IRC logs of #shogun for Wednesday, 2014-11-19

« Tuesday, 2014-11-18 Index Thursday, 2014-11-20 »
--- Log opened Wed Nov 19 00:00:15 2014
-!- ChanServ [ChanServ@services.] has quit [shutting down]00:12
-!- ChanServ [ChanServ@services.] has joined #shogun00:15
-!- ServerMode/#shogun [+o ChanServ] by orwell.freenode.net00:15
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection]00:18
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has joined #shogun01:11
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has quit [Quit: PirosB3]01:16
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has joined #shogun01:25
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun01:30
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection]01:30
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Ping timeout: 255 seconds]01:34
-!- Phawx [~Phawx@cpe-68-203-5-71.austin.res.rr.com] has joined #shogun03:33
-!- Phawx [~Phawx@cpe-68-203-5-71.austin.res.rr.com] has quit [Ping timeout: 250 seconds]03:40
-!- DSrupt [~DSrupt@73.6.109.86] has joined #shogun04:53
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has joined #shogun05:25
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection]07:34
-!- DSrupt [~DSrupt@73.6.109.86] has quit [Quit: (null)]08:27
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has joined #shogun10:01
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has quit [Quit: PirosB3]10:32
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has joined #shogun10:35
-!- HeikoS [~heiko@host-92-25-65-248.as13285.net] has joined #shogun12:17
-!- mode/#shogun [+o HeikoS] by ChanServ12:17
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun12:21
shogun-notifier-shogun: Wu Lin :develop * 9de1677 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/9de16777f38b2da942c0c331127742831fe5a04e12:21
shogun-notifier-shogun: update the KL notebook12:21
shogun-notifier-shogun: Heiko Strathmann :develop * 95787bd / / (4 files): https://github.com/shogun-toolbox/shogun/commit/95787bdd499a9904d9b90d377f4c3d42ed79f94e12:21
shogun-notifier-shogun: Merge pull request #2613 from yorkerlin/develop12:21
shogun-notifier-shogun:12:21
shogun-notifier-shogun: update the KL notebook12:21
-!- travis-ci [~travis-ci@ec2-54-92-156-116.compute-1.amazonaws.com] has joined #shogun12:22
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/4146876312:22
-!- travis-ci [~travis-ci@ec2-54-92-156-116.compute-1.amazonaws.com] has left #shogun []12:22
-!- HeikoS1 [~heiko@host-92-25-65-248.as13285.net] has joined #shogun12:24
-!- HeikoS [~heiko@host-92-25-65-248.as13285.net] has quit [Quit: Leaving.]12:24
@lisitsynHeikoS1: hey12:29
HeikoS1lisitsyn:  heya12:29
@lisitsynHeikoS1:can you please guide me to some work on release? :D12:29
@lisitsynI feel absolutely no control of what is going so I have no clue12:30
HeikoS1lisitsyn: there are not many things missing12:30
HeikoS1wiking's swig issue12:30
HeikoS1wiking's osx issue12:30
@lisitsynHeikoS1: I am like on sick leave for a few days more12:30
HeikoS1but what you could do12:30
HeikoS1is some doc fixes in doxygen12:30
@lisitsynHeikoS1: warnings?12:31
HeikoS1its a bit messy atm, broken links12:31
@lisitsynlike doc warnnings?12:31
HeikoS1and link our md files properly12:31
HeikoS1I opened an issue a while ago12:31
HeikoS1thats good for release12:31
@lisitsynok let me do that then12:31
@lisitsynHeikoS1: I will write a script for d-ptrs then12:31
HeikoS1http://www.shogun-toolbox.org/doc/en/latest/12:31
@lisitsynafter that12:31
HeikoS1lisitsyn: please focus on release, its important :)12:31
HeikoS1some links are broken12:32
@lisitsynHeikoS1: yes12:32
HeikoS1dptrs prototype also good12:32
HeikoS1lisitsyn: there are also tons of clean ups to do12:32
HeikoS1check my latest opned issues for the doc stuff12:32
HeikoS1thats actually important for release12:32
@lisitsynok let me check then12:32
HeikoS1like get rid of many doxygen things and replace with md text files that are editable from wiki12:32
HeikoS1clean up etc12:32
HeikoS1https://github.com/shogun-toolbox/shogun/issues/260212:33
HeikoS1https://github.com/shogun-toolbox/shogun/issues/259912:33
HeikoS1lisitsyn: update the install guide12:34
HeikoS1so that doxygen contains the same as md files12:34
HeikoS1actually we dont want anything in doxygen files, it all should come from md12:34
HeikoS1so that its easier to maintain12:34
@lisitsynoh lets stop at this point :D12:34
HeikoS1lisitsyn: oh, and: push wiking ;)12:34
@lisitsynsounds like I am loaded :)12:34
HeikoS1basically the tutorial on doxygen sould also be migrated to md files /wiki12:35
HeikoS1you need to be creative :D12:35
HeikoS1ok ill stop12:35
@lisitsynHeikoS1: not a matter of creativity :)12:36
@lisitsynHeikoS1: I tend to think I am creative sometimes ;)12:36
HeikoS1lisitsyn: what I mean: you know there are lots of bits and pieces to fix everywhere so just pick among those and come up with cleaner solutions12:36
@lisitsynthe thing is that I don't have enough focus12:37
HeikoS1lisitsyn: i actually think you are super creative!12:37
HeikoS1well then, important thing is the doc.12:37
HeikoS1lisitsyn: oh12:37
HeikoS1make notebooks work12:37
HeikoS1also good12:37
HeikoS1and a release blocker12:37
@lisitsynhehe okay I've got what are in priority12:37
HeikoS1cool12:37
HeikoS1"great" that you are sick ;) hehe12:37
HeikoS1and get well soon :)12:38
@lisitsynHeikoS1: I've got that thing causing crazy cough :D12:38
@lisitsynHeikoS1: ah if you're not busy I wanted to share some thought12:40
HeikoS1im busy, but share anyway  !12:41
@lisitsynHeikoS1: hehe okay12:42
@lisitsynHeikoS1: so the thing I started to notice people claim neural network applied to some dataset X is essentially learning some known representation (like gabor filters in images or matrix factorization in word2vec)12:44
HeikoS1ok?12:44
@lisitsynI am curious whether it could be moving to the point neural nets are just discovering some representations for us12:44
@lisitsynand then we still build domain models on top of it12:44
@lisitsynbut it would be still engineered based on the thing it learn12:45
@lisitsynHeikoS1: like in the end we still do something similar to bayesian (first engineer our priors adjusted to what top neural models do)12:46
shogun-buildbotbuild #888 of precise - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/888  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>12:46
HeikoS1I see12:47
HeikoS1well its kind of orthogonal idea12:47
HeikoS1once you learned a representation12:47
HeikoS1you have it12:47
HeikoS1so why model it12:47
HeikoS1why would one build a model of it?12:47
HeikoS1lisitsyn: careful with bayesian, once you have seen the data once, you cannot use it again12:49
HeikoS1otherwise you get inconsistetn12:49
@lisitsynHeikoS1: I mean12:49
@lisitsynHeikoS1: neural nets tend to learn some representations like gabor filters12:49
@lisitsynbut it is just approximate12:49
@lisitsyncan't we just use gabor filters then whatever it learns12:49
shogun-buildbotbuild #915 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/915  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>12:50
HeikoS1I guess the idea is that you dont specify this beforehand12:50
HeikoS1and it turns out that some models are good, so the nn recovers it12:50
@lisitsynHeikoS1: yes and then we can jump back to the models it tries to imitate12:51
@lisitsynHeikoS1: I see some mission in killing neural nets :D what's why I am bothered12:52
HeikoS1but why?12:52
HeikoS1its good12:52
@lisitsynI don't like its nearly infinite capacity and that it replaces everything12:52
@lisitsynI actually like what VC says about models12:53
HeikoS1it doesnt replace everything12:53
@lisitsyntries to :D12:53
HeikoS1these things are more engineering kind of things12:53
HeikoS1no12:54
HeikoS1completely non-interpretable12:54
@lisitsynHeikoS1: adepts would say it doesn't matter12:54
HeikoS1so not good for science, just a tool for engineering12:54
HeikoS1for solving concrete problems12:54
HeikoS1so just another tool12:54
@lisitsynHeikoS1: what I mean is that it could be used to discover exact models12:55
HeikoS1I dont think so12:55
@lisitsynhmm that's good you don't think so12:56
HeikoS1its quite different say learning a graphical model of data and representation learning of a nn12:56
@lisitsynI mean then it is not obvious :)12:56
HeikoS1also keep in mind nns are just gradient descent12:56
HeikoS1so just universal function approximator12:56
HeikoS1thats all12:56
@lisitsynhmm yeah12:56
HeikoS1but a model is something more12:56
HeikoS1which comes at some costs12:56
@lisitsynHeikoS1: well once we learn that function12:57
@lisitsyncan't we learn the exact form of it from the approximator?12:57
@lisitsynwouldn't it be more reliable and have less complexity12:57
HeikoS1but the exact form is useless, it doesnt tell you anything12:58
@lisitsynHeikoS1: why?12:58
HeikoS1you have a representation of the exact form, but its just some weights that dont mean things12:59
@lisitsynHeikoS1: no12:59
@lisitsynI don't mean w1x1 + ..12:59
@lisitsynI mean like analogies with some known procedures12:59
@lisitsynas I said, filters, matrix factorizations12:59
@lisitsynnot really like f(g(p(...))) but some tractable procedure that does the same thing13:00
@lisitsynHeikoS1: my main concern is that such a model is much more 'simpler' in means of capacity13:01
HeikoS1maybe I dont get you correctly13:02
HeikoS1lisitsyn: wanna write it down? :)13:02
@lisitsynHeikoS1: yeah will try13:03
HeikoS1I get what you mean maybe13:04
HeikoS1but I think its really a different goal to learn a representation and to fit a model13:05
HeikoS1you can use things for one another, but the idea behind it is different13:05
shogun-buildbotbuild #122 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/122  blamelist: Wu Lin <yorker.lin@gmail.com>13:14
@lisitsynHeikoS1: I can't stop wondering that I keep being surprised that I know basically nothing about the field I am working in :D13:19
HeikoS1why that?13:19
HeikoS1I think you know a lot13:19
HeikoS1Its not like I know more13:19
HeikoS1just my opinion13:20
@lisitsynHeikoS1: for sure you do :)13:20
HeikoS1but I agree its important to know things about big picture13:20
HeikoS1not just details13:20
@lisitsynHeikoS1: I was really following svms but that's it13:20
@lisitsynand I can't really understand most of bayesian methods now13:20
@lisitsyntrying to keep up with that13:20
@lisitsynbut the thing is that the field gets really big13:21
HeikoS1it does13:21
HeikoS1but as said13:21
HeikoS1its all in bishop ;)13:21
@lisitsynhah13:21
@lisitsynHeikoS1: well for example there is a PGM book by koller13:21
@lisitsyn1280 pages13:21
@lisitsynjust for pgms :D13:22
@lisitsynHeikoS1: or say a book by hastie13:23
@lisitsynESL13:23
@lisitsynit is big as hell13:23
HeikoS1its not important to read everything13:23
@lisitsynand still doesn't cover what modern things do13:23
HeikoS1just pick something that you get interested and try to understand it13:23
HeikoS1check whats happening on the ML conferences13:23
HeikoS1and read papers and references13:23
HeikoS1then youll get a good pictire13:23
shogun-buildbotbuild #123 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/123  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:25
@lisitsynHeikoS1: yeah that should work but still its moving quite fast :)13:25
shogun-buildbotbuild #2537 of deb3 - modular_interfaces is complete: Failure [failed python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2537  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>13:36
-!- HeikoS1 [~heiko@host-92-25-65-248.as13285.net] has quit [Quit: Leaving.]14:04
shogun-buildbotbuild #533 of deb4 - python3 is complete: Failure [failed python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/533  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>14:39
-!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has joined #shogun14:51
-!- mode/#shogun [+o HeikoS] by ChanServ14:51
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]15:21
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has joined #shogun15:43
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection]15:49
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun16:01
shogun-notifier-shogun: Sergey Lisitsyn :develop * 5a4abb8 / src/shogun/io/ProtobufFile.h: https://github.com/shogun-toolbox/shogun/commit/5a4abb8b5299dbc9f73e58a651dd0c7e83e2ee7016:01
shogun-notifier-shogun: Fixed compilation error under Mac OS X16:01
shogun-notifier-shogun: Sergey Lisitsyn :develop * e911b4a / src/shogun/io/ProtobufFile.h: https://github.com/shogun-toolbox/shogun/commit/e911b4aaaf08e720ab3c85ae46ed72be3b85d7a216:03
shogun-notifier-shogun: Adjusted condition when TYPE_BOOL is restored16:03
shogun-notifier-shogun: [skip ci]16:03
-!- travis-ci [~travis-ci@ec2-54-161-195-167.compute-1.amazonaws.com] has joined #shogun16:04
travis-ciit's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/4148864416:04
-!- travis-ci [~travis-ci@ec2-54-161-195-167.compute-1.amazonaws.com] has left #shogun []16:04
shogun-buildbotbuild #3262 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3262  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>16:15
@lisitsynehm I should do this better16:16
shogun-buildbotbuild #3263 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3263  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>16:21
-!- pickle27 [~pickle27@216.191.105.146] has joined #shogun17:05
-!- Phawx [~Phawx@cpe-68-203-5-71.austin.res.rr.com] has joined #shogun17:38
-!- Phawx [~Phawx@cpe-68-203-5-71.austin.res.rr.com] has quit [Ping timeout: 258 seconds]18:14
-!- PirosB3_ [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has joined #shogun18:38
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has quit [Read error: Connection reset by peer]18:38
-!- PirosB3_ is now known as PirosB318:38
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]19:03
-!- abhi [671a3516@gateway/web/freenode/ip.103.26.53.22] has joined #shogun19:51
abhihi everyone19:52
abhii would like to work on machine learning and contribute to shogun19:53
abhii would like help on this to get started19:53
-!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 240 seconds]20:29
-!- PirosB3 [~pirosb3@ip-66.net-81-220-115.brest.rev.numericable.fr] has quit [Quit: PirosB3]20:36
-!- abhi [671a3516@gateway/web/freenode/ip.103.26.53.22] has quit [Quit: Page closed]20:48
-!- DSrupt [~DSrupt@73.6.109.86] has joined #shogun22:01
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun23:00
shogun-notifier-shogun: lisitsyn :develop * c6c1c7e / src/shogun/io/ProtobufFile.h: https://github.com/shogun-toolbox/shogun/commit/c6c1c7e48f34242f775c36ae2bf3cdc8f6b7b76923:00
shogun-notifier-shogun: Fixed macosx ifdef in ProtobufFile23:00
shogun-notifier-shogun: Sergey Lisitsyn :develop * d63277d / src/shogun/io/ProtobufFile.h: https://github.com/shogun-toolbox/shogun/commit/d63277d488cc352f41d3904c66d17f92961b767e23:00
shogun-notifier-shogun: Merge pull request #2621 from lisitsyn/fix/protobuf-macosx23:00
shogun-notifier-shogun:23:00
shogun-notifier-shogun: Fixed macosx ifdef in ProtobufFile23:00
shogun-buildbotbuild #3264 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/326423:04
shogun-buildbotbuild #889 of precise - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/889  blamelist: lisitsyn <lisitsyn@yandex-team.ru>23:06
shogun-buildbotbuild #916 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/916  blamelist: lisitsyn <lisitsyn@yandex-team.ru>23:07
shogun-buildbotbuild #890 of precise - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/890  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>23:07
shogun-buildbotbuild #917 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/917  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>23:12
shogun-buildbotbuild #124 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/124  blamelist: lisitsyn <lisitsyn@yandex-team.ru>23:26
shogun-buildbotbuild #125 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/125  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>23:34
-!- HeikoS [~heiko@0545399b.skybroadband.com] has joined #shogun23:37
-!- mode/#shogun [+o HeikoS] by ChanServ23:37
shogun-buildbotbuild #2538 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/253823:46
shogun-buildbotbuild #534 of deb4 - python3 is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/53423:55
--- Log closed Thu Nov 20 00:00:16 2014
« Tuesday, 2014-11-18 Index Thursday, 2014-11-20 »

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