--- Log opened Fri Nov 25 00:00:30 2016 | ||
@wiking | i mean in the first place | 00:01 |
---|---|---|
@wiking | why the fuck we call it examples | 00:01 |
@wiking | when it's cookbook | 00:01 |
@wiking | i dont even get this fucking logic | 00:01 |
@wiking | a name of a link can be totally different | 00:01 |
@wiking | then the link itself | 00:01 |
@sukey | New Commit "General clean up and link fixes" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/20b7983e43b284aa188fb64797a9182d1bde2a46 | 00:04 |
@sukey | New Commit "remove quicklinks | 00:05 |
@sukey | File is very short these days so no need" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/111c389512767e64b7fb269d66b977d957fc2423 | 00:05 |
@HeikoS | "cookbook" doenst have a meaning for website | 00:06 |
@HeikoS | "examples" does | 00:06 |
@HeikoS | and it should be called examples :) | 00:06 |
@HeikoS | we should git grep "cookbook" and rename it | 00:06 |
@HeikoS | same for git grep "meta" | 00:07 |
@wiking | but man | 00:11 |
@wiking | i get it | 00:11 |
@wiking | but fuck | 00:11 |
@wiking | <a href="cookbook">examples</a> | 00:11 |
@wiking | what's the problem with this? :) | 00:11 |
@wiking | i mean we are actually doing this | 00:13 |
@wiking | with the redirect | 00:13 |
@wiking | which is a foobar | 00:13 |
@HeikoS | I dont know | 00:13 |
@HeikoS | it shouldnt be | 00:13 |
@HeikoS | the question "why did we do this" doesnt make sense :) | 00:13 |
@HeikoS | what makes sense is "I fixed it" :D | 00:13 |
@HeikoS | the wiki is totally borken | 00:14 |
@HeikoS | every second link is 404 | 00:14 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 00:17 |
@wiking | HeikoS, well if it doesn make fucking sense to do that | 00:19 |
@wiking | then why would we need to fix it | 00:19 |
@wiking | that's my point | 00:19 |
@wiking | i dont see the point | 00:19 |
@wiking | why we keep fucking redirecting http calls | 00:19 |
@wiking | when this si the first fucking time | 00:19 |
@wiking | we defined this link | 00:19 |
@wiking | then why dont we use the correct naming | 00:20 |
@wiking | redirect is for when you've moved content somewhere | 00:20 |
@wiking | not that when you actually created a new link | 00:20 |
@wiking | but you already redirect | 00:20 |
@wiking | it's like defining a function | 00:20 |
@wiking | for the first time | 00:20 |
@wiking | and mark it deprecated right away | 00:20 |
@sukey | Wiki page: GSoC-follow-up-blog-posts edited on shogun-toolbox/shogun by karlnapf | 00:21 |
@HeikoS | ohoh | 00:28 |
@HeikoS | 502 bad gateway | 00:28 |
@sukey | Wiki page: Testing created on shogun-toolbox/shogun by karlnapf | 00:32 |
@wiking | yes | 00:35 |
@wiking | working on the fucking webstie | 00:35 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 00:42 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 00:44 |
@wiking | i really hope that soon we can throw out this website | 00:47 |
@wiking | flask is a piece of shit | 00:47 |
@HeikoS | wiking: jo | 00:54 |
@HeikoS | saw this | 00:54 |
@HeikoS | https://github.com/shogun-toolbox/shogun/wiki/CodingStyle | 00:54 |
@HeikoS | ? | 00:55 |
@wiking | yeah this should be thrown out | 00:55 |
@wiking | and use clang-format | 00:55 |
@HeikoS | shall I delete? | 00:55 |
@HeikoS | look at this, even worse | 00:55 |
@HeikoS | https://github.com/shogun-toolbox/shogun/wiki/Code-style | 00:55 |
@wiking | it's getting there | 00:55 |
@wiking | we should use these | 00:55 |
@wiking | as the basis of writing the formatter | 00:55 |
@HeikoS | ok i keep this page and refer to it | 00:56 |
@HeikoS | kk | 00:56 |
@wiking | and once that's done | 00:56 |
@wiking | we throw out these pages | 00:56 |
@sukey | Wiki page: Developing-Shogun created on shogun-toolbox/shogun by karlnapf | 00:58 |
@HeikoS | wiking: https://github.com/shogun-toolbox/shogun/wiki/Developing-Shogun/_edit | 01:00 |
@HeikoS | any comments? | 01:00 |
@HeikoS | ah sorry | 01:00 |
@HeikoS | proper link | 01:00 |
@wiking | use | 01:00 |
@wiking | shorter headers | 01:00 |
@wiking | Shogun development cycle in a nutshell - “How to get my code into Shogun” | 01:00 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:00 |
@HeikoS | https://github.com/shogun-toolbox/shogun/wiki/Developing-Shogun | 01:00 |
@wiking | use links | 01:00 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:00 |
@wiking | (fork on github) | 01:00 |
@HeikoS | I didnt on purpose in fact | 01:01 |
@wiking | what using links? | 01:01 |
@HeikoS | yes in here | 01:01 |
@HeikoS | minimal | 01:01 |
@wiking | jsut to feel like it's 85? | 01:01 |
@wiking | where we dont have references? | 01:01 |
@HeikoS | no maintainence | 01:01 |
@HeikoS | ance | 01:01 |
@wiking | what that link is going to exist | 01:01 |
@wiking | forever | 01:01 |
@wiking | worst case | 01:01 |
@HeikoS | fork yes | 01:01 |
@wiking | redirect | 01:01 |
@wiking | what will change then? | 01:02 |
@HeikoS | I want the students to read the guides | 01:02 |
@HeikoS | github guides are quite good | 01:02 |
@HeikoS | so if they dont know how to fork | 01:02 |
@HeikoS | they read on it | 01:02 |
@wiking | yeah then give a link to that | 01:02 |
@wiking | now there's nothing | 01:02 |
@HeikoS | its there | 01:02 |
@wiking | ah above | 01:02 |
@HeikoS | please google, please read :) | 01:02 |
@wiking | fuck you really | 01:02 |
@wiking | think that they gonna do that | 01:03 |
@HeikoS | well its wiki | 01:03 |
@wiking | if it's not next to the line? | 01:03 |
@wiking | :))) | 01:03 |
@HeikoS | haha | 01:03 |
@HeikoS | lots of the wiki stuff i wrote is too verbose and has all 404 links | 01:03 |
@HeikoS | so just keeping it down | 01:03 |
@HeikoS | anyways | 01:03 |
@HeikoS | time for testing page | 01:03 |
@wiking | mmm | 01:04 |
@wiking | can i ask a question? | 01:04 |
@HeikoS | what kind of question is that? :) | 01:05 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/GaussianKernel.cpp#L85-L91 | 01:05 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/ShiftInvariantKernel.cpp#L54-L60 | 01:06 |
@wiking | is called twice | 01:06 |
@wiking | m_distance->init(l, r); | 01:06 |
@wiking | right? | 01:06 |
@wiking | is there a sane reason for it? | 01:06 |
@wiking | since m_distance is protected | 01:07 |
@wiking | ShiftInvariantKernel.cpp | 01:07 |
@HeikoS | no thats a refactoring bug | 01:07 |
@wiking | ok great | 01:07 |
@wiking | fuck thi s shit | 01:07 |
@wiking | :) | 01:07 |
@HeikoS | haha | 01:07 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 01:10 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 01:10 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:11 |
@wiking | HeikoS, | 01:13 |
@wiking | another question | 01:13 |
@HeikoS | wiking: yep | 01:16 |
@HeikoS | I also have one | 01:16 |
@wiking | shoot | 01:16 |
@HeikoS | things like dev readme | 01:16 |
@wiking | i'm just looking into what would be a sane unit test | 01:16 |
@HeikoS | should be in wiki or doc/readme | 01:16 |
@wiking | for get_matrix | 01:16 |
@wiking | for gaussian | 01:16 |
@wiking | or for which matrix | 01:16 |
@wiking | should we actually do a test | 01:16 |
@HeikoS | not sure I understand | 01:16 |
@HeikoS | but I would do n=2 d=2 | 01:16 |
@HeikoS | so a 2x2 matrix | 01:17 |
@HeikoS | and then compare against python or so | 01:17 |
@HeikoS | or | 01:17 |
@HeikoS | you compare against kernel->compute | 01:17 |
@HeikoS | thats better actually | 01:17 |
@HeikoS | and then just test kernel->compute | 01:17 |
@HeikoS | with data | 01:17 |
@HeikoS | and in that case, data should be random and numbers slightly larger | 01:17 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:18 |
@wiking | well we have the parallel shit all over Kernel.cpp | 01:18 |
@wiking | for example even in the case of get_matrix() | 01:18 |
@wiking | i mean get_kernel_matrix | 01:19 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:19 |
@wiking | hence a 2x2 matrix | 01:19 |
@wiking | is pretty shitty | 01:19 |
@HeikoS | nono | 01:19 |
@wiking | for testing parallel computation | 01:19 |
@wiking | :) | 01:19 |
@wiking | and testing whether it is true | 01:19 |
@HeikoS | so there you want to test the matrix aggregation | 01:19 |
@wiking | i mean works correctly | 01:19 |
@wiking | so basically i just wanna do | 01:20 |
@HeikoS | so just create some random data and compute a slightly bigger matrix on multiple threads | 01:20 |
@wiking | WhateverKernel(l,r) | 01:20 |
@wiking | k->get_kernel_matrix() | 01:20 |
@HeikoS | you wanna do that for all kernels? | 01:20 |
@HeikoS | or any? | 01:21 |
@HeikoS | well anyways | 01:21 |
@HeikoS | do that, and then compare output to a nested loop that calls k->compute | 01:21 |
@wiking | nono | 01:23 |
@wiking | any kernel | 01:23 |
@wiking | i mean the thing is that | 01:23 |
@wiking | there are multiple places to fail | 01:23 |
@wiking | due to parallel stuff | 01:24 |
@wiking | and i need a big enough kernel | 01:24 |
@sukey | Wiki page: AUTHORS edited on shogun-toolbox/shogun by karlnapf | 01:24 |
@wiking | so that i could test n > 2 threads | 01:24 |
@wiking | and i really wonder | 01:30 |
@wiking | why GaussianKernel tetss | 01:30 |
@wiking | are in Kernel_unittest.cc | 01:31 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 01:34 |
@HeikoS | wiking: they shouldnt | 01:34 |
@HeikoS | n=1000 should be ok | 01:34 |
@HeikoS | thats like 10^6 elements | 01:35 |
@HeikoS | even 100 should be ok for test | 01:35 |
@wiking | dim? | 01:35 |
@wiking | 3? | 01:35 |
@wiking | or 2 | 01:35 |
@wiking | ? | 01:35 |
@wiking | dohe | 01:37 |
@wiking | compute is protected | 01:37 |
@wiking | ah ok kernel(i,j) | 01:38 |
@wiking | HeikoS, heeeeeeeeeelp | 01:40 |
@wiking | ah ok | 01:41 |
@wiking | found the bug | 01:41 |
@wiking | i mean that at least the unit test now finds the bug | 01:41 |
@wiking | wonder if i should actually add 2 tests | 01:41 |
@wiking | 1 fixed to 1 thread, and second unit test where the num thread is num cpu | 01:42 |
@wiking | lol | 01:42 |
@wiking | you will not believe | 01:42 |
@wiking | :) | 01:42 |
@wiking | unit tests passes with num threads = 1, 2, 4, 5 | 01:43 |
@wiking | but fails with 3, 6 | 01:43 |
@HeikoS | haha | 01:51 |
@HeikoS | consistently? | 01:51 |
@HeikoS | dim = 100 | 01:53 |
@sukey | Wiki page: Testing edited on shogun-toolbox/shogun by karlnapf | 02:08 |
@wiking | yes yes | 02:10 |
@wiking | consistently | 02:10 |
@wiking | dim = | 02:10 |
@wiking | 3 | 02:10 |
@wiking | with dim = 100 does not happen | 02:11 |
@wiking | but of course this is not a solution :) | 02:11 |
@wiking | if i set dim = 3 | 02:12 |
@wiking | i mean dim = 4 | 02:12 |
@sukey | Wiki page: README_developer edited on shogun-toolbox/shogun by karlnapf | 02:12 |
@wiking | same pattern | 02:12 |
@sukey | Wiki page: Developing-Shogun edited on shogun-toolbox/shogun by karlnapf | 02:30 |
@sukey | Wiki page: Testing edited on shogun-toolbox/shogun by karlnapf | 02:35 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 250 seconds] | 02:40 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 04:21 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 04:21 | |
@sukey | New branch feature/KernelOMP created on shogun-toolbox/shogun | 07:09 |
@sukey | New Commit "Add some tolerance for testing matrix equality in PeriodicKernel_unittest" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/1dd88f0d5d62fb3db89c037dde4f3601d639ec88 | 07:09 |
-!- travis-ci [~travis-ci@ec2-54-159-250-106.compute-1.amazonaws.com] has joined #shogun | 07:58 | |
travis-ci | it'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/178768831 | 07:58 |
-!- travis-ci [~travis-ci@ec2-54-159-250-106.compute-1.amazonaws.com] has left #shogun [] | 07:58 | |
@wiking | CaBa, man you are lucky that you are using osx :))))0 | 09:44 |
@wiking | https://github.com/shogun-toolbox/shogun/commit/a6d7ae16c667a454f6ab7301e1b44ae2fa1ca6ed | 09:44 |
CaBa | wiking: hm? how "not correct" was the detected number on linux? | 12:03 |
@wiking | it was = 1 | 12:03 |
@wiking | :) | 12:03 |
CaBa | wiking: well in my case better that most high values ;) | 12:22 |
CaBa | but anyway not a mac issue i understand ;) | 12:22 |
@wiking | yeah in mac it was ok | 12:25 |
@sukey | Pull Request #3564 "Temp" opened by sorig - https://github.com/shogun-toolbox/shogun/pull/3564 | 12:48 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 12:53 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 12:53 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Quit: Leaving.] | 14:07 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 14:07 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:07 | |
@sukey | New Commit "Add some tolerance for testing matrix equality in PeriodicKernel_unittest" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/7c22cc640ad808805b70b4eb433ce85577c3a4b3 | 14:29 |
@wiking | HeikoS, oooooooooo | 14:54 |
@HeikoS | wiking: ? | 14:54 |
@HeikoS | man these dev docs | 14:54 |
@HeikoS | messy :) | 14:54 |
@wiking | HeikoS, have u seen my email? :) | 14:54 |
@HeikoS | yes | 14:54 |
@HeikoS | I am happy :)D | 14:54 |
@wiking | :0 | 14:54 |
@wiking | btw | 14:54 |
@wiking | i'm afraid of | 14:54 |
@wiking | kmeans rebase | 14:54 |
@wiking | (it fails) | 14:54 |
@wiking | so i thinks that Distance is not thread safe :) | 14:55 |
@HeikoS | ouch | 14:56 |
@wiking | but i reckon if we manage to fix now the failing shit in this feature branch | 15:10 |
@wiking | suddenly windows branch will work as well | 15:10 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 15:11 |
@HeikoS | cool! | 15:13 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has joined #shogun | 15:17 | |
travis-ci | it'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/178841791 | 15:17 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has left #shogun [] | 15:17 | |
CaBa | My program currently uses basically gridsearch ms, strat xval and libsvm from shogun | 15:42 |
CaBa | am i right that i therefore dont use any parallel features that are implemented directly in shogun but those that are implemented in libsvm? | 15:42 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 15:54 |
@sukey | New Commit "add development guide" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/dd48b82348cec309495fd7224a6f507bd2a48b32 | 15:54 |
shogun-buildbot | build #86 of trusty - libshogun - viennacl is complete: Failure [failed] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/86 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 15:54 |
shogun-buildbot | build #71 of deb1 - libshogun is complete: Failure [failed] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/71 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 15:54 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 15:55 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 15:57 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has joined #shogun | 15:58 | |
travis-ci | it's Heiko Strathmann'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/178860797 | 15:58 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has left #shogun [] | 15:58 | |
@HeikoS | wiking: https://github.com/shogun-toolbox/docs/blob/master/DEVELOPING.md | 15:58 |
@sukey | Wiki page: Shogun-development-guidelines edited on shogun-toolbox/shogun by karlnapf | 15:59 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 16:02 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 16:03 |
@sukey | Wiki page: Shogun-development-guidelines edited on shogun-toolbox/shogun by karlnapf | 16:03 |
@sukey | Issue #2600 "Shogun (documented & API) examples" closed by karlnapf - https://github.com/shogun-toolbox/shogun/issues/2600 | 16:04 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 16:05 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 16:10 |
@sukey | New Commit "some updates to readmes" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/86d233e9a905d4b2f6584abdf5fc05d1bffd1a55 | 16:23 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 16:37 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 16:38 |
@sukey | Wiki page: Google-Summer-of-Code-2016-Projects edited on shogun-toolbox/shogun by karlnapf | 16:49 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 16:52 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 16:53 |
@sukey | Wiki page: Shogun-development-guidelines edited on shogun-toolbox/shogun by karlnapf | 16:54 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 16:55 |
@sukey | Wiki page: Shogun-development-guidelines edited on shogun-toolbox/shogun by karlnapf | 16:56 |
@sukey | Wiki page: Shogun-development edited on shogun-toolbox/shogun by karlnapf | 16:56 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 16:57 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 16:57 |
@sukey | Wiki page: Dev-Tips edited on shogun-toolbox/shogun by karlnapf | 17:00 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 17:00 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 17:01 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 17:01 |
@sukey | Wiki page: Home edited on shogun-toolbox/shogun by karlnapf | 17:01 |
@sukey | Wiki page: Dev-Tips edited on shogun-toolbox/shogun by karlnapf | 17:02 |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has joined #shogun | 17:12 | |
travis-ci | it's Heiko Strathmann'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/178868257 | 17:12 |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has left #shogun [] | 17:12 | |
@sukey | Issue #3000 "Replace an undocumented example by a cookbook page"- https://github.com/shogun-toolbox/shogun/issues/3000 | 17:27 |
@sukey | Issue #3555 "Port all existing (python, libshogun) examples to meta examples"- https://github.com/shogun-toolbox/shogun/issues/3555 | 17:27 |
@sukey | Wiki page: GSoC-follow-up-blog-posts edited on shogun-toolbox/shogun by karlnapf | 17:54 |
@sukey | Issue #3565 "Awesome new Wiki, Thanks. But I don't know where the details about SG_REF" opened by hongguangguo - https://github.com/shogun-toolbox/shogun/issues/3565 | 17:58 |
CaBa | wiking: ping | 18:12 |
@wiking | pong | 18:14 |
CaBa | wiking: for CombinedKernel - do i have to add the individual kernel weights manually to the model selection process or is learning the kernel weights part of the kernel internals? | 18:15 |
@wiking | mmm | 18:15 |
@wiking | good question | 18:15 |
@wiking | m_parameters->add_vector(&sv_weight, &sv_count, "sv_weight", | 18:17 |
@wiking | "Support vector weights."); | 18:17 |
@wiking | SG_ADD(&append_subkernel_weights, "append_subkernel_weights", | 18:17 |
@wiking | "If subkernel weights are appended.", MS_AVAILABLE); | 18:17 |
@wiking | so this is what you want | 18:17 |
@sukey | New Commit "updated EXAMPLES.MD and fixes" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/10037267ceb2aed7ebbe47c05bfe798564bca5ad | 18:18 |
CaBa | wiking: yeah i was afraid that would mean that... :( | 18:18 |
@HeikoS | wiking: jo | 18:18 |
@HeikoS | check the updated dev docs | 18:18 |
@HeikoS | from e.g. wiki | 18:18 |
@HeikoS | we now have INSTALL, INTERFACES, EXAMPLES, DEVELOPING | 18:18 |
CaBa | so that means for MKL i have to define myself all kinds of combinations of in my case 3 weights for the 3 kernels? | 18:19 |
@HeikoS | I think they contain all that is needed for people that are joining newly | 18:19 |
shogun-buildbot | build #88 of trusty - libshogun - viennacl is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/88 | 18:19 |
@HeikoS | CaBa: you still have problems with the model selection? | 18:19 |
@HeikoS | CaBa: did you try to do it in a manual loop? | 18:19 |
@HeikoS | and just using cross-validation inside? | 18:19 |
shogun-buildbot | build #73 of deb1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/73 | 18:20 |
CaBa | HeikoS: right now i use this [1] workaround. adding "kernel" multiple times to the MS tree. is that legit? | 18:20 |
CaBa | [1] https://gitlab.unique-internet.de/snippets/9 | 18:20 |
@HeikoS | yes thats valid | 18:22 |
@HeikoS | but my question is: | 18:22 |
@HeikoS | you can also run a loop over all parameters manually | 18:22 |
@HeikoS | like | 18:22 |
@HeikoS | for p1 in params1: | 18:23 |
@HeikoS | for p2 in params2: | 18:23 |
@HeikoS | machine = instantiate_machine(p1,p2,data) | 18:23 |
@HeikoS | result[p1,p2]=xvalidate(machine) | 18:23 |
@HeikoS | and then take the argmax over the results | 18:23 |
CaBa | HeikoS: yeah, as i in not use the ms code of shogun alltogether you mean? | 18:23 |
@HeikoS | the model selection framework in shogun is not really needed to do grid-searches, as they just require simple loops | 18:23 |
@HeikoS | exactly | 18:24 |
@HeikoS | especially for such complex scenarios as yours, might be easier | 18:24 |
CaBa | HeikoS: sure. i was just still hoping it would be parallelized in the near future ;) | 18:24 |
@HeikoS | the MS code is more for very simple things ... | 18:24 |
@HeikoS | CaBa: yes, it is on the list :) | 18:24 |
CaBa | HeikoS: Saurabh7 was working on that i understand... but it was removed from develop again | 18:24 |
@HeikoS | you can parallelise the loop easiy if you are working in C++ | 18:24 |
@HeikoS | especially if you instantiate a new machine in the most inside loop | 18:24 |
@HeikoS | but I get your point | 18:25 |
@HeikoS | it would be better if shogun could do all that | 18:25 |
CaBa | HeikoS: i am - but i have no knowledge about the thread safety of all these things and data locking etc etc... so i'd like to stay on the "customer" side ;) | 18:25 |
@HeikoS | just trying to help you since you seem blocked by this | 18:25 |
@HeikoS | kk | 18:25 |
@HeikoS | CaBa: just spent the whole day updating dev docs | 18:25 |
@HeikoS | wanted to reply to your Sack overflow question tomorrow | 18:25 |
CaBa | cool! :) | 18:25 |
@HeikoS | and the parallel stuff is also on the list | 18:26 |
CaBa | HeikoS: yeah well that is my current workaround and if it's legit i can use it for now | 18:26 |
@HeikoS | sure as long as it does the thing you want it to do | 18:26 |
CaBa | HeikoS: disadvantage is that print_state=true doesn't report to me all the parameters used | 18:26 |
@HeikoS | i see | 18:26 |
CaBa | HeikoS: so i'm using this patch right now https://github.com/shogun-toolbox/shogun/compare/develop...lkuchenb:feature/printModselResults | 18:26 |
CaBa | HeikoS: which allows me to pass a function that takes all intermediate results for statistics | 18:27 |
CaBa | like that i can also build a properly formated file / output to make parameter heatmaps after MS, which was tricky with the print_state output before ;) | 18:28 |
CaBa | so i kinda anyways had to do that | 18:28 |
@sukey | Wiki page: Getting-involved edited on shogun-toolbox/shogun by karlnapf | 18:28 |
@HeikoS | yes thats ok | 18:28 |
@HeikoS | we would love to have something like this neatly integrated in fact :) | 18:28 |
@HeikoS | callback functions | 18:29 |
CaBa | wiking: can i use this instead? http://www.shogun-toolbox.org/api/latest/classshogun_1_1CMKLClassification.html | 18:30 |
CaBa | wiking: instead of doing model selection on the kernel weights... | 18:30 |
CaBa | it kinda looks as if it was for that ;) | 18:30 |
CaBa | HeikoS: yeah i figured other that print_state there was nothing to get more than the end result out of model selection, right? | 18:31 |
@wiking | yeah | 18:31 |
@sukey | Issue #3565 "Awesome new Wiki, Thanks. But I don't know where the details about SG_REF" closed by karlnapf - https://github.com/shogun-toolbox/shogun/issues/3565 | 18:31 |
@sukey | Wiki page: Dev-Tips edited on shogun-toolbox/shogun by karlnapf | 18:34 |
CaBa | so if i use CMKLClassification() around CLibSVM(), then the machine actually *changes* the kernel during training? | 18:35 |
CaBa | as in - it's parameters | 18:36 |
CaBa | s/changes/alters/ | 18:36 |
@sukey | Wiki page: Minimal-library-usage created on shogun-toolbox/shogun by karlnapf | 18:45 |
@sukey | Wiki page: Minimal-library-usage edited on shogun-toolbox/shogun by karlnapf | 18:46 |
@sukey | Wiki page: README_linalg edited on shogun-toolbox/shogun by karlnapf | 18:49 |
@sukey | Issue #1251 "Drop modelselection framework" closed by karlnapf - https://github.com/shogun-toolbox/shogun/issues/1251 | 18:54 |
@sukey | Wiki page: _Sidebar edited on shogun-toolbox/shogun by karlnapf | 18:54 |
@HeikoS | wiking: you there? | 18:59 |
CaBa | could one of you confirm that i got that correctly? | 19:00 |
@HeikoS | sorry | 19:00 |
@HeikoS | CaBa: I dont know | 19:01 |
@HeikoS | ah yes | 19:01 |
@HeikoS | it changes the weights | 19:01 |
CaBa | no problem, i know you all have other things on your mind ;) | 19:01 |
@HeikoS | no it *does* change the weights | 19:01 |
@HeikoS | i know that for sure | 19:01 |
@wiking | yes | 19:01 |
@HeikoS | check the notebook e.g. | 19:01 |
@wiking | whatsup | 19:01 |
@HeikoS | wiking: wiki problems | 19:01 |
@wiking | ? | 19:01 |
@wiking | what probems? | 19:01 |
@HeikoS | wiking: I edited via github a lot | 19:01 |
@wiking | and? | 19:02 |
CaBa | HeikoS: ok, and it comes with it's own strategy on how to traverse the space of possible weight combinations? | 19:02 |
@HeikoS | and now doing git pull in doc/wiki | 19:02 |
@HeikoS | doesnt give these updates | 19:02 |
@HeikoS | so diverged ... | 19:02 |
@wiking | https://github.com/shogun-toolbox/shogun-wiki/commits/master | 19:03 |
@wiking | :) | 19:03 |
@wiking | see the last commit there | 19:03 |
@HeikoS | yes | 19:03 |
@wiking | dunno wtf is with it honestly | 19:03 |
@HeikoS | probably comes from using the web interface for editing | 19:03 |
@HeikoS | and it doesnt pull/rebase properly or so | 19:04 |
@HeikoS | though it displays the image | 19:04 |
@HeikoS | from the last patch: | 19:04 |
@HeikoS | https://github.com/shogun-toolbox/shogun/wiki/GSoC-2016-summit-discussion | 19:04 |
@HeikoS | wiking: i acutaly think | 19:08 |
@HeikoS | that the wikis are not connected | 19:08 |
@HeikoS | I mean the submodule | 19:08 |
@wiking | could be | 19:09 |
@HeikoS | is just a clone of the github one | 19:09 |
@HeikoS | argh | 19:09 |
@HeikoS | so changing the submodule doesnt change the github one | 19:09 |
@HeikoS | and vice versa | 19:09 |
@HeikoS | the only thing that this submodule thing allows | 19:09 |
@HeikoS | is that one can accept PRs | 19:09 |
@HeikoS | and then copy them over by hand | 19:09 |
@HeikoS | thats what I remember from this solution that we did (remember when I said no I think this is not the way to go?) | 19:10 |
@HeikoS | wiking: ill remove the submodule | 19:10 |
@HeikoS | doesnt make sense | 19:10 |
@HeikoS | check the commits, there is nothing | 19:11 |
@sukey | New Commit "remove wiki submodule" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/17af6d8658ebff9c42df23877cb98e4ce98d9fd5 | 19:12 |
shogun-buildbot | build #74 of deb1 - libshogun is complete: Failure [failed] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/74 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:12 |
shogun-buildbot | build #89 of trusty - libshogun - viennacl is complete: Failure [failed] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/89 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:12 |
@sukey | New Commit "remove old wiki folder" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/3820a3bfba94476d3f13a83dfe62e3ee384d5bad | 19:13 |
shogun-buildbot | build #90 of trusty - libshogun - viennacl is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/90 | 19:14 |
shogun-buildbot | build #75 of deb1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/75 | 19:15 |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has joined #shogun | 19:16 | |
travis-ci | it's Heiko Strathmann'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/178906296 | 19:16 |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has left #shogun [] | 19:16 | |
@sukey | New Commit "Add dev readme" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/7ab14373b836a7d230a7feb13dd2b162d8d91c14 | 19:20 |
@sukey | New Commit "Cleanup" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/ee91a81719a6de4cd86c2131dbb9622426625b6c | 19:22 |
@sukey | New Commit "Mention new dev docs" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/7506acd5325656ddf378fa5f98d21a955b395575 | 19:26 |
@sukey | New Commit "Mention R build fix" to shogun-toolbox/shogun by karlnapf: https://github.com/shogun-toolbox/shogun/commit/49c91765ddc396c6fd8f4d18fe26d572d6647011 | 19:27 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Quit: Leaving.] | 19:29 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 19:29 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:29 | |
CaBa | can one of you tell me what interleaved vs. wrapper optimization in mklclassification is? | 19:29 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 240 seconds] | 19:36 | |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has joined #shogun | 20:43 | |
travis-ci | it's Heiko Strathmann'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/178907417 | 20:43 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has left #shogun [] | 20:43 | |
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun | 20:57 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 20:59 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:59 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 258 seconds] | 21:08 | |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has joined #shogun | 21:17 | |
travis-ci | it's Heiko Strathmann'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/178907844 | 21:17 |
-!- travis-ci [~travis-ci@ec2-54-81-231-103.compute-1.amazonaws.com] has left #shogun [] | 21:17 | |
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Quit: Leaving] | 21:32 | |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has joined #shogun | 21:56 | |
travis-ci | it's Heiko Strathmann'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/178908281 | 21:56 |
-!- travis-ci [~travis-ci@ec2-54-205-110-146.compute-1.amazonaws.com] has left #shogun [] | 21:56 | |
-!- travis-ci [~travis-ci@ec2-54-196-72-69.compute-1.amazonaws.com] has joined #shogun | 22:46 | |
travis-ci | it's Heiko Strathmann'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/178908511 | 22:46 |
-!- travis-ci [~travis-ci@ec2-54-196-72-69.compute-1.amazonaws.com] has left #shogun [] | 22:46 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 23:58 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 23:58 | |
--- Log closed Sat Nov 26 00:00:32 2016 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!