| --- Log opened Wed Feb 19 00:00:05 2014 | ||
| @sonney2k | wiking, maybe I just need to prefix it with \$ENV{DESTDIR} | 00:04 |
|---|---|---|
| @sonney2k | wiking, yes that fixes it | 00:06 |
| -!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 00:14 | |
| shogun-notifier- | shogun: Soeren Sonnenburg :develop * 5f9be33 / src/interfaces/r_modular/CMakeLists.txt,src/interfaces/r_static/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/5f9be33cbb93d25de0a3cc201a25393216417dfc | 00:14 |
| shogun-notifier- | shogun: fix install of r_static/modular interfaces when DESTDIR is set | 00:14 |
| -!- new_lido [~walid@41.218.177.240] has quit [Ping timeout: 252 seconds] | 00:30 | |
| -!- new_lido [~walid@41.218.175.159] has joined #shogun | 00:43 | |
| shogun-notifier- | shogun: Soeren Sonnenburg :develop * 3b51ef5 / src/interfaces/r_modular/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/3b51ef534dd1d5b8ea859bd5356e630e1752b559 | 01:21 |
| shogun-notifier- | shogun: add another missing DESTDIR | 01:21 |
| -!- HeikoS [~heiko@2.219.252.217] has joined #shogun | 01:38 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 01:38 | |
| -!- travis-ci [~travis-ci@ec2-23-20-20-64.compute-1.amazonaws.com] has joined #shogun | 02:05 | |
| travis-ci | [travis-ci] it's Soeren Sonnenburg'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/19152963 | 02:05 |
| -!- travis-ci [~travis-ci@ec2-23-20-20-64.compute-1.amazonaws.com] has left #shogun [] | 02:05 | |
| -!- alamages [~alamages@46-81-89.adsl.cyta.gr] has quit [Quit: leaving] | 02:10 | |
| @HeikoS | sonney2k: see example projects description in google doc, pls update and add what you think is needed | 02:59 |
| -!- HeikoS [~heiko@2.219.252.217] has left #shogun [] | 03:05 | |
| -!- tricksy___ [uid12458@gateway/web/irccloud.com/x-bnbwjtszpxgvbgbc] has quit [Ping timeout: 252 seconds] | 04:13 | |
| -!- tricksy___ [uid12458@gateway/web/irccloud.com/x-cptxpgonpcltlmzr] has joined #shogun | 04:14 | |
| -!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 04:21 | |
| -!- zxtx [~zv@129-79-241-148.dhcp-bl.indiana.edu] has quit [Ping timeout: 240 seconds] | 05:32 | |
| -!- zxtx [~zv@c-98-223-196-32.hsd1.in.comcast.net] has joined #shogun | 06:11 | |
| -!- new_lido [~walid@41.218.175.159] has quit [Ping timeout: 260 seconds] | 07:22 | |
| -!- new_lido [~walid@41.218.175.12] has joined #shogun | 07:35 | |
| -!- new_lido [~walid@41.218.175.12] has quit [Ping timeout: 260 seconds] | 07:41 | |
| -!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun | 07:44 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 07:44 | |
| -!- lisitsyn [~lisitsyn@80.252.20.67] has quit [Ping timeout: 245 seconds] | 07:50 | |
| -!- new_lido [~walid@41.218.176.107] has joined #shogun | 07:57 | |
| -!- new_lido [~walid@41.218.176.107] has quit [Ping timeout: 240 seconds] | 08:08 | |
| -!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Remote host closed the connection] | 08:10 | |
| -!- new_lido [~walid@41.218.177.66] has joined #shogun | 08:20 | |
| wiking | fuck | 08:36 |
| wiking | :d | 08:36 |
| -!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 08:56 | |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has joined #shogun | 09:30 | |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has quit [Client Quit] | 09:33 | |
| -!- v101 [~vodka101@115.112.95.154] has joined #shogun | 09:51 | |
| -!- v101 [~vodka101@115.112.95.154] has left #shogun [] | 09:52 | |
| -!- lambday [67157f4e@gateway/web/freenode/ip.103.21.127.78] has joined #shogun | 10:00 | |
| -!- lambday_ [67157e4e@gateway/web/freenode/ip.103.21.126.78] has joined #shogun | 10:22 | |
| -!- lambday [67157f4e@gateway/web/freenode/ip.103.21.127.78] has quit [Ping timeout: 245 seconds] | 10:23 | |
| -!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 10:26 | |
| -!- bumblebee is now known as Guest44527 | 10:26 | |
| -!- Guest44527 [~quassel@202.78.172.162] has quit [Client Quit] | 10:27 | |
| -!- bumblebee_ [~quassel@202.78.172.162] has joined #shogun | 10:27 | |
| -!- tom__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has joined #shogun | 10:45 | |
| tom__ | hi! | 10:45 |
| tom__ | I would like some advice on a multiclass unbalanced problem | 10:46 |
| tom__ | I would like to use shogun for multiclass one-vs-rest classification | 10:47 |
| -!- bumblebee_ [~quassel@202.78.172.162] has quit [Ping timeout: 260 seconds] | 10:47 | |
| tom__ | Is ts possible to assign weights for pos/neg samples for each binary svm machine inside multiclassmachine shogun framework ? | 10:48 |
| adrin | tom__: I'm not very familiar with shogun, but sometimes depending on the interface language you're using, available parameters change. | 11:03 |
| adrin | tom__: I guess nowadays python is fairly complete in terms of feature support of the underlying C++ code, if I'm not totally wrong. | 11:04 |
| -!- bumblebee_ [~quassel@202.78.175.199] has joined #shogun | 11:42 | |
| tom__ | I am using c++ interface | 11:45 |
| sonne|work | tom__: I don't think this is possible ATM | 11:46 |
| tom__ | Is it possible to use CMulticlassSVM class and set K CMKLClassification svm machine as K binary classifier ? | 11:46 |
| tom__ | I give a try and I got "train_machine is not yet implemented for SVM!" | 11:47 |
| sonne|work | tom__: but of course do-able. one could basically just set some C matrix which has a size of #examples * #classes | 11:47 |
| sonne|work | tom__: that splitting is done automatically | 11:49 |
| sonne|work | tom__: if you went that far you could just train your K machines yourself with whatever custom C settings | 11:50 |
| sonne|work | tom__: and then later put them into multiclass svm to use its apply() or do it yourself there too | 11:50 |
| tom__ | allright | 11:52 |
| tom__ | thank you | 11:54 |
| -!- tom__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has quit [Quit: Page closed] | 11:54 | |
| sonne|work | adrin: indeed there shouldn't be much difference between C++ and python_modular except that classes have no C prefix ;) | 11:59 |
| -!- tom__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has joined #shogun | 11:59 | |
| -!- bumblebee_ [~quassel@202.78.175.199] has quit [Ping timeout: 265 seconds] | 11:59 | |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has joined #shogun | 12:03 | |
| parijat | sonne|work: online? good time to talk? | 12:04 |
| sonne|work | parijat: will be off for lunch soon | 12:05 |
| sonne|work | parijat: but maybe start typing! | 12:05 |
| sonne|work | parijat: and I reply when I am back or so | 12:05 |
| parijat | sonne|work: sorry! | 12:07 |
| sonne|work | parijat: no reason to be sorry - I am happy that you ask here! | 12:08 |
| parijat | sonne|work: i wanted to ask about the pCA PR | 12:08 |
| sonne|work | I guess so | 12:08 |
| parijat | what did you mean by in-place? | 12:08 |
| parijat | i mean i can get rid of the feature_matrix_centered. will that work? | 12:08 |
| sonne|work | parijat: previously the Feature matrix was directly modified | 12:08 |
| sonne|work | parijat: that is a bit hacky for sure but did make it work in low memory conditions | 12:09 |
| parijat | ok. yes. but the feature matrix dimension is not correct | 12:09 |
| sonne|work | parijat: yes that would have to be corrected | 12:09 |
| sonne|work | but that is it | 12:10 |
| parijat | so, in any case we would need a new matrix right? | 12:10 |
| parijat | more so because the final portion of matrix needed is not continuous block in the old matrix | 12:11 |
| parijat | sonne|work: did I make sense? | 12:12 |
| sonne|work | parijat: yes but ... | 12:14 |
| sonne|work | parijat: I did 'hack' that by just moving the matrix columns to the very front and changing sizes | 12:15 |
| sonne|work | parijat: it certainly is a hack | 12:15 |
| sonne|work | but it did help back then | 12:15 |
| sonne|work | parijat: sry afk | 12:17 |
| -!- bumblebee_ [~quassel@202.78.175.199] has joined #shogun | 12:18 | |
| parijat | ok! but the final matrix to be returned should be of lesser dimension right? were you somehow able to return a submatrix | 12:19 |
| -!- HeikoS1 [~heiko@2.219.252.217] has joined #shogun | 12:29 | |
| parijat | sonne|work: whenever you are bach . :) My point was, I am directly modifying feature_matrix (m in our case) via Eigen3 interface matrix 'feature_matrix'. but the final result I am storing in a different matrix with proper target dimension which I return | 12:31 |
| parijat | HeikoS1: hello! :) | 12:32 |
| HeikoS1 | parijat: hi! | 12:33 |
| parijat | HeikoS1: i was just discussing about the in-place PCA thing with sonney2k :) | 12:33 |
| -!- alamages [~alamages@46-81-89.adsl.cyta.gr] has joined #shogun | 12:34 | |
| HeikoS1 | parijat: cool, any results on that? | 12:35 |
| parijat | HeikoS1: may be I am missing something trivial but you would need a new matrix anyways right? because the dimension of current matrix is not correct? | 12:36 |
| HeikoS1 | parijat: maybe realloc helps? since the new matrix is smaller | 12:36 |
| HeikoS1 | this should be in place | 12:36 |
| HeikoS1 | parijat: tricky thing :) it only worked before since there was this size bug :) | 12:37 |
| parijat | HeikoS1: I mean the method returns SGMatrix of smaller size. I will have to form that maatrix anyways right? | 12:37 |
| HeikoS1 | parijat: just to make sure I get everything, what are the matrices involved, how was it before, how is it now? | 12:38 |
| parijat | HeikoS1: earlier there was only this SGMatrix m obtained from features and it was returned | 12:40 |
| -!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 12:40 | |
| shogun-notifier- | shogun: lambday :develop * 4fc837f / / (10 files): https://github.com/shogun-toolbox/shogun/commit/4fc837f7b031501ac975b1c3aec3061dfaeee388 | 12:40 |
| shogun-notifier- | shogun: ssk renamed to SubsequenceStringKernel for naming consistency | 12:40 |
| shogun-notifier- | shogun: Heiko Strathmann :develop * a2e27d9 / / (10 files): https://github.com/shogun-toolbox/shogun/commit/a2e27d91c220aa6961e737b1e540e2ea53b1912a | 12:40 |
| shogun-notifier- | shogun: Merge pull request #1879 from lambday/develop | 12:40 |
| shogun-notifier- | shogun: | 12:40 |
| shogun-notifier- | shogun: ssk renamed to SubsequenceStringKernel for naming consistency | 12:40 |
| parijat | HeikoS1: it had wrong dimensions | 12:40 |
| HeikoS1 | parijat: my question is: is it possible to do this in place? | 12:40 |
| HeikoS1 | via resizing something? | 12:41 |
| HeikoS1 | or is that impossible | 12:41 |
| HeikoS1 | am let me look at the code .... | 12:41 |
| shogun-notifier- | shogun-data: Parijat Mazumdar :master * a7c153f / testsuite/python2-tests/preprocessor_pca_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/a7c153f611cfba4ffcaea68c6fbb5f0f22490466 | 12:41 |
| shogun-notifier- | shogun-data: PCA python-2 data changed | 12:41 |
| shogun-notifier- | shogun-data: Parijat Mazumdar :master * 865f0d2 / testsuite/python3-tests/preprocessor_pca_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/865f0d2594b23719d93eb375d48ea9ae11a2db90 | 12:41 |
| shogun-notifier- | shogun-data: PCA python-3 data changed | 12:41 |
| shogun-notifier- | shogun-data: Heiko Strathmann :master * 70491d2 / testsuite/ (2 files): https://github.com/shogun-toolbox/shogun-data/commit/70491d2fed0bce5561fde94a5e7b157d41c8dd2f | 12:41 |
| shogun-notifier- | shogun-data: Merge pull request #32 from mazumdarparijat/master | 12:41 |
| shogun-notifier- | shogun-data: | 12:41 |
| shogun-notifier- | shogun-data: PCA test data change | 12:41 |
| parijat | HeikoS1: I am not sure! let me check resizing of matrices | 12:42 |
| parijat | HeikoS1: unless we can resize an SGMatrix, I think no. | 12:44 |
| HeikoS1 | parijat: its possible to resize, but dont we loose things then? | 12:44 |
| parijat | HeikoS1: no we want to lose some things | 12:44 |
| HeikoS1 | parijat: looking at code currently | 12:45 |
| HeikoS1 | afk for 5 mins, ill be back soon and then check | 12:45 |
| -!- HeikoS1 [~heiko@2.219.252.217] has quit [Quit: Leaving.] | 12:45 | |
| parijat | HeikoS1: num_vectors remains same num_features changes | 12:45 |
| parijat | alright | 12:45 |
| -!- HeikoS1 [~heiko@2.219.252.217] has joined #shogun | 12:49 | |
| lambday_ | HeikoS1: hi! | 12:56 |
| lambday_ | HeikoS1: I didn't get what you mean by update the data :( | 12:57 |
| lambday_ | HeikoS1: the matrices are being returned as well as the kernel obj, right? | 12:57 |
| HeikoS1 | lambday_: hi! | 12:58 |
| HeikoS1 | lambday_: if the kernel object is returned in the function, then the integration tests will fail after a class name change | 12:59 |
| HeikoS1 | lambday_: so data needs to be updated | 12:59 |
| HeikoS1 | with the new names | 12:59 |
| lambday_ | HeikoS1: oh I didn't know that! | 13:00 |
| lambday_ | checking | 13:00 |
| lambday_ | HeikoS1: by data you mean shogun-data ? | 13:01 |
| HeikoS1 | lambday_: yes | 13:01 |
| parijat | HeikoS1: the PCA data that you have already merged will cause travis failure for other PRs? because the PCA change in not merged yet | 13:07 |
| lambday_ | HeikoS1: umm.. I am not getting - tests worked fine locally... you mean updating the submodule in buildbot? | 13:07 |
| HeikoS1 | parijat: yes :) | 13:07 |
| HeikoS1 | parijat: so better hurry up :D | 13:07 |
| HeikoS1 | lambday_: mmh weird, this should have broken things, lets just wait and i fit stays green things are fine. Have an eye on the buildbot | 13:07 |
| -!- bumblebee_ [~quassel@202.78.175.199] has quit [Ping timeout: 245 seconds] | 13:07 | |
| lambday_ | HeikoS1: alrighty | 13:08 |
| HeikoS1 | parijat: did you get my point with the realloc or doing vector wise projectsions? | 13:08 |
| parijat | HeikoS1: nope. could you explain again plz? | 13:08 |
| HeikoS1 | parijat: ok, 1 sec .... | 13:09 |
| HeikoS1 | finishing emali | 13:09 |
| parijat | HeikoS1: specifically how can I generate the final feature matrix with some of the cells of original feature matrix missing? | 13:11 |
| parijat | HeikoS1: I think I may have uunderstood your point! let me see if it works. | 13:16 |
| HeikoS1 | parijat: you can use eigen3 views to access only subsets of the matrix | 13:17 |
| HeikoS1 | but that depends on how its stored in memory | 13:17 |
| HeikoS1 | and then you can do stuff, and aferwards cut off the remaining part with realloc | 13:17 |
| parijat | HeikoS1: yes. I got your point. | 13:19 |
| HeikoS1 | parijat: if all this doesnt work, you can do it like the existing implementation, i.e. vector wise | 13:19 |
| HeikoS1 | thats same computational costs | 13:19 |
| -!- tom__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has quit [Quit: Page closed] | 13:19 | |
| HeikoS1 | (if done in serial) | 13:19 |
| parijat | yes actually I am trying to avoid looping! lets see | 13:20 |
| HeikoS1 | parijat: would be best to do this in eigen3 since it might parallelise the operation internally | 13:22 |
| HeikoS1 | so with the view, thats the best solution | 13:22 |
| HeikoS1 | I am not sure about the cutting off | 13:22 |
| HeikoS1 | parijat: the problem is that we reduce dimensions | 13:22 |
| HeikoS1 | i.e. number of rows | 13:22 |
| HeikoS1 | but we store matrices column by column | 13:23 |
| parijat | HeikoS1: yup! i am in absolutely 0 mood to let go of eigen3 | 13:23 |
| HeikoS1 | so we cannot just cut off at the end | 13:23 |
| parijat | :) | 13:23 |
| HeikoS1 | we need to remove element from within the memory block | 13:23 |
| HeikoS1 | ah thats messy | 13:23 |
| HeikoS1 | maybe transpose before | 13:23 |
| parijat | HeikoS1: yes! I am thinking of taking transpose | 13:23 |
| parijat | yeah my point | 13:23 |
| HeikoS1 | parijat: but transpose is slow | 13:23 |
| parijat | :) | 13:23 |
| HeikoS1 | no we dont want to do that | 13:23 |
| HeikoS1 | I suggest to do this then: | 13:24 |
| parijat | ok! | 13:24 |
| HeikoS1 | mmh tricky ..... | 13:25 |
| HeikoS1 | actually I dont see a way without transposing | 13:25 |
| HeikoS1 | that doesnt allocate a new matrix | 13:25 |
| HeikoS1 | ok then, transpose! | 13:25 |
| HeikoS1 | its sensible since matrix multiplication is more expensive | 13:25 |
| HeikoS1 | and allocating memory is also more expensive | 13:26 |
| HeikoS1 | ah | 13:26 |
| HeikoS1 | wait | 13:26 |
| HeikoS1 | transposing back has the same problem when the matrix is not square anymore | 13:26 |
| parijat | HeikoS1: I can also loop element by element to fill up the first elements of feature matrix and then get rid of rest. and also change num_features and num_vectors in feature | 13:27 |
| parijat | would that be less expensive? | 13:27 |
| parijat | got my point? | 13:27 |
| HeikoS1 | the "getting rid" is the problem | 13:27 |
| HeikoS1 | cutting of in place is not possible in shoguns matrix format | 13:27 |
| HeikoS1 | cutting off dimensions I mean | 13:28 |
| HeikoS1 | so you could transpose and do it then | 13:28 |
| HeikoS1 | and then transpose back | 13:28 |
| HeikoS1 | the tranposing back should work in place in fact | 13:28 |
| HeikoS1 | I was wrong earlier | 13:28 |
| HeikoS1 | sorry for the mess ;) | 13:29 |
| HeikoS1 | I think transpose is best, but make sure its in place and on the real matrix memory, not the eigen3 view | 13:29 |
| parijat | no if i line up the needed cells in the pointer to matrix and then use realloc?? | 13:29 |
| parijat | lining up will be element by element operation | 13:29 |
| HeikoS1 | 1.) transpose memory 2.)eigen3 view on subset 3.) projection view eigen3 multiplication 4.) realloc on the matrix memory to get rid of the last columns (aka dimensions since transposed) 4.) transpose back ..... all transposes should be in place | 13:30 |
| HeikoS1 | parijat: no try to avoid loops over matrix elements, eigen3 calls are better | 13:30 |
| HeikoS1 | parijat: since we can replace eigen3 with a multicore eigen3 or GPU eigen3 | 13:31 |
| HeikoS1 | without having to change our code | 13:31 |
| parijat | i think i am not able to convey my idea clearly! let me push my patch. In case you dont like it, I got the transpose thing. I will implement that | 13:32 |
| parijat | no i am not doing away with eigen3! | 13:32 |
| HeikoS1 | parijat: hehe sorry | 13:42 |
| HeikoS1 | could you explain me again? :) | 13:42 |
| HeikoS1 | or just send, I will be able to have a look later today | 13:43 |
| parijat | HeikoS1: lets try with an example. | 13:43 |
| HeikoS1 | parijat: interesting stuff this efficient linear algebra in c++, isnt it? :) | 13:43 |
| parijat | HeikoS1: yeah! :) | 13:43 |
| HeikoS1 | cant do this stuff if matlab/python/R :) | 13:43 |
| HeikoS1 | its always slower | 13:43 |
| HeikoS1 | apart from the loops | 13:43 |
| HeikoS1 | anyways, I gotta go, will be back later today, see you! | 13:43 |
| parijat | HeikoS1: yeah sure. I will push code for your perusal | 13:44 |
| parijat | bbye! | 13:44 |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has quit [Quit: Page closed] | 13:47 | |
| -!- bumblebee_ [~quassel@202.78.175.199] has joined #shogun | 13:57 | |
| -!- bumblebee_ [~quassel@202.78.175.199] has quit [Client Quit] | 13:57 | |
| -!- bumblebee__ [~quassel@202.78.175.199] has joined #shogun | 13:58 | |
| -!- HeikoS1 [~heiko@2.219.252.217] has quit [Quit: Leaving.] | 13:59 | |
| -!- new_lido [~walid@41.218.177.66] has quit [Ping timeout: 252 seconds] | 14:24 | |
| -!- new_lido [~walid@41.218.172.173] has joined #shogun | 14:37 | |
| wiking | btw | 14:44 |
| wiking | this site is the awesomeness factor 2.0: http://techcrunch.com/2014/01/26/google-deepmind/ | 14:44 |
| wiking | no | 14:44 |
| wiking | http://deepmind.com/ | 14:44 |
| wiking | shogun refactor should look like this imo | 14:44 |
| sonne|work | wiking: you mean no content? | 14:56 |
| sonne|work | totally awesome! | 14:56 |
| wiking | i was refering to infinite canvas | 14:58 |
| -!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has quit [Quit: Leaving.] | 15:04 | |
| -!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has joined #shogun | 15:04 | |
| lisitsyn | wiking: yeah the site is awesome though no info as sonne|work said :D | 15:16 |
| wiking | lisitsyn: was never talking about the content | 15:17 |
| -!- HeikoS1 [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has joined #shogun | 15:21 | |
| lambday_ | HeikoS1: hey.. you were right.. changing name did break the integration test.. I had the something wrong at my side | 15:25 |
| lambday_ | HeikoS1: tested locally again.. failing tests | 15:26 |
| HeikoS1 | lambday_: thats good news then, all as expected :) | 15:26 |
| lambday_ | HeikoS1: btw I started working on the multitask notebook.. reading malsar doc as well... thinking of including some real world example... | 15:27 |
| HeikoS1 | lambday_: man thats great! | 15:27 |
| lisitsyn | lambday_: hmm I have some draft of this notebook | 15:27 |
| HeikoS1 | your other notebook was great, so looking forward to see it | 15:27 |
| lisitsyn | if you could continue with it - would be fantastic :D | 15:28 |
| HeikoS1 | lambday_, lisitsyn writing those notebooks is a great thing to practice writing about ML btw, it really helps for later | 15:28 |
| HeikoS1 | lisitsyn: speaking of writing, we should write the GSoC veteran blog post | 15:28 |
| lambday_ | yep! | 15:28 |
| HeikoS1 | lisitsyn: any suggestions? | 15:28 |
| HeikoS1 | we should make something catchy, like a dialog between us or a fake interview or so | 15:28 |
| lisitsyn | HeikoS1: what time works for you? | 15:28 |
| HeikoS1 | lisitsyn: what time? | 15:28 |
| lisitsyn | HeikoS1: to take care of this :) | 15:29 |
| HeikoS1 | weekend? | 15:29 |
| lisitsyn | yeah! | 15:29 |
| HeikoS1 | lisitsyn: saturday afternoon? | 15:29 |
| lisitsyn | good | 15:29 |
| lisitsyn | should be possible | 15:29 |
| lambday_ | lisitsyn: could you please share with me what you already have? or.. you wanna finish this? | 15:29 |
| lisitsyn | lambday_: I want but have time pressure | 15:30 |
| -!- lambday_ is now known as lambday | 15:30 | |
| lambday | lisitsyn: umm... you'll do it best - its all your code... | 15:30 |
| lisitsyn | lambday: sure | 15:30 |
| HeikoS1 | lisitsyn: btw we desperately need a multiclass notebook | 15:31 |
| HeikoS1 | lisitsyn: or extend the existing one | 15:31 |
| HeikoS1 | with all the OvO and accuracy stuff | 15:31 |
| HeikoS1 | like the guy in the mail asked | 15:31 |
| HeikoS1 | lisitsyn: you should do that! you know best | 15:31 |
| HeikoS1 | lisitsyn: and then you should also document all classes :) | 15:31 |
| lisitsyn | HeikoS1: I'll answer the mail tonight | 15:31 |
| lambday | HeikoS1: umm should I go ahead with string kernel noteboook then? or focus on gsoc topics - lol I am all confused! :D | 15:31 |
| lambday | sonney2k: I can create shogun tag on stack overflow I guess! | 15:32 |
| HeikoS1 | lambday: what ever you want! | 15:33 |
| HeikoS1 | lambday: you can start doing gsoc stuff when we are accepted :) | 15:33 |
| lambday | HeikoS1: alright then - I'll start the string kernel one then - this one is gonna be huge! | 15:34 |
| lisitsyn | HeikoS1: lambday: so I mean | 15:34 |
| lisitsyn | we should not duplicate the work on multitask notebook | 15:34 |
| lambday | lisitsyn: yeah totally makes sense | 15:34 |
| HeikoS1 | lisitsyn: just put the notebook into your git and then update there, then lambday can join forces if he wants | 15:34 |
| lisitsyn | ok | 15:35 |
| lambday | yeah that sounds good :) | 15:35 |
| * lambday goes back reading stuffs | 15:35 | |
| HeikoS1 | lisitsyn: https://github.com/shogun-toolbox/shogun/issues/1880 | 15:37 |
| lisitsyn | HeikoS1: haha another duty ;) | 15:37 |
| lisitsyn | alright | 15:37 |
| -!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 15:41 | |
| lambday | HeikoS1: lisitsyn: can the notebook documentation page have titles for the notebooks and then one can put a toc sort of thing at the top so that its easier to browse? | 15:42 |
| lambday | like the ideas page | 15:43 |
| HeikoS1 | lambday: great idea! could you hunt fernanod down on this? like pusing them to do it? :) I have not idea how the website works | 15:44 |
| lambday | HeikoS1: neither do I or I would have done it! I can create an issue though and tag him :P | 15:45 |
| HeikoS1 | lambday: thats good! but not enough to make it happen ;) | 15:46 |
| lambday | HeikoS1: lol okay I will catch him at irc :) | 15:47 |
| HeikoS1 | lambday: every day :) | 15:47 |
| lambday | the shogun-web repo is not the one that is used currently I guess :-/ | 15:48 |
| lambday | last updated a month ago | 15:48 |
| -!- new_lido [~walid@41.218.172.173] has quit [Quit: Leaving] | 16:03 | |
| -!- lambday [67157e4e@gateway/web/freenode/ip.103.21.126.78] has quit [Ping timeout: 245 seconds] | 16:03 | |
| -!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 16:05 | |
| -!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has joined #shogun | 16:05 | |
| sonne|work | lambday: it is the one | 16:25 |
| sonne|work | lambday: could you please add the string kernel notebook? | 16:26 |
| sonne|work | I mean write one :D | 16:26 |
| sonne|work | or start at least | 16:26 |
| lambday | sonne|work: hehe.. yeah I will asap.. | 16:27 |
| lambday | sonne|work: I found your thesis somewhere that had all the references to most of the string kernels that we have | 16:27 |
| sonne|work | lambday: I can give you the source code for string kernels if you do more than 1 kernel :D | 16:27 |
| sonne|work | lambday: the latex code I mean :D | 16:27 |
| lambday | sonne|work: that would be really helpful :) I will also get to know about them :D | 16:28 |
| lambday | sonne|work: oh btw do you agree that having a toc for the notebooks in the documentation page is a good idea? | 16:29 |
| lambday | imo mini desc is helpful but toc would help largely in browsing | 16:29 |
| lambday | brb after dinner :) | 16:39 |
| sonne|work | lambday: sure! | 16:39 |
| sonne|work | lambday: just do it | 16:40 |
| sonne|work | lambday: and yes the code is up-to-date :D | 16:40 |
| lambday | sonne|work: lol its django and I don't know anything - will try :D | 16:40 |
| sonne|work | lambday: hehe then better don't try it :) | 16:40 |
| lambday | sonne|work: hehe! :D | 16:41 |
| lambday | brb :) | 16:41 |
| -!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has quit [Quit: Leaving.] | 16:42 | |
| -!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has joined #shogun | 16:42 | |
| -!- HeikoS1 [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 16:58 | |
| -!- bumblebee__ [~quassel@202.78.175.199] has quit [Ping timeout: 252 seconds] | 16:58 | |
| -!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:07 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 17:07 | |
| -!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Client Quit] | 17:08 | |
| -!- lisitsyn [~lisitsyn@80.252.20.67] has joined #shogun | 17:24 | |
| -!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has quit [Ping timeout: 245 seconds] | 17:29 | |
| -!- alamages [~alamages@46-81-89.adsl.cyta.gr] has quit [Ping timeout: 260 seconds] | 18:42 | |
| -!- HeikoS [~heiko@nat-184-104.internal.eduroam.ucl.ac.uk] has joined #shogun | 18:42 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 18:42 | |
| -!- [paladin] [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun | 18:48 | |
| -!- alamages [~alamages@213.249.12.26] has joined #shogun | 18:48 | |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has joined #shogun | 19:20 | |
| parijat | HeikoS: I have pushed in the new code https://github.com/shogun-toolbox/shogun/pull/1874 | 19:22 |
| parijat | HeikoS: I believe this is better than taking transpose 2 times | 19:23 |
| parijat | :) | 19:23 |
| -!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun | 19:42 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 19:42 | |
| parijat | iglesiasg: hello! hows life? :) | 19:47 |
| @iglesiasg | hey parijat | 19:48 |
| @iglesiasg | life's good | 19:48 |
| @iglesiasg | haha | 19:48 |
| @iglesiasg | parijat, what about you? | 19:48 |
| parijat | iglesiasg: good so far! :D | 19:48 |
| @iglesiasg | nice to hear | 19:48 |
| parijat | iglesiasg: i have been working on fixin the pca bug. | 19:48 |
| @iglesiasg | parijat, yeah, I have seen the commits and the comments from HeikoS and sonney2k | 19:49 |
| parijat | iglesiasg: running into troubles now n then! lets see how it turns out finally :) | 19:49 |
| @iglesiasg | it will be good ;) | 19:50 |
| parijat | iglesiasg: hoping so. ;) | 19:50 |
| wiking | iglesiasg: hey man | 20:09 |
| wiking | iglesiasg: so if i'm good then i should just copy-paste the html in your repo right? | 20:09 |
| -!- [paladin] [73f91219@gateway/web/freenode/ip.115.249.18.25] has left #shogun [] | 20:09 | |
| wiking | iglesiasg: i'll send you soon a PR with the other 2 projects and then i'll do the copy-paste... | 20:09 |
| -!- HeikoS [~heiko@nat-184-104.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 272 seconds] | 20:10 | |
| -!- HeikoS [~heiko@nat-184-104.internal.eduroam.ucl.ac.uk] has joined #shogun | 20:11 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 20:11 | |
| -!- HeikoS [~heiko@nat-184-104.internal.eduroam.ucl.ac.uk] has quit [Remote host closed the connection] | 20:11 | |
| @iglesiasg | wiking, cool | 20:11 |
| @iglesiasg | wiking, sure do that, get the file and copy-paste | 20:12 |
| parijat | iglesiasg: I am getting this wierd exception in travis EXCEPTION 'module' object has no attribute '_Unpickler' | 20:17 |
| parijat | iglesiasg: any clue about it? | 20:17 |
| @iglesiasg | parijat, maybe it is an issue with serialization in Python? | 20:20 |
| parijat | sorry whats that? | 20:20 |
| @iglesiasg | parijat, serialization? | 20:21 |
| parijat | iglesiasg: yes | 20:22 |
| @iglesiasg | parijat, when the state of an object or a data structure in general is stored so that it can be used outside the program | 20:22 |
| @iglesiasg | parijat, maybe because you want to recover that state later, or maybe because you want to send over the network, etc | 20:23 |
| @iglesiasg | the pickle module is used in Python for serialization, that's why I guessed so from Unpickler | 20:23 |
| parijat | iglesiasg: there might be a problem with the format in which output expected is stored? | 20:23 |
| parijat | iglesiasg: thats why I might be getting this exception? | 20:24 |
| @iglesiasg | parijat, I am not sure at all. Have you detected when was the first time the error happen? | 20:24 |
| @iglesiasg | send me the link to the travis log you mean also please | 20:24 |
| parijat | iglesiasg: here https://s3.amazonaws.com/archive.travis-ci.org/jobs/19203009/log.txt | 20:25 |
| parijat | iglesiasg: test 173 | 20:25 |
| @iglesiasg | parijat, does it happen to you locally? Is it with python2 or python3? | 20:26 |
| parijat | iglesiasg: yes! with python 2. I thought it might be some random thing | 20:27 |
| @iglesiasg | parijat, I can restart the travis build and see if next time happens. Is it in your opened PR? | 20:28 |
| parijat | yes! | 20:29 |
| parijat | iglesiasg: thanks. | 20:29 |
| @iglesiasg | parijat, although I think it might be something else... mm | 20:30 |
| @iglesiasg | parijat, it is an integration test what is failing I think | 20:30 |
| @iglesiasg | integration, indeed yeah | 20:30 |
| parijat | parijat: yup its integration test. | 20:31 |
| wiking | iglesiasg: pr sent | 20:31 |
| @iglesiasg | parijat, have you tried running that test locally? | 20:31 |
| @iglesiasg | wiking, done! | 20:31 |
| wiking | cool i'll now copy-paste | 20:32 |
| @iglesiasg | wiking, sure, thanks! | 20:32 |
| wiking | hope i dont delete some ideas :) | 20:32 |
| @iglesiasg | haha man that's been such a mistake from me, sorry really | 20:32 |
| wiking | iglesiasg: mmm do u know that not the full html is in the repo? | 20:33 |
| @iglesiasg | but it didn't cross my mind at all | 20:33 |
| @iglesiasg | wiking, what? it is it is | 20:33 |
| @iglesiasg | the one you have to copy for the ideas is | 20:33 |
| wiking | </tbody> | 20:33 |
| wiking | </table> | 20:33 |
| wiking | is missing | 20:33 |
| wiking | oh wait | 20:33 |
| wiking | yeah | 20:33 |
| wiking | that part is missing | 20:33 |
| wiking | the closing tags i've mentioned | 20:34 |
| wiking | no worries i'll add it like that | 20:34 |
| @iglesiasg | ok | 20:34 |
| wiking | updated | 20:34 |
| wiking | http://shogun-toolbox.org/page/Events/gsoc2014_ideas | 20:34 |
| wiking | but i have a backup of the prev one | 20:34 |
| wiking | if needed | 20:34 |
| @iglesiasg | sounds good | 20:34 |
| -!- parijat [671b082a@gateway/web/freenode/ip.103.27.8.42] has quit [Ping timeout: 245 seconds] | 20:35 | |
| -!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has quit [Quit: Leaving] | 20:48 | |
| alamages | /clear | 20:49 |
| -!- alamages [~alamages@213.249.12.26] has quit [Quit: leaving] | 20:53 | |
| -!- new_lido [~walid@41.218.176.133] has joined #shogun | 20:58 | |
| -!- parijat [6adf654d@gateway/web/freenode/ip.106.223.101.77] has joined #shogun | 21:08 | |
| -!- parijat [6adf654d@gateway/web/freenode/ip.106.223.101.77] has quit [Client Quit] | 21:11 | |
| -!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has joined #shogun | 22:13 | |
| -!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has quit [Ping timeout: 245 seconds] | 22:34 | |
| -!- alamages [~alamages@213.249.12.26] has joined #shogun | 22:37 | |
| -!- iglesiasg [~iglesiasg@82.74.224.167] has joined #shogun | 23:00 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 23:00 | |
| --- Log closed Thu Feb 20 00:00:06 2014 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!