--- Log opened Wed Oct 17 00:00:17 2012 | ||
-!- heiko1 [~heiko@host86-177-117-85.range86-177.btcentralplus.com] has left #shogun [] | 00:35 | |
-!- blackburn [~blackburn@37.61.180.136] has quit [Quit: Leaving.] | 00:36 | |
-!- heiko [~heiko@host86-177-117-85.range86-177.btcentralplus.com] has joined #shogun | 01:12 | |
shogun-buildbot | build #118 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/118 | 03:02 |
---|---|---|
shogun-buildbot | build #135 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/135 | 03:58 |
-!- adoniscik [~emre@c-67-180-103-118.hsd1.ca.comcast.net] has joined #shogun | 05:29 | |
-!- heiko [~heiko@host86-177-117-85.range86-177.btcentralplus.com] has quit [Quit: Leaving.] | 08:47 | |
-!- Dave [863ab323@gateway/web/freenode/ip.134.58.179.35] has joined #shogun | 09:06 | |
-!- Dave is now known as Guest23074 | 09:06 | |
Guest23074 | The original LibSVM has support for class weights (to compensate for unbalanced problems). Can Shogun also do that? Couldn't find anything in the documentation though.. | 09:08 |
Guest23074 | svm.set_C(c_neg,c_pos) -- nice :) | 09:29 |
Guest23074 | thx | 09:29 |
sonne|work | he | 09:31 |
sonne|work | h | 09:31 |
Guest23074 | :) | 09:47 |
-!- Guest23074 [863ab323@gateway/web/freenode/ip.134.58.179.35] has quit [Quit: Page closed] | 10:02 | |
-!- adoniscik [~emre@c-67-180-103-118.hsd1.ca.comcast.net] has quit [Ping timeout: 260 seconds] | 11:41 | |
wiking | somebody review my PR and possibly apply: https://github.com/shogun-toolbox/shogun/pull/794 | 11:47 |
sonne|work | wiking: Operators berween? | 12:51 |
wiking | sonne|work: yep | 13:07 |
wiking | but yeah i'll do some extra codes soon | 13:07 |
wiking | sonne|work: you reckon that SGSparseVectorEntry<T>* is sorted in SGSparsevector by feat_index? | 13:12 |
wiking | sonne|work: or should we add an option into SGSparseVector to do a sorting.... | 13:16 |
wiking | ? | 13:16 |
sonne|work | wiking: we should assume it is *always* sorted | 13:24 |
sonne|work | but document that we assume this! | 13:24 |
wiking | sonne|work: oke | 13:26 |
-!- Dav_ [863ab324@gateway/web/freenode/ip.134.58.179.36] has joined #shogun | 15:30 | |
-!- Dav_ [863ab324@gateway/web/freenode/ip.134.58.179.36] has quit [Client Quit] | 15:30 | |
-!- blackburn [~blackburn@83.234.54.218] has joined #shogun | 15:50 | |
-!- David [863ab324@gateway/web/freenode/ip.134.58.179.36] has joined #shogun | 16:01 | |
-!- David is now known as Guest88873 | 16:01 | |
Guest88873 | Is it correct that there's no CEvaluation subclass that allows setting specific losses for TP,TF,FP and FP ? | 16:05 |
blackburn | Guest88873: hey, could you please elaborate? | 16:06 |
blackburn | what do you mean by specific losses? | 16:06 |
Guest88873 | I' like to parameters of an SVM based on its 'loss', where e.g. a false negative is considered 10 times worse than a false positive | 16:08 |
Guest88873 | That would be: Loss = 10*FN +1*FP | 16:09 |
blackburn | Guest88873: oh got it, unfortunately that wasn't coded yet | 16:09 |
Guest88873 | ok, thx | 16:09 |
blackburn | Guest88873: we have class computing all fps,tps,fns,tns so it could be easy to combine | 16:10 |
Guest88873 | related question, could I add such a class on the run-time using Python (I think not, but I hope I'm wrong:) ) | 16:11 |
Guest88873 | No indeed, it wouldn't be that in C, but adding it in Python would be handier for experiments | 16:12 |
blackburn | Guest88873: yeah you could but it would require some C++ code again | 16:44 |
blackburn | Guest88873: we have such classes for structured risk models and kernels | 16:47 |
blackburn | it is called 'director' classes | 16:47 |
blackburn | i.e. it has a virtual method in C++ that may be overloaded by python code | 16:48 |
blackburn | and then used together with all the rest C++ code | 16:48 |
Guest88873 | sounds perfect, I'll check it out | 16:48 |
Guest88873 | thanks! | 16:48 |
blackburn | Guest88873: I'll add such class and example a little later today | 16:57 |
Guest88873 | Thanks blackburn, that would be nice, since - for the moment - I don't get which classes this concerns and how to do it.. | 16:58 |
wiking | blackburn: when u have time check out https://github.com/shogun-toolbox/shogun/pull/794 ;) | 17:08 |
blackburn | wiking: yeah wanted to | 17:28 |
blackburn | will do in 5 mins | 17:28 |
wiking | okok cool thnx | 17:33 |
wiking | oh yeah i have a warning supresser fix | 17:53 |
wiking | blackburn: http://pastebin.com/afyeSf6N | 17:54 |
wiking | anything against it? | 17:54 |
blackburn | no, looks ok | 17:54 |
blackburn | what is the warning? | 17:54 |
wiking | well clang constantly giving me the warning | 17:55 |
wiking | when i compile SpareFeatures.cpp | 17:55 |
blackburn | ohh | 17:55 |
blackburn | that won't work! | 17:55 |
wiking | why not? | 17:55 |
blackburn | I mean it is broken already | 17:55 |
blackburn | one cannot simply copy array of referenced data | 17:56 |
wiking | well it's a copy constructor :) | 17:56 |
wiking | aaah ok | 17:56 |
wiking | i see | 17:56 |
wiking | because we store reference counter there in the beginning of that array | 17:57 |
wiking | right? | 17:57 |
wiking | orig.sparse_feature_matrix | 17:57 |
wiking | or why? | 17:59 |
blackburn | I mean we have to increase ref count of each vector | 18:02 |
blackburn | here comes da fix | 18:03 |
wiking | :> | 18:03 |
blackburn | CIA passed away again | 18:04 |
wiking | :( | 18:04 |
wiking | eh | 18:05 |
wiking | new (&sparse_feature_matrix[i]) SGSparseVector<ST>(); | 18:05 |
wiking | i don't get this :D | 18:05 |
blackburn | wiking: in-place new shit | 18:06 |
wiking | so this is equivalent of sparse_feature_matrix[i] = new SGSparseVector<ST>(); | 18:06 |
wiking | ? | 18:06 |
blackburn | wiking: calloc doesn't call constructors | 18:06 |
blackburn | no | 18:06 |
blackburn | sparse_feature_matrix[i] is not a pointer it is a memory area already | 18:07 |
blackburn | wiking: new creates object using memory provided by something, here we provide where to store the object | 18:08 |
blackburn | another hardcore option to cut off a leg | 18:08 |
blackburn | we like such things here for a reason I do not understand :D | 18:09 |
wiking | :> | 18:09 |
wiking | ah ok | 18:09 |
wiking | so you basically have this "placement new" | 18:09 |
blackburn | some of shogun's code should be posted in some pornhub or whatever like that | 18:09 |
wiking | i.e. to just run a ctor on the already allocated memory | 18:09 |
blackburn | category 'extreme hardcore' | 18:09 |
blackburn | yes | 18:09 |
wiking | http://www.parashift.com/c++-faq-lite/placement-new.html | 18:10 |
wiking | :D | 18:10 |
wiking | :DDD | 18:10 |
wiking | // Line #3 (see "DANGER" below) | 18:10 |
wiking | :> | 18:10 |
wiking | lol "(if you don't know what "alignment" means, please don't use the placement new syntax). You have been warned." | 18:10 |
blackburn | // here your leg is being cutted off | 18:10 |
wiking | well yeah | 18:11 |
wiking | and is it aligned? :) | 18:12 |
wiking | on all the possible os/arch variations :) | 18:12 |
blackburn | good question | 18:12 |
wiking | as i know that osx malloc will give always give u an aligned memory | 18:12 |
blackburn | does sizeof takes alignment into account? | 18:12 |
blackburn | I hope it does | 18:12 |
wiking | but afaik that's only true for osx | 18:12 |
wiking | i remember when i had troubles with alignment (3 years ago) | 18:13 |
wiking | aligned memory by malloc was only true for osx | 18:13 |
wiking | linux and win did not do that | 18:13 |
blackburn | well actually I know it works now | 18:13 |
blackburn | on linux :) | 18:13 |
wiking | good :) | 18:14 |
blackburn | but better way is to use new[] here | 18:14 |
wiking | btw | 18:14 |
wiking | do u know why does it only work (clang_complete) | 18:14 |
wiking | if u do the include <eigen3/Eigen/Dense> | 18:14 |
wiking | ? | 18:14 |
blackburn | oh some paths issue | 18:14 |
wiking | mmm yeah i guess so but why | 18:14 |
blackburn | well | 18:14 |
wiking | i cannot get any error out of vim | 18:14 |
blackburn | /usr/include/eigen3/ | 18:15 |
blackburn | that's the path on my machine IIRC | 18:15 |
blackburn | not /usr/include/Eigen/ | 18:15 |
wiking | for me it's /opt/local/include/eigen3 | 18:15 |
blackburn | see what I mean? | 18:15 |
wiking | yeah | 18:15 |
wiking | but if you put | 18:15 |
wiking | -I/usr/include/eigen3 | 18:15 |
wiking | in your clang_complete? | 18:15 |
wiking | .clang_complete | 18:15 |
blackburn | I didn't try | 18:15 |
blackburn | did you? | 18:15 |
wiking | does it work for you | 18:15 |
wiking | because i've tried | 18:15 |
wiking | and i get NO errors | 18:16 |
blackburn | so works seamless? | 18:16 |
wiking | but still clang complete doesn't want to work with SGVector or SGSparseVector for that matter | 18:16 |
blackburn | heh | 18:16 |
blackburn | crazy | 18:16 |
wiking | could u try plz? | 18:16 |
blackburn | okay let me try | 18:16 |
shogun-buildbot | build #555 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/555 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Viktor Gal <viktor.gal@maeth.com> | 18:16 |
blackburn | boom | 18:16 |
blackburn | wiking: where to put .clang_complete? | 18:17 |
wiking | blackburn: hahaha have u seen that ? :) | 18:17 |
wiking | i mean the error | 18:17 |
wiking | http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/555/steps/test%20libshogun/logs/stdio | 18:17 |
wiking | it's some crazy shit | 18:17 |
blackburn | heiko's thing | 18:17 |
wiking | ah ok | 18:17 |
wiking | then i ignore | 18:17 |
blackburn | he got into that forest with an axe | 18:18 |
wiking | i put .clang_complete right in the root of shogun | 18:18 |
wiking | shogun/.clang_complete | 18:18 |
blackburn | in a root or src? | 18:18 |
blackburn | cause I work from src | 18:18 |
blackburn | usually | 18:18 |
wiking | -I./src | 18:18 |
wiking | -I/Users/wiking/mosek/6/tools/platform/osx64x86/h | 18:18 |
wiking | -I/opt/local/include | 18:18 |
wiking | -I/opt/local/include/eigen3 | 18:18 |
wiking | -std=c++11 | 18:18 |
wiking | -stdlib=libc++ | 18:18 |
wiking | it looks like this for me | 18:18 |
wiking | in the root | 18:18 |
wiking | not the src | 18:18 |
wiking | since this way it works as well | 18:18 |
wiking | for examples | 18:18 |
wiking | as well | 18:18 |
blackburn | I see | 18:19 |
wiking | but u can put it in ./src | 18:19 |
wiking | the only difference would be then the first -I | 18:19 |
wiking | and i guess u don't need the last two flags | 18:19 |
wiking | as that's for c++11 support in clang | 18:20 |
blackburn | wiking: works | 18:20 |
wiking | no fucking way :D | 18:20 |
wiking | so you can do a this-> | 18:20 |
wiking | in SGVector.cpp | 18:20 |
wiking | and it gives u the list | 18:20 |
wiking | w/o errors | 18:20 |
wiking | ? | 18:20 |
blackburn | this-> doesn't work | 18:20 |
blackburn | hmm | 18:21 |
wiking | hehehe | 18:21 |
wiking | then i guess it's the same for u as well | 18:21 |
blackburn | wiking: some things work | 18:21 |
wiking | in SGVector.cpp:123 | 18:21 |
wiking | add a line | 18:21 |
wiking | and put | 18:21 |
wiking | x. | 18:21 |
blackburn | no, doesn't work | 18:21 |
wiking | and u should be able to get a list for SGSparseVector<T> | 18:21 |
wiking | ah yeah | 18:21 |
wiking | and what does | 18:22 |
wiking | :copen | 18:22 |
wiking | say | 18:22 |
wiking | ? | 18:22 |
blackburn | clapack not found | 18:22 |
blackburn | :D | 18:22 |
wiking | i get | 18:22 |
wiking | the usual pattern not found | 18:22 |
wiking | but when i do :copen | 18:22 |
wiking | there's noooo error at all | 18:23 |
blackburn | heh | 18:23 |
wiking | do u have? | 18:23 |
blackburn | yeah clapack thing | 18:23 |
blackburn | may be it is the cause | 18:23 |
wiking | mmmm | 18:24 |
blackburn | let me fix that | 18:24 |
wiking | hehehe why don't i get any errors :D | 18:24 |
blackburn | sitll | 18:25 |
blackburn | still | 18:25 |
wiking | yep | 18:27 |
wiking | same here :) | 18:27 |
wiking | ok there's a python scrupt | 18:27 |
wiking | bin/cc_args.py | 18:27 |
wiking | in clang_complete | 18:27 |
wiking | it suppose to generate u the right .clang_complete | 18:27 |
wiking | :) | 18:27 |
wiking | now i'm gonna try that | 18:28 |
blackburn | hmm | 18:30 |
shogun-buildbot | build #556 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/556 blamelist: Viktor Gal <viktor.gal@maeth.com> | 18:34 |
wiking | ok i have to change that | 18:39 |
wiking | i mean in configure script | 18:39 |
wiking | streaming_vw_createcache_modular.py was broken before me right? | 18:39 |
blackburn | yeah I think so | 18:41 |
wiking | ok | 18:45 |
wiking | that python script doesn't help me at all | 18:45 |
wiking | :( | 18:45 |
shogun-buildbot | build #557 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/557 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 18:54 |
wiking | mm | 18:56 |
wiking | let g:clang_use_library = 1 | 18:56 |
wiking | the completion is faster for me | 18:56 |
wiking | i wonder how can one debug clang_complete | 18:58 |
wiking | ah ok this is why | 18:58 |
wiking | "Instead of calling the clang/clang++ tool use libclang directly. This gives | 18:58 |
wiking | access to many more clang features. Furthermore it automatically caches all | 18:59 |
wiking | includes in memory. Updates after changes in the same file will therefore be a | 18:59 |
wiking | lot faster." | 18:59 |
-!- Guest88873 [863ab324@gateway/web/freenode/ip.134.58.179.36] has quit [Ping timeout: 245 seconds] | 19:05 | |
wiking | i'll never find out what's the problem here :D | 19:25 |
-!- adoniscik [~emre@c-67-180-103-118.hsd1.ca.comcast.net] has joined #shogun | 19:30 | |
blackburn | wiking: that option is killing my vim :) | 20:15 |
wiking | blackburn: does it have python support compiled? | 20:15 |
wiking | w/o that it won't work | 20:15 |
blackburn | how can I? | 20:15 |
blackburn | do tha | 20:16 |
wiking | i don't know how u can check it if it has python support added | 20:17 |
wiking | but for the configure of vim you should add something like this | 20:17 |
wiking | --enable-pythoninterp --with-python=${prefix}/bin/python2.7 | 20:17 |
wiking | or wherever your python interpreter is | 20:17 |
wiking | can u do this in your vim | 20:18 |
wiking | :python print "helo" | 20:18 |
wiking | ? | 20:18 |
wiking | if not then pythoninterp is not enabled | 20:20 |
wiking | that's why the g:clang_use_library = 1 does not work | 20:20 |
wiking | otherwise i have no idea | 20:20 |
wiking | i had this problem but my vim didn't crash just got an error msg | 20:20 |
blackburn | I see thanks | 20:49 |
blackburn | it works | 20:49 |
wiking | heheh :) | 20:49 |
wiking | faster? :) | 20:49 |
blackburn | not yet | 20:49 |
blackburn | I meant python thing | 20:50 |
blackburn | oh works | 20:50 |
blackburn | okay let me try | 20:50 |
blackburn | argh -> doesn't work for me no matter where | 20:51 |
wiking | mmm | 20:53 |
wiking | maybe you need to set the library path | 20:53 |
wiking | where's your libclang.so? | 20:53 |
wiking | moreover, do you have one? :) | 20:53 |
blackburn | I think I have to compile shogun now | 21:01 |
wiking | no | 21:01 |
wiking | libclang.so should deal with it | 21:01 |
wiking | afaik | 21:01 |
wiking | but try | 21:01 |
wiking | i might be really wrong ehre :) | 21:01 |
blackburn | my installed headers are quite different from actual | 21:02 |
blackburn | so I have to install I think | 21:02 |
-!- mode/#shogun [+o sonney2k] by ChanServ | 21:59 | |
wiking | dude | 23:50 |
wiking | blackburn: still here? | 23:50 |
blackburn | wiking: yeao | 23:55 |
blackburn | yeap | 23:56 |
blackburn | wiking: I'm watching my parents' flight via flightradar24 :D | 23:57 |
wiking | blackburn: ehehe | 23:58 |
wiking | blackburn: i think the new latent so-svm just overexceeds any previously published solver's performance on one of the benchmark datasets :) | 23:59 |
wiking | both in training time and precision :) | 23:59 |
blackburn | O_O | 23:59 |
wiking | nyappp | 23:59 |
wiking | \o/ | 23:59 |
--- Log closed Thu Oct 18 00:00:05 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!