| --- Log opened Wed Jul 12 00:00:51 2017 | ||
| -!- travis-ci [~travis-ci@ec2-54-224-193-41.compute-1.amazonaws.com] has joined #shogun | 00:10 | |
| 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/252567639 | 00:10 | 
|---|---|---|
| -!- travis-ci [~travis-ci@ec2-54-224-193-41.compute-1.amazonaws.com] has left #shogun [] | 00:10 | |
| @sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3924 opened by micmn | 00:44 | 
| -!- travis-ci [~travis-ci@ec2-54-224-193-41.compute-1.amazonaws.com] has joined #shogun | 00:55 | |
| 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/252567639 | 00:55 | 
| -!- travis-ci [~travis-ci@ec2-54-224-193-41.compute-1.amazonaws.com] has left #shogun [] | 00:55 | |
| -shogun-buildbot:#shogun- Build nightly_all #15 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/22/builds/15 | 04:06 | |
| -!- mikeling [uid89706@gateway/web/irccloud.com/x-siynttqijkgauifg] has joined #shogun | 04:52 | |
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has joined #shogun | 09:23 | |
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has quit [Ping timeout: 260 seconds] | 10:50 | |
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has joined #shogun | 11:04 | |
| @sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3906 synchronized by MikeLing | 11:57 | 
| @sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3906 synchronized by MikeLing | 12:01 | 
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has quit [Ping timeout: 260 seconds] | 12:14 | |
| @wiking | motheruuucker!!!!! | 13:17 | 
| -!- HeikoS [~heiko@eduroam-int-pat-8-96.ucl.ac.uk] has joined #shogun | 13:18 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 13:18 | |
| @wiking | HeikoS, type in "machine learning library" into goog | 13:19 | 
| lisitsyn | yay | 13:24 | 
| Trixis | wiking: cant see shogun on the first 2 pages :( | 13:24 | 
| @wiking | Trixis, but on the top :) | 13:24 | 
| Trixis | oh its because i skipped the google suggestions thing | 13:24 | 
| Trixis | lol | 13:24 | 
| Trixis | congrats | 13:24 | 
| @wiking | good stuff :) | 13:28 | 
| @wiking | lisitsyn, if u check our analytics webpage it is obvious that we need win packages :) | 13:31 | 
| @wiking | lisitsyn, do u know anybody familiar wiht msvc? | 13:32 | 
| -!- HeikoS [~heiko@eduroam-int-pat-8-96.ucl.ac.uk] has quit [Quit: Leaving.] | 13:32 | |
| -!- besser82_ [~besser82@fedora/besser82] has joined #shogun | 14:42 | |
| -!- mode/#shogun [+o besser82_] by ChanServ | 14:42 | |
| -!- rcurtin [~ryan@knife.lugatgt.org] has quit [Ping timeout: 240 seconds] | 14:43 | |
| -!- besser82 [~besser82@fedora/besser82] has quit [Excess Flood] | 14:43 | |
| -!- besser82_ is now known as besser82 | 14:43 | |
| mikeling | wiking: Hi, for the prng things, do we want to ensure all the meta tests still the same? I mean get same result wit h the reference data | 14:44 | 
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has joined #shogun | 14:46 | |
| -!- rcurtin [~ryan@knife.lugatgt.org] has joined #shogun | 14:50 | |
| -!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun | 15:02 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 15:02 | |
| -!- rcurtin [~ryan@knife.lugatgt.org] has quit [Ping timeout: 248 seconds] | 15:08 | |
| @sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3925 opened by geektoni | 15:15 | 
| geektoni | ping HeikoS | 15:24 | 
| @HeikoS | geektoni: jojo | 15:25 | 
| geektoni | HeikoS: regarding the xvalidation thing | 15:25 | 
| @HeikoS | geektoni: yes! | 15:25 | 
| geektoni | what do you want to monitor exactly? :) | 15:25 | 
| @HeikoS | so what would be cool would be to get the results of | 15:26 | 
| @HeikoS | -each fold | 15:26 | 
| @HeikoS | -weights/parameters of the model for each fold | 15:26 | 
| @HeikoS | (e.g. MKL weights) | 15:26 | 
| @HeikoS | -average fold performance over multiple runs | 15:26 | 
| @HeikoS | etc | 15:26 | 
| @HeikoS | geektoni: and then there is a kernel section framework in the statistical_testing impl code | 15:27 | 
| @HeikoS | that has similar things, there we want to extract certain parameters when doing the selection | 15:27 | 
| geektoni | mmh I see | 15:27 | 
| geektoni | ok | 15:27 | 
| geektoni | so I'll start digging into xvalidation code | 15:27 | 
| @HeikoS | geektoni: it is quite simple | 15:28 | 
| @HeikoS | you have a CMachine | 15:28 | 
| @HeikoS | and CFeatures | 15:28 | 
| @HeikoS | and CLabels | 15:28 | 
| @HeikoS | then you split the data into say 5 folds | 15:28 | 
| @HeikoS | train on 4, test on 1 | 15:28 | 
| @HeikoS | thats a fold result (say accuracy) | 15:29 | 
| @HeikoS | then you do that for all combinations, so you get 5 fold results | 15:29 | 
| @HeikoS | (you migh want to store the fold incdices as well) | 15:29 | 
| @HeikoS | then you might repeat with a different random splitting | 15:29 | 
| @HeikoS | in addition to the say test performance, you might also want to look at say the w vector of a linear model for each fold | 15:29 | 
| @HeikoS | geektoni: but start simple: result of each fold :) | 15:29 | 
| geektoni | HeikoS: aha yeah, one thing at a time | 15:31 | 
| @HeikoS | geektoni: so the thing is | 15:31 | 
| @HeikoS | I believe you register model parameters | 15:31 | 
| @HeikoS | ? | 15:31 | 
| @HeikoS | geektoni: so the fold results might not be a parameter of x-validation | 15:31 | 
| @HeikoS | so it might need some refactoring | 15:31 | 
| @HeikoS | but that should be easy | 15:31 | 
| geektoni | actually no, there is no such thing. There is a general method which emits a value, without worrying about if the value emitted comes from a registered parameter or not. | 15:32 | 
| geektoni | HeikoS: ^ | 15:33 | 
| @HeikoS | geektoni: ah cok | 15:33 | 
| @HeikoS | then it might be easier | 15:33 | 
| @HeikoS | wanna start with the results of each fold? | 15:33 | 
| @HeikoS | that is most useful | 15:33 | 
| @HeikoS | (and we have API for that I think) | 15:33 | 
| @HeikoS | there is a class like CrossValidationResults or so | 15:34 | 
| @HeikoS | http://www.shogun-toolbox.org/api/latest/classshogun_1_1CCrossValidationResult.html | 15:34 | 
| @HeikoS | this should be deleted | 15:34 | 
| @HeikoS | and the mean should be returned | 15:34 | 
| @wiking | :> | 15:34 | 
| @wiking | muss | 15:34 | 
| @HeikoS | and if one wants more (standard deviation for example) | 15:35 | 
| @HeikoS | one registers an observer | 15:35 | 
| @HeikoS | and boonm | 15:35 | 
| -!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun | 15:35 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 15:35 | |
| @wiking | badabiiiiiiiim | 15:35 | 
| @wiking | mikeling, nono we can change it | 15:36 | 
| @wiking | mikeling, but not yet | 15:36 | 
| @wiking | mikeling, because as soon as we change to c++11 random | 15:36 | 
| @wiking | anyways many things will change | 15:36 | 
| @wiking | :) | 15:36 | 
| @HeikoS | ah man I should rewrite the xvalidation class | 15:37 | 
| geektoni | HeikoS: ok, I'll start doing things then ;) | 15:37 | 
| @HeikoS | geektoni: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/evaluation/CrossValidationMKLStorage.h | 15:38 | 
| @wiking | mikeling, here? | 15:38 | 
| @HeikoS | please please delete this! :) | 15:38 | 
| @HeikoS | this is exactly the kind of sh** we were doing before the nice observables | 15:38 | 
| @wiking | HeikoS, :D | 15:38 | 
| @HeikoS | and there is more of those | 15:39 | 
| @HeikoS | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/statistical_testing/kernelselection/KernelSelectionStrategy.h#L85 | 15:39 | 
| @HeikoS | etc etc | 15:39 | 
| @HeikoS | geektoni: please put them on some todo list | 15:39 | 
| geektoni | HeikoS: sure | 15:40 | 
| @HeikoS | wiking: whats up with some? | 15:41 | 
| @HeikoS | is there any perspective on this? | 15:41 | 
| @wiking | some is some :( | 15:41 | 
| @HeikoS | wiking: how to SWIG it? | 15:44 | 
| @HeikoS | how to get rid of raw pointers? | 15:44 | 
| mikeling | wiking: hi, yes | 15:44 | 
| @HeikoS | wiking: you have some ideas there? :D | 15:44 | 
| @HeikoS | wiking: some thing? | 15:44 | 
| @wiking | geektoni, https://travis-ci.org/shogun-toolbox/shogun/jobs/252804570#L974 | 15:44 | 
| @wiking | mooooooooooo | 15:44 | 
| Trixis | haha cross validation | 15:46 | 
| Trixis | coding grid search w/ cross validation is more work than i expected. (because custom kernels :|) | 15:46 | 
| geektoni | wiking: ahhh that's clang fault :> | 15:47 | 
| @sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3925 synchronized by geektoni | 15:50 | 
| -!- rcurtin [~ryan@knife.lugatgt.org] has joined #shogun | 16:23 | |
| -!- geektoni [c1cdd220@gateway/web/freenode/ip.193.205.210.32] has quit [Quit: Page closed] | 16:48 | |
| -!- dennyjooo [~dennyjooo@183.163.21.191] has joined #shogun | 16:49 | |
| -!- OXPHOS [92bd305b@gateway/web/freenode/ip.146.189.48.91] has joined #shogun | 17:41 | |
| -!- dennyjooo [~dennyjooo@183.163.21.191] has quit [] | 17:42 | |
| -!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 248 seconds] | 18:32 | |
| -!- sanuj [~sanuj@43.224.128.19] has joined #shogun | 19:34 | |
| -!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has joined #shogun | 21:53 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 21:53 | |
| -!- sanuj [~sanuj@43.224.128.19] has quit [Remote host closed the connection] | 23:15 | |
| -!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has quit [Quit: Leaving.] | 23:33 | |
| --- Log closed Thu Jul 13 00:00:53 2017 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!