Open in new window / Try shogun cloud
--- Log opened Thu Jan 19 00:00:47 2017
CaBaanyone up?00:06
CaBawiking, HeikoS?00:06
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Quit: Leaving.]00:24
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 255 seconds]00:25
-!- Netsplit *.net <-> *.split quits: @sukey00:26
-!- Netsplit over, joins: @sukey00:27
@sukeyNew Commit "some minor fixes for Random Forest" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/48fabe18e9c9735b8fab4a2e082c00a3f334bdd301:57
-!- mikeling [uid89706@gateway/web/irccloud.com/x-wpvemmgnfzrkvjde] has joined #shogun02:18
@sukeyNew Commit "some minor fixes for Random Forest" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/dadd09e05f34981ec94b160c40a2c07d471849b602:32
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has joined #shogun03:19
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/19325096803:19
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has left #shogun []03:20
@sukeyNew Commit "Merge branch 'feature/KernelOMP' into develop" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/db05cba946f4b163ff57097832e8391cc0b1c8e703:22
--- Log closed Thu Jan 19 03:38:45 2017
--- Log opened Thu Jan 19 03:38:53 2017
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun03:38
-!- Irssi: #shogun: Total of 16 nicks [4 ops, 0 halfops, 0 voices, 12 normal]03:38
-!- Irssi: Join to #shogun was synced in 6 secs03:38
-!- lisitsyn_ [~lisitsyn@37.139.2.75] has joined #shogun03:41
-!- Netsplit *.net <-> *.split quits: rcurtin, lisitsyn, shogun-t1olbox03:43
-!- lisitsyn_ is now known as lisitsyn03:43
@sukeyNew Commit "Update MS Dirent implementation" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/36c124481d7e1b6c6b7eb1e1bc3b46d108b791e703:53
-!- Netsplit over, joins: rcurtin03:58
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has joined #shogun04:11
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/19325837304:11
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has left #shogun []04:11
shogun-buildbotbuild #98 of deb4 - python3 is complete: Failure [failed test python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/98  blamelist: Viktor Gal <viktor.gal@maeth.com>04:11
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has joined #shogun04:50
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/19326063704:50
-!- travis-ci [~travis-ci@ec2-54-167-177-79.compute-1.amazonaws.com] has left #shogun []04:50
-!- suhas2go [uid201652@gateway/web/irccloud.com/x-spdssovllzhubohi] has joined #shogun08:50
-!- praisethemoon [~praisethe@197.3.11.178] has joined #shogun08:51
-!- besser82 [~besser82@fedora/besser82] has joined #shogun09:47
-!- mode/#shogun [+o besser82] by ChanServ09:47
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun09:52
-!- mode/#shogun [+o HeikoS] by ChanServ09:53
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 255 seconds]10:17
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 256 seconds]10:20
-!- besser82 [~besser82@fedora/besser82] has joined #shogun10:31
-!- mode/#shogun [+o besser82] by ChanServ10:32
-!- besser82 [~besser82@fedora/besser82] has quit [Remote host closed the connection]10:36
-!- besser82 [~besser82@fedora/besser82] has joined #shogun10:36
-!- mode/#shogun [+o besser82] by ChanServ10:36
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 245 seconds]10:42
-!- besser82 [~besser82@fedora/besser82] has joined #shogun11:17
-!- mode/#shogun [+o besser82] by ChanServ11:17
-!- besser82 [~besser82@fedora/besser82] has quit [Remote host closed the connection]11:36
-!- besser82 [~besser82@fedora/besser82] has joined #shogun11:36
-!- mode/#shogun [+o besser82] by ChanServ11:36
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 256 seconds]11:41
-!- besser82 [~besser82@fedora/besser82] has joined #shogun12:00
-!- mode/#shogun [+o besser82] by ChanServ12:00
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 245 seconds]12:10
-!- suhas2go [uid201652@gateway/web/irccloud.com/x-spdssovllzhubohi] has quit [Quit: Connection closed for inactivity]12:27
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun13:23
-!- mode/#shogun [+o HeikoS] by ChanServ13:23
CaBahey HeikoS13:23
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 252 seconds]13:27
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun13:37
-!- mode/#shogun [+o HeikoS] by ChanServ13:37
CaBayou guys need better internet connectivity ;)13:38
@HeikoShehe ;)13:39
@HeikoSwell life, work and all that is also going on13:39
CaBaHeikoS: :)13:39
CaBaHeikoS: can i ask you a quick question nontheless? :P13:39
CaBaHeikoS: i have a custom string kernel that currently works on StringFeatures<char>... now i want to enhance it's performance by precomputing some indices over those sequences and use them instead of the sequences itself in ::compute().13:41
@HeikoSsure13:41
@HeikoSCaBa: I think for that you would have to modify the code a bit13:41
CaBaHeikoS: my problem is - i only realized afterwards that the lhs and rhs members of CKernel are *not* exclusively modified through CKernel::init() - i thought that was the only place where i had to update the indices13:42
CaBaHeikoS: i was thinking... i should precompute everything and then have a kernel that works on the indices only... but those are complex objects. is there a CFeatures class for that?13:43
CaBaHeikoS: or how about CStringFeatures<T>? Does that actually work for *any* T, as long as compute() can handle T?13:44
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 276 seconds]13:46
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun13:59
-!- mode/#shogun [+o HeikoS] by ChanServ14:00
-!- besser82 [~besser82@fedora/besser82] has joined #shogun14:06
-!- mode/#shogun [+o besser82] by ChanServ14:06
CaBaHeikoS: still got my last 3 msgs?14:13
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 248 seconds]14:15
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun14:21
-!- mode/#shogun [+o HeikoS] by ChanServ14:21
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 276 seconds]14:25
-!- besser82 [~besser82@fedora/besser82] has quit [Remote host closed the connection]14:27
-!- mizari [~mizari@user-79-136-131-162.tomtelnet.ru] has quit [Read error: Connection reset by peer]14:43
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun14:44
-!- mode/#shogun [+o HeikoS] by ChanServ14:44
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 255 seconds]14:52
-!- besser82 [~besser82@fedora/besser82] has joined #shogun14:54
-!- mode/#shogun [+o besser82] by ChanServ14:54
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has joined #shogun15:08
-!- mode/#shogun [+o HeikoS] by ChanServ15:08
-!- HeikoS [~heiko@host81-148-236-30.range81-148.btcentralplus.com] has quit [Ping timeout: 240 seconds]15:12
-!- mizari [~mizari@user-79-136-131-162.tomtelnet.ru] has joined #shogun15:40
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 276 seconds]16:11
-!- besser82 [~besser82@fedora/besser82] has joined #shogun16:12
-!- mode/#shogun [+o besser82] by ChanServ16:12
-!- abhinav [~abhinav@103.196.221.113] has joined #shogun16:35
-!- HeikoS [~heiko@eduroam-int-pat-8-240.ucl.ac.uk] has joined #shogun16:36
-!- mode/#shogun [+o HeikoS] by ChanServ16:36
-!- HeikoS [~heiko@eduroam-int-pat-8-240.ucl.ac.uk] has quit [Ping timeout: 240 seconds]16:41
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 276 seconds]16:50
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun16:54
-!- mode/#shogun [+o HeikoS] by ChanServ16:54
-!- abhinav [~abhinav@103.196.221.113] has quit [Ping timeout: 240 seconds]17:08
-!- mikeling [uid89706@gateway/web/irccloud.com/x-wpvemmgnfzrkvjde] has quit [Quit: Connection closed for inactivity]17:19
-!- besser82 [~besser82@fedora/besser82] has joined #shogun17:54
-!- mode/#shogun [+o besser82] by ChanServ17:54
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 258 seconds]18:00
-!- praisethemoon [~praisethe@197.3.11.178] has quit [Ping timeout: 240 seconds]18:08
CaBaHeikoS: ping18:19
@HeikoSjo18:19
CaBaHeikoS: tell me... SGString, SGStringList - they are not implemented for any type T?18:20
CaBalike - the generic version is non functional?18:20
@HeikoSCaBa: only for basic types I think18:20
@HeikoSbut not sure18:20
CaBa>.<18:20
@HeikoSwhat type did you try?18:20
CaBamy own class18:20
@HeikoSah18:20
@HeikoSreason why you do that?18:21
CaBawell... i want to use the index data structures as features and CKernel::compute() on that...18:21
@HeikoSmmh18:21
@HeikoSI see18:21
CaBaHeikoS: is there any feature type that can take "whatever" per sample and let me handle the kernel function myself in ::compute() and that's it?18:22
@HeikoSI wanted to kick out supporting strings of vectors/matrices a long time ago18:22
@HeikoSnope that is not there18:23
@HeikoSI mean it can be added18:23
CaBa>.<18:23
@HeikoSand precomputing the kernel?18:23
@HeikoSCaBa: your think would be a hack18:23
CaBawhat's the precomputing the kernel option?18:23
CaBaCCustomKernel?18:24
@HeikoSas the kernel behaves differently based on what feature type is in there (at least in the current way we do this)18:24
@HeikoSCaBa: yes precomputing the matrix18:24
@HeikoSand using that18:24
@HeikoSvia CustmKernel18:24
CaBaHeikoS: is CustomKernel fully functional? can i use it in CCombinedKernel?18:25
@HeikoSeven works with cross-validation18:25
@HeikoSCaBa: I did that once, but it is a few years ago18:25
CaBaHeikoS: ok, so subsetting works in CCustomKernel?18:25
@HeikoSyes18:25
@HeikoSbut xvalidation and custom kernel AND combined kernel18:25
@HeikoSI dont know18:25
@HeikoSprobably not18:25
@HeikoSI think I wrote an example for this at some point18:26
@HeikoSand added the feature to the news18:26
@HeikoSbut that must be 2011 or so18:26
-!- besser82 [~besser82@fedora/besser82] has joined #shogun18:27
-!- mode/#shogun [+o besser82] by ChanServ18:27
CaBaHeikoS: for custom kernel you mean?18:27
@HeikoScustom kernel xvalidation18:29
@HeikoSand also custom kernel and combined kernel (I think)18:29
@HeikoSAll of this stuff is quite useful to sort out generally18:29
CaBaHeikoS: so maybe it works alltogether?18:29
@HeikoSso we would be quite happy to have that18:29
@HeikoSI mean give it a try18:29
@HeikoSif it runs, I am sure it does sensible things18:29
CaBaHeikoS: so with customkernel there is nothing i need to do other than compute the SGMatrix<> and construct CCustomKernel(matrix)?18:32
CaBaHeikoS: does it still handle normalization through set_normalizer()?18:32
CaBaor do i have to do that beforehand, too?18:32
@HeikoSbefore18:32
@HeikoSyou can just pass a CKernel in the constructor of CCustomKernel I think18:32
@HeikoSand it does the job for you18:32
CaBawhat do you mean?18:33
CaBawhat CKernel should i pass?18:33
@HeikoSany shogun kernel18:34
@HeikoSbut matrix is fine as well18:34
CaBaHeikoS: so can it handle the normalization for me?18:34
@HeikoSnope18:34
@HeikoSit just stores numbers18:34
@HeikoS(it might but better day no for now)18:35
CaBayee, i should have compiled every now and then... took me a full day of implementation to realize that SGString<T> won't compile for any 'T'18:36
@HeikoSoh yes, alwayw compile all the time18:36
@HeikoSuse ccache to make it fast18:36
CaBai'm just compiling my own tool, that's just 10 compilation units or so :P18:37
@HeikoSah i see18:37
@HeikoSokok18:37
CaBa:'(18:38
CaBawell here we go, attempt #3 with CCustomKernel...18:38
@HeikoSlet me know how it gues18:38
@HeikoSshare results if you  make improvements18:38
CaBaHeikoS: hmmm - so normally normalization would be applied on every subset during x-val, right?18:41
@HeikoSI think that is what is done18:41
@HeikoSbut better check the code18:42
CaBaHeikoS: that cannot be achieved with a pre-computed and pre-normalized matrix if the customkernel cannot do it18:42
@HeikoSyep18:42
@HeikoStrue18:42
@HeikoSvery true18:42
@HeikoSso then no normalisation18:42
@HeikoSin that context18:42
@HeikoSor the custom kernel does that on its subsets18:42
@HeikoSbut that I know it doesnt yet (would be easy to add)18:42
CaBa:(18:45
-!- abhinav [~abhinav@103.196.221.113] has joined #shogun18:45
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 252 seconds]18:47
-!- besser82 [~besser82@fedora/besser82] has quit [Remote host closed the connection]18:52
-!- besser82 [~besser82@fedora/besser82] has joined #shogun18:52
-!- mode/#shogun [+o besser82] by ChanServ18:52
-!- abhinav [~abhinav@103.196.221.113] has quit [Quit: Leaving]18:55
-!- abhinav [~abhinav@103.196.221.113] has joined #shogun18:55
-!- abhinav [~abhinav@103.196.221.113] has quit [Quit: Leaving]20:04
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 256 seconds]21:48
-!- besser82 [~besser82@fedora/besser82] has joined #shogun21:50
-!- mode/#shogun [+o besser82] by ChanServ21:50
@besser82lisitsyn, you there?21:51
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 255 seconds]22:08
-!- besser82 [~besser82@fedora/besser82] has joined #shogun23:06
-!- mode/#shogun [+o besser82] by ChanServ23:06
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 255 seconds]23:18
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun23:23
-!- mode/#shogun [+o HeikoS] by ChanServ23:23
CaBaHeikoS: welcome home :P23:29
--- Log closed Fri Jan 20 00:00:49 2017