| --- Log opened Mon Jun 20 00:00:01 2011 | ||
| blackburn | I really hate when variables is called qdsad_dsdwqrtwe_312 | 00:00 |
|---|---|---|
| blackburn | I guess alesis came to Lithuania, cause he finished passing exams | 00:01 |
| @sonney2k | dammed naming clash | 00:03 |
| @sonney2k | R has a function called length | 00:03 |
| @sonney2k | so that clashes with SGVector definition | 00:03 |
| @sonney2k | length | 00:03 |
| @sonney2k | I guess I will rename it to either vector_length or vlength | 00:03 |
| @sonney2k | blackburn, ^ any preferences | 00:04 |
| blackburn | shorter! | 00:04 |
| @sonney2k | vlength then it is | 00:04 |
| blackburn | may be even vlen | 00:04 |
| blackburn | we will use it in loops and so on | 00:04 |
| @sonney2k | ok then vlen | 00:04 |
| blackburn | e.g. for (i=0;i<mean_vector.vector_length;i++) | 00:05 |
| blackburn | classifier/svm is ok | 00:05 |
| @sonney2k | you can always make temporaty variables | 00:05 |
| @sonney2k | dammed! now I have to change half of shogun because of this | 00:05 |
| blackburn | hahaha it will be pretty hard | 00:06 |
| blackburn | ::std::vector<float64_t> & weights, const int32_t index | 00:08 |
| blackburn | O_O | 00:08 |
| blackburn | std in shogun.. | 00:09 |
| blackburn | classifier seems to be done | 00:09 |
| blackburn | only KNN required changes | 00:09 |
| blackburn | regression | 00:10 |
| @sonney2k | I will only get this thing to compile and then better stop - I am getting too sleepy | 00:10 |
| @sonney2k | 'only' that is | 00:10 |
| @sonney2k | I guess I will have to touch every file... | 00:10 |
| blackburn | can you avoid this naming clash? | 00:11 |
| @sonney2k | how? | 00:11 |
| @sonney2k | I think you will be able to safely merge | 00:12 |
| blackburn | I don't know, so asking | 00:12 |
| blackburn | safely merge what? | 00:12 |
| blackburn | sonney2k: may be we will just remove typemaps in our 'packages' one-by-another? | 00:13 |
| blackburn | it will be too hard to resolve | 00:14 |
| blackburn | for example now I will remove that from classifier | 00:15 |
| blackburn | and check classifier.. | 00:15 |
| @sonney2k | still compiling... | 00:18 |
| blackburn | tried to use vim :D | 00:19 |
| blackburn | hey I can't quit | 00:19 |
| blackburn | don't react on anything | 00:19 |
| @sonney2k | *lol* | 00:21 |
| @sonney2k | try esc then : then q | 00:22 |
| blackburn | tried, it did nothing :) | 00:22 |
| @sonney2k | ok static infterfaces compile | 00:22 |
| @sonney2k | then esc twice :) | 00:22 |
| @sonney2k | or ctrl+c | 00:22 |
| @sonney2k | octave_modular too | 00:23 |
| @sonney2k | python_modular too | 00:23 |
| blackburn | sonney2k: why MultiClassSVM doesn't have apply(CFeatures)?? | 00:23 |
| @sonney2k | r-Modular too | 00:23 |
| blackburn | sonney2k: how to exit with save? | 00:24 |
| @sonney2k | I thought that kernelmachine has this and is calling set_feautres + apply? | 00:24 |
| @sonney2k | :x | 00:24 |
| blackburn | NotImplementedError: Wrong number of arguments for overloaded function 'MultiClassSVM_apply'. | 00:24 |
| blackburn | Possible C/C++ prototypes are: | 00:24 |
| blackburn | apply(shogun::CMultiClassSVM *) | 00:24 |
| blackburn | apply(shogun::CMultiClassSVM *,int32_t) | 00:24 |
| blackburn | I just made apply() | 00:24 |
| @sonney2k | .......java_modular all good | 00:24 |
| @sonney2k | blackburn, wierd | 00:25 |
| @sonney2k | cshard died | 00:25 |
| blackburn | sonney2k: same thing with larank | 00:25 |
| @sonney2k | blackburn, larank is also multiclass | 00:26 |
| blackburn | oh! found one thing to be sgvectorized | 00:27 |
| blackburn | oh I'm going to check all svms now | 00:32 |
| blackburn | fffffffff | 00:32 |
| blackburn | FUUUUUUUUUUUU | 00:32 |
| @sonney2k | ffff here too | 00:34 |
| @sonney2k | some csharp stuff is just broken... | 00:34 |
| -!- blackburn [~blackburn@31.28.40.202] has quit [Ping timeout: 255 seconds] | 00:37 | |
| -!- blackburn [~blackburn@31.28.40.202] has joined #shogun | 00:38 | |
| blackburn | wifi lost connection FFUUUU | 00:38 |
| @sonney2k | blackburn, merge please | 00:42 |
| CIA-32 | shogun: Soeren Sonnenburg master * r9c9e74a / (29 files in 14 dirs): get rid of legacy T* x, int32_t len / or T** typemaps and transition everything to SGMatrix and friends - http://bit.ly/inZrJ7 | 00:42 |
| blackburn | is it compiling? | 00:43 |
| @sonney2k | yes | 00:43 |
| @sonney2k | for all $LANG | 00:43 |
| blackburn | nice | 00:43 |
| blackburn | will check tests :) | 00:43 |
| @sonney2k | not much works | 00:43 |
| blackburn | now I fixed some svms | 00:43 |
| @sonney2k | simply typemaps are missing | 00:43 |
| @sonney2k | err SGVector stuff | 00:43 |
| @sonney2k | typemaps are there | 00:44 |
| blackburn | okay now only larank is ERROR | 00:44 |
| blackburn | will merge now | 00:44 |
| @sonney2k | blackburn, please try to merge and see if you get not too many conflicts | 00:44 |
| @sonney2k | better commit before attempting | 00:44 |
| @sonney2k | ahh you have too anyway with git | 00:44 |
| blackburn | there wouldn't be any | 00:45 |
| @sonney2k | no conflicts? | 00:45 |
| blackburn | no, everything is ok | 00:45 |
| @sonney2k | great | 00:46 |
| blackburn | compiling | 00:46 |
| blackburn | will see what's up with tests now | 00:46 |
| @sonney2k | I will have to go to bed now (I am falling asleep with my eyes open) | 00:46 |
| blackburn | sonney2k: bam | 00:46 |
| blackburn | classifier/svm/SVM.cpp:336: error: ‘struct shogun::SGVector<double>’ has no member named ‘length’ | 00:46 |
| @sonney2k | so please send a pull request with what you have | 00:46 |
| blackburn | have you renamed? | 00:46 |
| @sonney2k | vlen | 00:46 |
| @sonney2k | :) | 00:46 |
| blackburn | already?? | 00:46 |
| @sonney2k | sure - I had too | 00:47 |
| @sonney2k | otherwise r_modular was not compiling | 00:47 |
| blackburn | wow | 00:47 |
| @sonney2k | that is why the above push is so big | 00:47 |
| blackburn | okay | 00:47 |
| blackburn | good night :) | 00:47 |
| blackburn | i will make a pull request | 00:48 |
| @sonney2k | blackburn, cu and thanks for your help - I think the transition will be finished end of this week latest :) | 00:55 |
| blackburn | I hope | 00:55 |
| blackburn | see yo | 00:55 |
| * sonney2k is closing his eyes | 00:56 | |
| blackburn | and falls down the floor | 00:56 |
| -!- blackburn [~blackburn@31.28.40.202] has quit [Quit: Leaving.] | 01:44 | |
| CIA-32 | shogun: Sergey Lisitsyn master * rbdc3e2d / (31 files in 15 dirs): Merge branch 'master' of https://github.com/shogun-toolbox/shogun - http://bit.ly/jwSHoG | 10:22 |
| CIA-32 | shogun: Sergey Lisitsyn master * rb95d2fa / (12 files in 5 dirs): Various fixes - http://bit.ly/kElqzR | 10:22 |
| -!- heiko [~heiko@infole-06.uni-duisburg.de] has joined #shogun | 10:25 | |
| -!- blackburn [~blackburn@31.28.40.202] has joined #shogun | 10:27 | |
| @sonney2k | blackburn, btw you have to modify libshogunui code too when you do that stuff | 10:28 |
| @sonney2k | I am fixing compilation now :) | 10:28 |
| blackburn | ooooooh I want to die | 10:29 |
| @sonney2k | blackburn, I cannot help you with that ;-) | 10:29 |
| @sonney2k | a shogun with appropriate sword might come handy here | 10:29 |
| blackburn | refactoring refactoring refactoring | 10:30 |
| heiko | hey | 10:30 |
| blackburn | hello | 10:30 |
| heiko | with the refactoring SGVector/SGMatrix stuff | 10:30 |
| heiko | sometimes one has to be carefull | 10:31 |
| heiko | a setter for featre_matrix in SimpleFeatures has to apply the new matrix sizes to addtional variables | 10:31 |
| blackburn | I'm going to check everything at CustomKernel and Kernel | 10:31 |
| blackburn | could somebody fix features? | 10:32 |
| @sonney2k | heiko, you are right | 10:32 |
| @sonney2k | but it is just non-trivial to do it all 100% in the first iteration | 10:32 |
| heiko | yes of course :) | 10:32 |
| @sonney2k | I think we have no choice but to go through some rough waters now | 10:32 |
| heiko | I just wanted to note, I will fix it | 10:32 |
| @sonney2k | also we still have the issue with get_ / set_ sometimes copying | 10:32 |
| @sonney2k | and we never want to copy if possible | 10:33 |
| heiko | btw why are there two variables for the feature matrix in simple features? | 10:33 |
| @sonney2k | heiko, where? | 10:34 |
| @sonney2k | heiko, btw in SimpleFeatures we no longer need num_vectors / num_features | 10:35 |
| heiko | but only the matrix dimensions? | 10:35 |
| @sonney2k | that would then all be in a SGMatrix<T> feature_matrix; object | 10:35 |
| heiko | well, ok, I meant these | 10:35 |
| heiko | currently there are | 10:35 |
| heiko | int32_t num_vectors; | 10:35 |
| heiko | AND | 10:36 |
| heiko | int32_t feature_matrix_num_vectors; | 10:36 |
| heiko | Then I will apply everything for that class | 10:36 |
| @sonney2k | simplefeatures don't necessarily need a feature matrix | 10:37 |
| @sonney2k | they can be just vectors read from disc or so | 10:37 |
| heiko | ah ok | 10:37 |
| heiko | so these double number stuff remains | 10:38 |
| * sonney2k continues fixing libshogunui | 10:40 | |
| blackburn | sonney2k: is it ok to not copy kernel matrices at CustomKernel | 10:44 |
| blackburn | ? | 10:44 |
| @sonney2k | heiko, thanks for the small patch - much more digestile | 10:44 |
| @sonney2k | blackburn, yes *never* copy | 10:44 |
| heiko | :) | 10:44 |
| @sonney2k | digestible | 10:44 |
| CIA-32 | shogun: Soeren Sonnenburg master * r1327d08 / (2 files): T** transition for kernels in libshogunui - http://bit.ly/lKaUGw | 10:45 |
| CIA-32 | shogun: Soeren Sonnenburg master * r532b5df / (7 files in 2 dirs): Merge branch 'master' of git://github.com/karlnapf/shogun - http://bit.ly/mna4Sc | 10:46 |
| blackburn | ZZZE | 10:47 |
| blackburn | sonney2k: I can't never copy cause there are float32_t kernel matrix | 10:48 |
| @sonney2k | return SGMatrix<float32_t> then | 10:49 |
| @sonney2k | and in float64_t case copy | 10:49 |
| @sonney2k | blackburn, please add a free flag | 10:49 |
| @sonney2k | and set it | 10:50 |
| blackburn | free flag?? | 10:50 |
| @sonney2k | blackburn, recall that we discussed the best way to deal with get_feature_matrix() - that returns a copy of the object? | 10:52 |
| blackburn | yes but I can't do two things at the moment :) | 10:52 |
| @sonney2k | blackburn, ok let me add it | 10:52 |
| blackburn | i've sticked at kernels. | 10:52 |
| @sonney2k | blackburn, what is your preferred name? | 10:53 |
| @sonney2k | do_free | 10:53 |
| @sonney2k | or free_matrix | 10:53 |
| blackburn | do_free ok | 10:53 |
| @sonney2k | or free_vector | 10:53 |
| @sonney2k | do_free then ok | 10:53 |
| heiko | sonney2k, what about the preprocs and subset? | 10:53 |
| blackburn | bool set_full_kernel_matrix_from_full( | 10:54 |
| blackburn | SGMatrix<float32_t> full_kernel_matrix) | 10:54 |
| blackburn | bool set_full_kernel_matrix_from_full( | 10:54 |
| blackburn | SGMatrix<float65_t> full_kernel_matrix) | 10:54 |
| blackburn | hahah | 10:54 |
| blackburn | 64 | 10:54 |
| blackburn | so wiil it work? | 10:54 |
| heiko | i changed apply_preproc to apply the preprocs to ALL features instead only to subset, is that ok? | 10:54 |
| blackburn | apply_preproc? do we have it?? | 10:55 |
| blackburn | I was fixing it to apply_preprocessor | 10:55 |
| heiko | SimpleFeatures | 10:55 |
| blackburn | FFFFFFUUUUUUUu we better rewrite all shogun | 10:56 |
| heiko | hehe :) relax, this just takes some time | 10:56 |
| blackburn | it will take infinite time | 10:57 |
| @sonney2k | blackburn, an infinitely long 1 week :) | 10:58 |
| blackburn | btw I will stop doing it today because I have last exam on 22nd | 10:58 |
| @sonney2k | blackburn, yes good idea | 10:59 |
| @sonney2k | exam is more important | 10:59 |
| heiko | which subject? | 11:00 |
| blackburn | numerical methods | 11:01 |
| heiko | ok :) success! I did this one last semester | 11:05 |
| @bettyboo | smart | 11:05 |
| @sonney2k | blackburn, https://gist.github.com/1035341 | 11:06 |
| @sonney2k | like this? | 11:06 |
| blackburn | do we really need do_free? | 11:06 |
| blackburn | I mean if we call free_vector we of course want it to free | 11:07 |
| @sonney2k | blackburn, I am open for suggestions | 11:07 |
| @sonney2k | blackburn, how do you know that it needs to be freed? | 11:07 |
| blackburn | no idea | 11:09 |
| @sonney2k | the only way I can think of is to mark a vector/matrix that it can be freed | 11:10 |
| @sonney2k | hence the do_free | 11:10 |
| @sonney2k | I mean we could have a different type like SGMatrixGC | 11:10 |
| @sonney2k | that does the garbage collection | 11:10 |
| @sonney2k | but hmhhh | 11:10 |
| @sonney2k | blackburn, any further thoughts? | 11:12 |
| blackburn | don't know.. | 11:13 |
| blackburn | sonney2k: why delete[] matrix.matrix will kaboom? | 11:20 |
| @sonney2k | no kaboom | 11:21 |
| blackburn | ok | 11:22 |
| blackburn | almost done with custom kernel | 11:29 |
| blackburn | yuppy! | 11:31 |
| blackburn | sonney2k: something is wrong with all string features | 11:31 |
| @sonney2k | :) | 11:31 |
| @sonney2k | blackburn, so shall I commit this change | 11:32 |
| blackburn | sonney2k: which? flag? | 11:32 |
| @sonney2k | blackburn, the do_free thing I meant | 11:33 |
| @sonney2k | yes | 11:33 |
| blackburn | why not :) | 11:34 |
| @bettyboo | :Q | 11:34 |
| CIA-32 | shogun: Soeren Sonnenburg master * r84caa72 / src/libshogun/lib/DataType.h : introduce a do_free flag in sgvector/sgmatrix - http://bit.ly/ioBKfm | 11:34 |
| blackburn | sonney2k: pull request with custom kernel is in | 11:34 |
| blackburn | test passing now | 11:35 |
| blackburn | don't sure everything is right :D but test is passed | 11:35 |
| blackburn | distribution tests 0/4 | 11:36 |
| blackburn | features FFFFUUUUUU | 11:37 |
| blackburn | kernels aren't working very well too hehe | 11:37 |
| blackburn | running mkl_multiclass_modular.py ..*** glibc detected *** python: double free or corruption (out): 0x095c59f8 *** | 11:37 |
| blackburn | oh... we are far away from 1.0.0 :D | 11:37 |
| blackburn | fixing clustering | 11:40 |
| heiko | hej blackburn | 12:02 |
| blackburn | I | 12:02 |
| heiko | CDistance, line 228, there is a do_free argument missing | 12:02 |
| heiko | I dont know whether it should be true or false there | 12:03 |
| blackburn | it should be freed I think | 12:04 |
| blackburn | it returns a copy of matrix | 12:04 |
| heiko | yes, ok | 12:05 |
| heiko | will you fix it? does not compile | 12:06 |
| @sonney2k | blackburn, we are in a transition not at the end of a release :) | 12:06 |
| @bettyboo | *g*! sonney2k | 12:06 |
| @sonney2k | blackburn, you can now use m_parameters->add_matrix(&kmatrix, "kmatrix", ...) heiko added this functionality | 12:09 |
| blackburn | hehe | 12:09 |
| heiko | actually without _matrix | 12:09 |
| heiko | simply add | 12:09 |
| @sonney2k | heiko, btw why do you include <string> ? | 12:10 |
| @sonney2k | what is it and can we avoid it? | 12:11 |
| heiko | did it because of memcpy, but yes, we can avoid it. I forgot to remove it | 12:12 |
| heiko | will do | 12:12 |
| blackburn | fixed | 12:12 |
| @sonney2k | heiko, but that is string.h | 12:13 |
| @sonney2k | not string | 12:13 |
| heiko | oh, ok | 12:13 |
| heiko | works here, sorry I did not really put attention on this | 12:13 |
| @sonney2k | heiko, no worries | 12:13 |
| heiko | :) | 12:14 |
| @sonney2k | heiko, do you need a copy of the subset? | 12:14 |
| @sonney2k | I see that we now have get_subset_copy | 12:14 |
| heiko | yes, at one place | 12:14 |
| @sonney2k | ok - then please set the do_free flag to true | 12:14 |
| blackburn | oh! I know that I forgot | 12:15 |
| blackburn | to have a breakfast :D | 12:15 |
| heiko | oh, I dont need it anymore, this was before when i backuped subsets but I dont do this anymore ... so it can go | 12:16 |
| @sonney2k | blackburn, get your next 2d - meal... | 12:16 |
| @sonney2k | ...pizza! | 12:16 |
| @sonney2k | heiko, very good | 12:16 |
| blackburn | :D | 12:16 |
| * sonney2k merges | 12:16 | |
| CIA-32 | shogun: Sergey Lisitsyn master * r136be95 / src/libshogun/clustering/KMeans.h : Fixes for KMeans - http://bit.ly/kHxsU8 | 12:17 |
| CIA-32 | shogun: Sergey Lisitsyn master * r176a1b2 / (4 files in 2 dirs): Fixes for Alesis code - http://bit.ly/kD4pR8 | 12:17 |
| CIA-32 | shogun: Sergey Lisitsyn master * r7525d58 / src/libshogun/kernel/CustomKernel.cpp : Fix for CustomKernel - http://bit.ly/jHYaOo | 12:17 |
| CIA-32 | shogun: Soeren Sonnenburg master * r1a4582d / (7 files in 3 dirs): Merge branch 'master' of git://github.com/SergeyLisitsyn/shogun - http://bit.ly/mUHdkx | 12:17 |
| @sonney2k | heiko, please fix and then I merge your's too | 12:17 |
| heiko | ok | 12:17 |
| @sonney2k | (string.h and copy that is) | 12:17 |
| @sonney2k | blackburn, SGMatrix will really simplify our lives | 12:18 |
| @sonney2k | for example we can check if dims match etc | 12:18 |
| @sonney2k | currently all done manually everywhere | 12:18 |
| @sonney2k | blackburn, so we can get down to 100kloc | 12:18 |
| @sonney2k | :D | 12:18 |
| blackburn | indeed.. | 12:19 |
| blackburn | but refactoring is such a pain in ass | 12:19 |
| @sonney2k | blackburn, I cannot think of a more intrusive change than this one | 12:20 |
| @sonney2k | blackburn, the good thing is that we can test if everything works by just running the test suite | 12:21 |
| blackburn | we can't rely only on tests.. | 12:21 |
| heiko | ok done | 12:21 |
| blackburn | in fact it is not ideal | 12:21 |
| blackburn | Hierarchical works! | 12:21 |
| CIA-32 | shogun: Heiko Strathmann master * r2eb94d6 / (5 files): Merge branch 'toCheckin' (+5 more commits...) - http://bit.ly/iBI1JG | 12:22 |
| @sonney2k | blackburn, test suite and brainz | 12:23 |
| * sonney2k changes the typemaps to support the new do_free thingy | 12:23 | |
| heiko | sonney2k, btw this brach stuff works pretty cool | 12:23 |
| heiko | sonney2k, blackburn, I am out for lunch now and then working at home (no internet) so see you | 12:23 |
| blackburn | see you | 12:23 |
| @sonney2k | heiko, thanks and cu! | 12:23 |
| blackburn | sonney2k: another one pull request | 12:24 |
| -!- sploving [~root@124.16.139.196] has joined #shogun | 12:24 | |
| sploving | git fetch upstream so slow? | 12:24 |
| blackburn | time for breakfast, I worth it ! :D | 12:25 |
| @sonney2k | blackburn, thanks merged | 12:25 |
| CIA-32 | shogun: Sergey Lisitsyn master * r3937a21 / src/libshogun/clustering/Hierarchical.h : Fixes for Hierarchical clustering - http://bit.ly/lgZ4kD | 12:25 |
| CIA-32 | shogun: Sergey Lisitsyn master * rc5a09ba / (10 files in 4 dirs): Merge branch 'master' of https://github.com/shogun-toolbox/shogun - http://bit.ly/lGtsM8 | 12:25 |
| CIA-32 | shogun: Soeren Sonnenburg master * r8901172 / src/libshogun/clustering/Hierarchical.h : Merge branch 'master' of git://github.com/SergeyLisitsyn/shogun - http://bit.ly/ktjEVZ | 12:25 |
| @sonney2k | sploving, hi! | 12:25 |
| @sonney2k | sploving, not here... | 12:25 |
| sploving | sonney2k hi | 12:25 |
| @bettyboo | welcome | 12:25 |
| sploving | sonney2k, fetch upsteam so slow | 12:27 |
| @sonney2k | how slow is slow? | 12:27 |
| @sonney2k | here it takes <1sec | 12:27 |
| -!- heiko [~heiko@infole-06.uni-duisburg.de] has left #shogun [] | 12:27 | |
| sploving | remote: Counting objects: 1043, done. | 12:27 |
| sploving | remote: Compressing objects: 100% (245/245), done. | 12:27 |
| sploving | Receiving objects: 25% (216/862) it stops there | 12:27 |
| blackburn | ~10s for me | 12:28 |
| CIA-32 | shogun: Soeren Sonnenburg master * r3082f37 / (2 files in 2 dirs): remove trailing whitespaces - http://bit.ly/jOmGeN | 12:34 |
| sploving | maybe there is merging, so it is slow now | 12:34 |
| @sonney2k | sploving, I removed a few trailing whitespaces in your commits. please avoid thx! | 12:34 |
| sploving | oh. thanks. sonney2k, I would support sparse matrix this week, and clean up the examples later | 12:35 |
| sploving | where could I learn the sparse matrix? | 12:38 |
| @sonney2k | sploving, lib/DataType.h | 12:42 |
| @sonney2k | sploving, or look at all the other typemaps python_modular for example how I did this | 12:42 |
| sploving | I need to understand what the date structure that sparse matrix use. | 12:43 |
| sploving | I will look at DataType.h | 12:44 |
| blackburn | hey it is not compiling :D | 12:47 |
| sploving | what is the matter? | 12:47 |
| blackburn | sonney2k: num_cols(orig.num_cols), do_free(free_matrix) { } | 12:49 |
| blackburn | what is free matrix?? | 12:49 |
| blackburn | ah I see | 12:49 |
| blackburn | in copy constructor it shouldn't be | 12:50 |
| CIA-32 | shogun: Soeren Sonnenburg master * rdf429eb / src/libshogun/lib/DataType.h : fix typo - http://bit.ly/lu8p8l | 12:50 |
| @sonney2k | blackburn, sorry - fixed | 12:50 |
| * sonney2k continues working on typemaps | 12:50 | |
| blackburn | Subset.cpp isnot compiling | 12:50 |
| blackburn | will fix now | 12:50 |
| blackburn | sonney2k: I think we should use some method to set params from other SGMatrix/SGVector | 12:51 |
| blackburn | i.e. not constructor but copy | 12:52 |
| @sonney2k | blackburn, copy constructor or copy of matrix? | 12:52 |
| blackburn | sonney2k: some routine to avoid this: | 12:53 |
| blackburn | m_subset.vector=subset.vector; | 12:53 |
| blackburn | m_subset.vlen=subset.vlen; | 12:53 |
| @sonney2k | blackburn, you can just write m_subset=subset; | 12:54 |
| @sonney2k | we have that already | 12:54 |
| blackburn | is it legal? | 12:54 |
| @sonney2k | yes | 12:54 |
| blackburn | hmm | 12:54 |
| blackburn | ok | 12:54 |
| blackburn | didn't know | 12:54 |
| @sonney2k | that is what the copy constructor is for | 12:54 |
| blackburn | I just thought it will create new instance of SGVector | 12:55 |
| blackburn | and replace | 12:55 |
| blackburn | will it? | 12:55 |
| @sonney2k | SGMatrix<xx> y; | 12:55 |
| @sonney2k | x=y; | 12:55 |
| blackburn | no memory leaks? | 12:55 |
| @sonney2k | means all things will be copied | 12:55 |
| @sonney2k | why? | 12:56 |
| blackburn | I don't know | 12:56 |
| @sonney2k | it is just doing | 12:56 |
| @sonney2k | <blackburn> m_subset.vector=subset.vector; | 12:56 |
| @sonney2k | <blackburn> m_subset.vlen=subset.vlen; | 12:56 |
| blackburn | ah | 12:56 |
| blackburn | I see | 12:56 |
| @sonney2k | exactly that | 12:56 |
| @sonney2k | look at the copy constructor - it is written there what it does | 12:56 |
| @sonney2k | it copies do_free too btw :) | 12:56 |
| @bettyboo | :^) | 12:56 |
| blackburn | hooray | 12:57 |
| blackburn | compiles | 12:57 |
| @sonney2k | blackburn, a rare occasion these days | 12:58 |
| blackburn | will do a pull request in a minute | 12:58 |
| blackburn | hehe I should focus on my project, landmark mds is waiting for me | 12:59 |
| sploving | not understanding SGSparseVectorEnty | 13:00 |
| sploving | does sparse matrix use data structure the same as matlab, python? | 13:06 |
| CIA-32 | shogun: Sergey Lisitsyn master * r2cf4b40 / (3 files in 2 dirs): vector length renaming - http://bit.ly/j5oPdS | 13:07 |
| @sonney2k | thanks blackburn | 13:07 |
| blackburn | it compiles!! | 13:07 |
| blackburn | heh | 13:07 |
| @sonney2k | sploving, python/matlab etc all use different ones | 13:08 |
| @sonney2k | sploving, a sparse vector is a tuple of index value | 13:09 |
| @sonney2k | err a sparse vector entry | 13:09 |
| @sonney2k | a sparse vector is an array of these entries | 13:09 |
| @sonney2k | a sparse matrix is an array of these vectors | 13:09 |
| blackburn | ImportError: cannot import name HDF5File | 13:09 |
| @sonney2k | that's it | 13:09 |
| blackburn | ше шы ыщьу шыыгуы ецшер | 13:10 |
| sploving | I will try to understand it thx | 13:10 |
| blackburn | damn | 13:10 |
| blackburn | it is some issue with my os? | 13:10 |
| @sonney2k | blackburn, did this ever work for you before? | 13:10 |
| blackburn | don't know | 13:10 |
| blackburn | I'll try to fix string features now | 13:10 |
| blackburn | and simple features too | 13:11 |
| sploving | sonney2k, sparsevector, for example, x = [0, 0, 1, 0, 0, 3] then feat_entries = 6? | 13:15 |
| @sonney2k | blackburn, thx | 13:15 |
| @sonney2k | x cannot contain zeros | 13:16 |
| sploving | features = [{3,1},{6,3}]? | 13:16 |
| sploving | I mean that x is a sparse vector. rows =1, cols =6 , and most of its element is zero | 13:16 |
| @sonney2k | sparse vector has no rows then | 13:17 |
| @sonney2k | would need a sparse matrix for that | 13:17 |
| @sonney2k | num_feat_entries would be 2 | 13:18 |
| sploving | then could you give me sparse vector example? | 13:18 |
| @sonney2k | feat_idx would be 2 with entry 1 | 13:18 |
| @sonney2k | and feat_idx woudl be 5 and value 3 | 13:18 |
| sploving | oh | 13:18 |
| sploving | what is vec_index? | 13:19 |
| sploving | oh I know | 13:20 |
| sploving | that is for matrix | 13:20 |
| @sonney2k | yes | 13:20 |
| @sonney2k | and num_vectors is number of sparsevectors in sparse matrix | 13:20 |
| @sonney2k | and num_features can be some large number with number of dims | 13:20 |
| @sonney2k | (not encoded in vector) | 13:20 |
| blackburn | development of sparse routines could kill | 13:21 |
| blackburn | :D | 13:21 |
| @sonney2k | blackburn, it is not really difficult ... | 13:21 |
| @sonney2k | but operations on that can be | 13:21 |
| blackburn | sonney2k: some lanczos on sparse matrices could be :D | 13:21 |
| @sonney2k | I meant: the format is not very difficult but oeprations | 13:21 |
| @sonney2k | yes | 13:21 |
| sploving | then what is the rows and cols of sparsematrix? | 13:22 |
| blackburn | even sparse dot is something terrible | 13:22 |
| sploving | not num_vector | 13:22 |
| sploving | rows = num_features, what is cols? | 13:24 |
| sploving | there is no way to know cols? | 13:27 |
| @sonney2k | got disconnect not sure if you read | 13:28 |
| @sonney2k | <sonney2k> blackburn, not so... | 13:28 |
| @sonney2k | <sonney2k> sploving, num_vec | 13:28 |
| @sonney2k | <sonney2k> though that is rather arbitrary | 13:28 |
| sploving | sonney2k, what is cols of sparse matrix | 13:29 |
| @sonney2k | <sonney2k> sploving, num_vec | 13:29 |
| sploving | num_features is rows | 13:29 |
| blackburn | sonney2k: do you know how can I run multiple python files? | 13:29 |
| blackburn | e.g. I want to run features_* | 13:29 |
| @sonney2k | blackburn: for i in features_*.py ; do python $i ; done | 13:30 |
| blackburn | cool | 13:30 |
| sploving | for example a sparse matrix [(0, 0, 2, 0), (7,0,0,0), (0, 8, 0, 0)] | 13:34 |
| sploving | rows =3, cols = 4 | 13:34 |
| sploving | you mean num_vector = 4 | 13:35 |
| @sonney2k | sploving, it depends on the meaning | 13:37 |
| @sonney2k | but yes | 13:37 |
| sploving | here vector mean colvector, sonney2k? | 13:38 |
| @sonney2k | in shogun vectors are column vectors and 1 example == 1 vector == 1 column | 13:39 |
| blackburn | sonney2k: SGMatrix<SGString<ST>>? | 13:42 |
| blackburn | SGString<ST>* p_features, int32_t p_num_vectors, | 13:42 |
| blackburn | int32_t p_max_string_length | 13:42 |
| @sonney2k | blackburn, ? | 13:42 |
| sploving | there is still a question. num_vectors = 4, but maybe the sparsematrix has more than 4 cols, the other cols are all 0 | 13:42 |
| blackburn | sonney2k: how can I replace | 13:42 |
| blackburn | SGString<ST>* p_features, int32_t p_num_vectors, | 13:42 |
| blackburn | int32_t p_max_string_length | 13:42 |
| blackburn | with no typemaps it doesn't work | 13:43 |
| @sonney2k | blackburn, lib/DataType.h -> SGStringList | 13:43 |
| blackburn | hoho | 13:43 |
| @sonney2k | sploving, sure | 13:44 |
| @sonney2k | then you have to set num_vectors / num_features to bigger values | 13:44 |
| @sonney2k | and use empty sparse vectors | 13:44 |
| sploving | I know. I think sparse vectors could not be empty | 13:45 |
| @sonney2k | sploving, well they can - just set num_feat_entries to 0 | 13:46 |
| @sonney2k | blackburn, these are all still structs though without fancy constructors... | 13:46 |
| blackburn | sonney2k: I don't need ones now | 13:47 |
| sploving | sonney2k, in numpy, there are seven available sparse matrix types. | 13:51 |
| sploving | we just need typemap one kind of? | 13:51 |
| @sonney2k | sploving, I know - but you only need to support ujmp | 13:51 |
| @sonney2k | numpy has no sparse matrix support | 13:51 |
| @sonney2k | scipy has | 13:52 |
| sploving | yeap | 13:52 |
| @sonney2k | and we support ccs already | 13:52 |
| @sonney2k | and that is sufficient | 13:52 |
| sploving | I saw. I am looking at the typemap | 13:52 |
| CIA-32 | shogun: Soeren Sonnenburg master * re72475c / (7 files in 7 dirs): utilize free_vector/free_matrix in typemaps - http://bit.ly/k4zoFm | 14:07 |
| blackburn | I'm in one step to make string features to work | 14:09 |
| blackburn | kernels are working except two! | 14:16 |
| -!- alesis-novik [~alesis@188.74.87.206] has joined #shogun | 14:19 | |
| blackburn | alesis-novik: we lost you! | 14:19 |
| sploving | sonney2k, indptr = array([0,2,3,6]) , indices = array([0,2,2,0,1,2]) , data = array([1,2,3,4,5,6]) , csc_matrix( (data,indices,indptr), shape=(3,3) ).todense() matrix([[1, 0, 4], [0, 0, 5], [2, 3, 6]]) | 14:19 |
| blackburn | sonney2k: pull request is in | 14:20 |
| sploving | I just do not understand indptr? | 14:20 |
| alesis-novik | blackburn, I was travelling the last week, and while I did do work, I couldn't idle in IRC. | 14:20 |
| sploving | does it mean, the sum num in the cloumn? | 14:20 |
| alesis-novik | I'll submit the report later today as well | 14:20 |
| blackburn | alesis-novik: I have modified your code a little - SGVector interface issues | 14:21 |
| @sonney2k | alesis-novik, please notify us next time that you are away - we all were very worried | 14:21 |
| blackburn | so be aware when will merge it | 14:21 |
| alesis-novik | sonney2k, sorry about not giving prior notice | 14:21 |
| alesis-novik | blackburn, what was the issue? | 14:22 |
| blackburn | https://github.com/shogun-toolbox/shogun/commit/176a1b2419fd125c621633d3e9fb3b078e508c33 | 14:22 |
| CIA-32 | shogun: Sergey Lisitsyn master * r478e05f / (3 files in 2 dirs): Fixes for Features and SpectrumRBFkernel - http://bit.ly/kvzEFY | 14:22 |
| CIA-32 | shogun: Soeren Sonnenburg master * ra71be6a / (3 files in 2 dirs): Merge branch 'master' of git://github.com/SergeyLisitsyn/shogun - http://bit.ly/lNS5NJ | 14:22 |
| @sonney2k | blackburn, so what is missing now? | 14:22 |
| @sonney2k | sparse matrix? | 14:22 |
| blackburn | sonney2k: what kind of missing? | 14:22 |
| alesis-novik | sonney2k, should I submit pull request even if I know that's not the final version of the code, or should I keep it in my local rep until I'm done? | 14:23 |
| @sonney2k | blackburn, you did all the libshogun refactoring - which files/dirs are missing | 14:23 |
| @sonney2k | alesis-novik, I really prefer *small* changes | 14:23 |
| blackburn | eh..? | 14:23 |
| @sonney2k | blackburn, missing to be refactored | 14:23 |
| blackburn | ah | 14:23 |
| sploving | sonney2k, what does indptr mean? | 14:23 |
| blackburn | well sparsefeatures | 14:23 |
| @sonney2k | and? | 14:23 |
| blackburn | distributions | 14:23 |
| @sonney2k | and? | 14:24 |
| blackburn | doing tests now | 14:24 |
| blackburn | ah | 14:24 |
| blackburn | some classifier issues | 14:24 |
| blackburn | larank doesn't work | 14:24 |
| @sonney2k | alesis-novik, so the smaller the better | 14:24 |
| blackburn | sparse kernels aren't working | 14:24 |
| @sonney2k | blackburn, but I guess because of the virtual function not appearing... | 14:24 |
| @sonney2k | blackburn, shall I do sparse now? | 14:24 |
| blackburn | yes, would be nice | 14:25 |
| blackburn | mkl_multiclass_modular causing memory problems :) | 14:25 |
| alesis-novik | Darn it... | 14:25 |
| alesis-novik | blackburn, the fixes are good, but you updated the old code. I'll apply the fixes to the rewritten classes. It seems straight forward | 14:26 |
| blackburn | alesis-novik: feel free to rewrite it :) we just care it should return SGVectors and SGMatrices | 14:27 |
| blackburn | because we removed typemaps | 14:27 |
| blackburn | for ST** things | 14:27 |
| alesis-novik | So essentially all arrays with appropriate sizes should be replaced with SGVectors and SGMatrices | 14:28 |
| blackburn | yes | 14:28 |
| blackburn | alesis-novik: could you please also add some very simple example for python_modular? | 14:28 |
| blackburn | or any other language you prefer :) | 14:28 |
| @bettyboo | yep | 14:28 |
| @sonney2k | blackburn, no python_modular definitely | 14:28 |
| @sonney2k | python_modular is the reference.... | 14:29 |
| blackburn | sonney2k: eh? | 14:29 |
| @sonney2k | blackburn, if sth. appears new in shogun it will be working outside of C++ first in python_modular | 14:29 |
| @sonney2k | so any example should be done first in python | 14:30 |
| blackburn | ah okay | 14:30 |
| alesis-novik | blackburn, I've been working on an Octave example | 14:30 |
| @sonney2k | alesis-novik, octave_modular? | 14:32 |
| @sonney2k | Ahh I guess I understand - vojtech is using matlab so octave is the GCD | 14:32 |
| -!- sploving [~root@124.16.139.196] has quit [Ping timeout: 240 seconds] | 14:38 | |
| @sonney2k | blackburn, for shogun-toolbox.org I need your ssh public key | 14:45 |
| blackburn | sonney2k:where should I send it? | 14:47 |
| @sonney2k | blackburn, gpg signed email | 14:48 |
| blackburn | sonney2k: never used 'gpg signed email' :) | 14:48 |
| @sonney2k | then send it plain text | 14:48 |
| @sonney2k | blackburn, should work now | 14:52 |
| blackburn | will test | 14:54 |
| CIA-32 | shogun: Sergey Lisitsyn master * rc7daae3 / src/libshogun/preprocessor/ClassicMDS.cpp : Small fix for ClassicMDS - http://bit.ly/mfbhlo | 15:51 |
| @sonney2k | blackburn, now that you can merge too - I guess we have to synchronize and always do git pull before | 15:55 |
| blackburn | sonney2k: exactly | 15:55 |
| * sonney2k prepares for retirement | 15:57 | |
| @sonney2k | :D | 15:57 |
| @bettyboo | HA | 15:57 |
| blackburn | heheh | 15:58 |
| blackburn | have to go to bus station, will be heading to samara :) | 15:58 |
| blackburn | see you | 15:58 |
| -!- blackburn [~blackburn@31.28.40.202] has left #shogun [] | 15:58 | |
| -!- in3xes [~in3xes@180.149.49.227] has quit [Quit: Leaving] | 18:37 | |
| -!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 18:38 | |
| -!- blackburn [~blackburn@188.168.3.18] has joined #shogun | 19:29 | |
| blackburn | sonney2k: houston we have a problem | 19:43 |
| blackburn | set_labels don't working I guess | 19:43 |
| CIA-32 | shogun: Sergey Lisitsyn master * rab158e1 / (3 files): Classify->apply transition for graphical python modular examples - http://bit.ly/ijk2Ua | 19:58 |
| @sonney2k | blackburn, what? | 20:18 |
| @sonney2k | set_labels no longer works? | 20:18 |
| blackburn | sonney2k: can't tell for sure but in graphical python examples - no | 20:35 |
| blackburn | it says labels aren't set | 20:36 |
| @sonney2k | which example exactly? | 20:37 |
| blackburn | svr_sync.py | 20:38 |
| blackburn | krr_sync.py | 20:38 |
| blackburn | ah sorry svr works | 20:38 |
| blackburn | krr_sync not | 20:38 |
| @sonney2k | blackburn, btw, please configure --interfaces=libshogunui,python too | 20:38 |
| @sonney2k | this way you know that it doesn't compile currently | 20:39 |
| * sonney2k is currently fixing this | 20:39 | |
| blackburn | it doesn't compile?? hehe | 20:39 |
| blackburn | oh | 20:46 |
| blackburn | GUIClassifier.cpp:1304: error: ‘class shogun::CKMeans’ has no member named ‘get_radi’ | 20:47 |
| blackburn | GUIClassifier.cpp:1310: error: ‘class shogun::CKMeans’ has no member named ‘get_centers’ | 20:47 |
| blackburn | GUIClassifier.cpp:1323: error: ‘class shogun::CHierarchical’ has no member named ‘get_merge_distance’ | 20:47 |
| blackburn | GUIClassifier.cpp:1328: error: ‘class shogun::CHierarchical’ has no member named ‘get_pairs’ | 20:47 |
| blackburn | sonney2k: that is my fault | 20:47 |
| @sonney2k | I fixed that already | 20:48 |
| @sonney2k | have tons of other probs now | 20:48 |
| blackburn | thanks | 20:48 |
| blackburn | sonney2k: is this related to SGVector conversion? | 20:53 |
| @sonney2k | and friends yes :) | 20:53 |
| CIA-32 | shogun: Soeren Sonnenburg master * rdd63a89 / (6 files in 3 dirs): continue T** transition and fix compilation of static interfaces - http://bit.ly/mtwhu4 | 21:01 |
| CIA-32 | shogun: Soeren Sonnenburg master * ra909a9f / data : require new data - http://bit.ly/jGZLtk | 21:01 |
| CIA-32 | shogun: Soeren Sonnenburg master * r880d418 / (4 files in 2 dirs): Merge branch 'master' of github.com:shogun-toolbox/shogun - http://bit.ly/iOE2in | 21:01 |
| @sonney2k | blackburn, ok compiles again | 21:01 |
| blackburn | sonney2k: hooray | 21:01 |
| @sonney2k | blackburn, I am not even half-way through sparse features though | 21:02 |
| blackburn | I see | 21:02 |
| @sonney2k | blackburn, did you do 'simple'features already? | 21:02 |
| blackburn | sonney2k: well tests are passing | 21:03 |
| blackburn | tests for stringfeatures too | 21:03 |
| * blackburn away | 21:03 | |
| @sonney2k | blackburn, krr*.py examples were broken due to a bug in KRR.cpp | 21:38 |
| blackburn | oh I see | 22:12 |
| blackburn | sonney2k: is it fixed? | 22:12 |
| @sonney2k | yes | 22:13 |
| CIA-32 | shogun: Soeren Sonnenburg master * r95ce789 / src/libshogun/regression/KRR.cpp : fix a bug in KRR label getting - http://bit.ly/immTFS | 22:13 |
| CIA-32 | shogun: Soeren Sonnenburg master * rf61db8e / src/csharp_modular/swig_typemaps.i : fix typo -> use free_vector() to free - http://bit.ly/mJZ7A6 | 22:13 |
| CIA-32 | shogun: Soeren Sonnenburg master * rb3f2f96 / src/libshogun/features/SparseFeatures.h : T** transition for sparse features - http://bit.ly/ifw9zm | 22:13 |
| CIA-32 | shogun: Soeren Sonnenburg master * rf5341be / src/csharp_modular/swig_typemaps.i : fix cshapr compile error - http://bit.ly/iRuV6i | 22:25 |
| blackburn | sonney2k: we could try issues for tasks | 22:26 |
| blackburn | on github | 22:26 |
| @sonney2k | blackburn, for sparse features I only added SGType stuff - didn't really remove the T* etc functions | 22:26 |
| blackburn | for example we can create a milestone 'Get the f-g SGVector transition' | 22:26 |
| blackburn | and then make a pull requests or so | 22:26 |
| blackburn | ah nevermind | 22:27 |
| blackburn | :D | 22:27 |
| blackburn | anyway we both doing it without pull requests now | 22:27 |
| blackburn | bad idea | 22:27 |
| CIA-32 | shogun: Soeren Sonnenburg master * r76c37e2 / examples/undocumented/libshogun/classifier_mklmulticlass.cpp : fix compile error in mkl multiclass example - http://bit.ly/jpRCm0 | 22:32 |
| @sonney2k | ahh I should have known that running tests now becomes a problem | 22:39 |
| blackburn | what's up with tests? | 22:40 |
| @sonney2k | blackburn, for python modular tests are run from shogun/testsuite/python_modular | 22:41 |
| @sonney2k | and there python tester.py | 22:41 |
| blackburn | never used it :) | 22:41 |
| @sonney2k | or python generator.py | 22:41 |
| @sonney2k | what they do is: every example is a function | 22:41 |
| @sonney2k | generator will run the function with all the partmeter settings in the example | 22:42 |
| @sonney2k | and then python pickle the returned objects | 22:42 |
| @sonney2k | tester.py will runt he example to but load the pickled data | 22:42 |
| @sonney2k | and then just compare | 22:42 |
| @sonney2k | now we have thanks to heiko subset support in labels etc | 22:42 |
| @sonney2k | these things didn't exist when the things were pickled | 22:43 |
| @sonney2k | so all tests fail no matter what | 22:43 |
| blackburn | I see | 22:43 |
| @sonney2k | which is bad - because that really is the only way to check that we are in good shape | 22:43 |
| @sonney2k | blackburn, I guess I can modify the tester to have an option to ignore unknown new things | 22:45 |
| @sonney2k | anyway first examples have to run at all | 22:46 |
| blackburn | sonney2k: I simply used ./check.sh at python_modular | 22:48 |
| @sonney2k | blackburn, yeah but many still don't run | 22:48 |
| blackburn | yeap | 22:48 |
| blackburn | larank | 22:54 |
| blackburn | distributions | 22:54 |
| blackburn | sparse features | 22:55 |
| blackburn | sonney2k: I did some changes with vim! :) | 23:08 |
| blackburn | sonney2k: and btw mkl_multiclass fails | 23:08 |
| blackburn | you modified something related to | 23:09 |
| @sonney2k | blackburn, yeah the example did not even compile | 23:13 |
| blackburn | ah I see | 23:13 |
| blackburn | python modular example fails with memory corruption | 23:13 |
| @sonney2k | blackburn, I guess double free | 23:19 |
| blackburn | don't know, i'm reading paper on landmarks now :) | 23:21 |
| @bettyboo | :) | 23:21 |
| @sonney2k | blackburn, yeah I will have a look but currently fixing wd kernel | 23:21 |
| --- Log closed Tue Jun 21 00:00:06 2011 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!