:35
HeikoS1sonney2k: dont worry, its all sorted now, the guy is doing a great job22:35
HeikoS1and enjoys it :)22:35
HeikoS1vojtechs idea is cool22:35
HeikoS1if we only had 10  more people ;)22:36
sonney2kyes tell him that he does a great job - he is *really* cool22:36
sonney2kHeikoS1, you say that every year22:36
sonney2kbut in fact we are a lot more now22:36
sonney2kand admittedly only due to gsoc!22:36
sonney2kHeikoS1, btw did you write your success thing with lisitsyn1?22:37
sonney2kHeikoS1, and did the gsoc money arrive already?22:37
HeikoS1sonney2k: not yet, actually we have a date today to start22:37
HeikoS1ah let me check22:37
HeikoS1sonney2k: nope22:38
sonney2kHeikoS1, btw we need to fix the equals method in parameter comparing doubles!22:38
HeikoS1sonney2k: is there an issue or a description?22:39
sonney2kHeikoS1, we do some (wrong!) abs(a-b)<eps22:39
HeikoS1sonney2k: what do you think about asking the new gsoc projects to be in BSD3?22:39
sonney2kto say stuff equals22:39
sonney2kwhich fails in about 1000000 cases22:39
HeikoS1sonney2k: could you point me to something concrete?22:39
sonney2kfor example NaN, inf etc22:40
sonney2kI fixed that in the python tester22:40
sonney2klet me point you to the code22:40
HeikoS1so whats your suggestion on that?22:40
HeikoS1sonney2k: could you open an issue assign it to me and add link to code in there?22:40
HeikoS1because I want to finish other things first22:40
HeikoS1like all these descriptions22:40
sonney2kLook at line 22 - inspired by
sonney2kHeikoS1, then please do so yourself22:41
HeikoS1sonney2k: ok22:41
HeikoS1is that true for other types too?22:42
HeikoS1no only floating point22:42
sonney2kyes only floating point22:44
sonney2kHeikoS1, so float32,float64,floatmax_t22:44
sonney2kHeikoS1, that was the reason we had failing integration tests on osx22:44
HeikoS1sonney2k: I see, well spotted then22:44
sonney2kHeikoS1, took me quite a while to figure that out...22:44
HeikoS1Ill make it an intro task22:44
HeikoS1I guess the comparison code should be in CMath22:45
sonney2kHeikoS1, yeah it is a good one. This function could live in CMath22:45
sonney2kand SGVector / matrix could use it too22:45
HeikoS1they do22:47
HeikoS1sonney2k: equals is based on comparing element by element22:47
HeikoS1ah you mean the vector->euqals method22:47
sonney2kHeikoS1, yes22:47
HeikoS1sonney2k: this one has no accuracy22:48
sonney2kHeikoS1, we should another function that has it then22:49
HeikoS1sonney2k: see #1888 and 188922:50
pickle27sonney2k: its working thanks!22:55
pickle27now to do some tweaking22:55
sonney2kpickle27, show us some screen demo :D22:56
pickle27yeah I'll push it later today22:56
HeikoS1pickle27: whats going on? :)22:58
pickle27hacking away on our website!22:58
pickle27whats up with you?22:58
-!- Guest33531 [quassel@gateway/shell/bnc4free/x-qadivtdumrsjwbwz] has joined #shogun23:06
HeikoS1pickle27: writing gsoc proposals :)23:08
HeikoS1sonney2k: what about some d3 action for ryans project, would be a neat addition23:08
sonney2kHeikoS1, ahh btw I think we can do the automated examples with jinja2 too - the same way like we do the clone etc tests23:08
sonney2kHeikoS1, and the same way pickle27 is doing the  website :)23:08
HeikoS1sonney2k: good idea, could you add that to the project description?23:09
HeikoS1and add references to the code that does things already?23:09
sonney2kHeikoS1, no we dont' have references23:09
sonney2kjust jinja223:09
HeikoS1sonney2k: well just point out that we generate automatic examples already23:10
HeikoS1and that this is a starting point23:10
sonney2knot examples but code for one particular test23:10
-!- Netsplit over, joins: naywhayare, adrin23:10
sonney2kHeikoS1, btw when is the notification deadline?23:10
HeikoS1sonney2k: dont know by heart23:12
HeikoS1sonney2k: add it anyways its a good starting point, the more stuff the better people23:12
sonney2k 24 February at 19:00 UTC.23:13
sonney2kHeikoS1, ^ just perfect start for the stammtisch23:14
HeikoS1yeah :)23:14
HeikoS1ah thats soon23:14
sonney2kHeikoS1, so I guess they are already done with the scoring23:14
HeikoS1yeah I guess23:15
-!- Netsplit *.net <-> *.split quits: Guest3353123:17
sonney2kHeikoS1, we are missing a project that gets r_modular into the best possible shape23:17
sonney2kI mean R is really popular still and shogun could be as good for R as it is for python now23:17
HeikoS1sonney2k: yep23:18
HeikoS1sonney2k: ideas?23:18
HeikoS1I cannot write 10 descriptions ;)23:18
sonney2kno we would need some R enthusiast23:18
sonney2khmmhh maybe thorsten hothorn or one of his gang23:19
HeikoS1someone for whom this is useful23:19
HeikoS1sonney2k: we also want julia interfaces, very upcoming thing that girl23:19
sonney2kHeikoS1, well julia is not swig supported...23:24
HeikoS1sonney2k: I know, and they dont want to be23:25
HeikoS1but it would still be good to think about this somehow23:25
HeikoS1but maybe later23:25
-!- Netsplit *.net <-> *.split quits: adrin, naywhayare23:25
-!- Netsplit over, joins: naywhayare23:26
-!- new_lido [~walid@] has quit [Remote host closed the connection]23:31
-!- zxtx [~zv@2607:ff38:3:811c:f4fd:b74f:f882:a9e6] has quit [Ping timeout: 272 seconds]23:42
-!- adrin [] has joined #shogun23:49
-!- pickle27 [c0008876@gateway/web/freenode/session] has quit [Quit: Page closed]23:50
-!- pickle27 [c0008876@gateway/web/freenode/session] has joined #shogun23:53
--- Log closed Sun Feb 23 00:00:11 2014