--- Log opened Wed Jun 08 00:00:08 2011 | ||
@sonney2k | serialhex, np! I understand that your life/job has priority! | 00:00 |
---|---|---|
serialhex | btw, i keep getting this error when i load anyof the ruby modules: | 00:01 |
serialhex | src/ruby_modular/Distribution.so: warning: rb_define_const: invalid name `__STDC_FORMAT_MACROS' for constant | 00:01 |
serialhex | that happens with all of them, and i'm not sure if that is going to be a time bomb and destroy things later on... i figured i'd let you know about it now before it explodes (if it does...) | 00:02 |
serialhex | all constants in ruby are have an initial capital letter so if tit was simply STDC_FORMAT_MACROS it'd be fine, but idk what it's for, or if you want me to dig into it to see what i can fix | 00:04 |
CIA-18 | shogun: Baozeng Ding master * r5d0f71b / examples/undocumented/java_modular/ClassifierLibsvmMinimalModular.java : add ClassifierLibsvmMinimalModular exmaple - http://bit.ly/jbNoVM | 00:30 |
CIA-18 | shogun: Soeren Sonnenburg master * r10af489 / src/libshogun/features/Labels.h : add SGVector based constructor - http://bit.ly/m0p8ev | 00:30 |
@sonney2k | serialhex, I don't know what `__STDC_FORMAT_MACROS' is | 00:37 |
@sonney2k | in shogun constants are also all uppercase | 00:37 |
@sonney2k | at least they should be | 00:37 |
serialhex | hmm, well i'll look into it later then | 00:38 |
@sonney2k | g'night | 01:24 |
-!- Netsplit *.net <-> *.split quits: @sonney2k, @bettyboo | 04:11 | |
-!- Netsplit over, joins: bettyboo, sonney2k | 04:11 | |
-!- mode/#shogun [+o bettyboo] by ChanServ | 04:11 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 04:11 | |
-!- blackburn [~blackburn@188.168.4.201] has joined #shogun | 08:02 | |
-!- f-x [~shashwat@117.192.210.115] has joined #shogun | 10:13 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 13:36 | |
sploving | anybody? | 13:36 |
sploving | sooney2k? | 13:37 |
sploving | around? | 13:37 |
sploving | sonney2k, how to integrate them into one module shogun | 13:38 |
blackburn | sploving: can I help you? | 13:38 |
sploving | java typemap issue | 13:39 |
blackburn | ah | 13:40 |
blackburn | I guess I can't :) | 13:40 |
sploving | what about your examples? | 13:40 |
blackburn | examples of? | 13:40 |
sploving | how is it going ? | 13:40 |
blackburn | ah | 13:40 |
sploving | exams | 13:40 |
blackburn | I have one on 10th | 13:41 |
blackburn | now preparing | 13:41 |
sploving | the last one? | 13:41 |
blackburn | 1/3 | 13:41 |
blackburn | last one on 21st | 13:42 |
blackburn | I hope I could do some work on my project after that first exam | 13:42 |
blackburn | sonney2k's cpu fire applet rocks | 13:42 |
sploving | what ?? | 13:43 |
blackburn | sploving: are you running ubuntu? | 13:43 |
sploving | no | 13:43 |
sploving | fedora | 13:43 |
blackburn | ah | 13:43 |
sploving | blackburn, And following up on this. If we don't get this resolved the only way | 13:44 |
sploving | would be to put everything into a single shogun module. This would | 13:44 |
sploving | resolve all issues except that we now will never be able to modularize | 13:44 |
sploving | the code (e.g. only ship a isolated Features module). | 13:44 |
sploving | what does that mean? | 13:44 |
blackburn | I don't understand about what it is | 13:44 |
blackburn | ah | 13:45 |
blackburn | I guess I see | 13:45 |
sploving | blackburn, i sent you the email | 13:46 |
blackburn | well seems you have a problem with inter-package relation in your interfaces, right? | 13:47 |
sploving | yeap | 13:48 |
sploving | one package may need call another package | 13:48 |
blackburn | so in the case it wasn't resolved we would have to place everything in one module | 13:49 |
blackburn | I hope it won't take place indeed | 13:49 |
sploving | why ? | 13:49 |
blackburn | if it was in one module there are would not be any conflicts | 13:49 |
blackburn | cause in this case we will lose structure | 13:50 |
blackburn | just mess of *.cpp and *.h | 13:50 |
-!- sploving [~root@124.16.139.196] has quit [Read error: Operation timed out] | 13:52 | |
blackburn | bam | 13:52 |
@sonney2k | sploving is not around... :( | 14:02 |
-!- sploving [~root@124.16.139.196] has joined #shogun | 14:08 | |
blackburn | sonney2k: now around | 14:14 |
@sonney2k | sploving, around? | 14:14 |
blackburn | hehe | 14:14 |
sploving | yeap | 14:14 |
sploving | sonney2k, I added "import org.shogun.*;". the error is the same | 14:15 |
@sonney2k | sploving, I managed to work around the error | 14:15 |
@sonney2k | one thing remains though: for vectors we don't have jblas support | 14:15 |
sploving | you mean jblas causes such error? | 14:16 |
@sonney2k | sploving, no but one cannot set labels with a jblas DoubleMatrix | 14:17 |
sploving | why not? I tested jblas, it could work with vector (VectorTest example) | 14:18 |
sploving | before rename classifier to labels, it works well | 14:18 |
sploving | sorry. I mean rename classifier to machine | 14:18 |
@sonney2k | VectoTest example still uses double[] | 14:19 |
sploving | sonney2k, it is great to fix the module interface example | 14:19 |
@sonney2k | not jblas DoubleMatrix | 14:19 |
@sonney2k | sploving, no Classifier is correct | 14:19 |
@sonney2k | gtb | 14:19 |
@sonney2k | be back in a minute | 14:19 |
sploving | okay | 14:20 |
sploving | It is not difficult to wrap vector in jblas | 14:22 |
sploving | but do we need to do so? | 14:23 |
@sonney2k | Re | 14:23 |
@sonney2k | sploving, yes | 14:24 |
@sonney2k | I would assume people only use jblas based matrices not double[] | 14:24 |
@sonney2k | sploving, I think you need to add a typecheck to test if it is really a vector or not | 14:25 |
sploving | ok. this is easy to add that | 14:25 |
sploving | sonney2k, I need rewrite the typemap for SGVector | 14:25 |
@sonney2k | IIRC jblas doc for DoubleMatrix says it has 0 colums or 1 (?) | 14:25 |
@sonney2k | so that would be the test | 14:26 |
@sonney2k | sploving, I managed to get things to compile by adding just %import's | 14:26 |
sploving | you mean the minimal example could run as expected? | 14:26 |
@sonney2k | though now python modular doesn't compile - and I will now have to add other fixes | 14:26 |
sploving | no five error? | 14:26 |
@sonney2k | sploving, no error except for Label | 14:27 |
@sonney2k | (no DoubleMatrix jblas support as I said) | 14:27 |
sploving | sonney2k, could you pushed it? or afer i add vector support? | 14:27 |
@sonney2k | sploving, I have to make sure that all interfaces work - not just java | 14:28 |
@sonney2k | so it will take more time | 14:28 |
sploving | you not only change swig_typemaps.i? | 14:28 |
sploving | sonney2k, why %imports effect python? | 14:29 |
@sonney2k | sploving, I changed all in modular/*.i | 14:29 |
@sonney2k | sploving, for some reason - when I do import, swig creates some kind of data type mapping table | 14:30 |
@sonney2k | but because these are not in %{ %include * %} C++ compilation fails | 14:30 |
@sonney2k | I think I will now move all the %include parts into separate files | 14:31 |
@sonney2k | and include these in the appropriate places | 14:31 |
sploving | so complex. why make java has its own .i file in the java_modular dir | 14:32 |
@sonney2k | sploving, for example one error is Classifier_wrap.cxx: In function ‘void* _p_shogun__CStringFileFeaturesT_unsigned_char_tTo_p_shogun__CStringFeaturesT_unsigned_char_t(void*, int*)’: | 14:32 |
@sonney2k | sploving, I don't understand | 14:33 |
@sonney2k | sploving, each interface has just swig_typemaps.i | 14:33 |
@sonney2k | and the rest is in modular/*.i | 14:33 |
@sonney2k | so modules are defined in modular/ | 14:33 |
@sonney2k | i/o typemaps in {python,java,...}_modular | 14:34 |
sploving | I mean we copy modular/*.i into java_dir and so we could not effect others if we changes these i | 14:34 |
@sonney2k | sploving, I could aswell add ifdef's in modular/*.i | 14:34 |
@sonney2k | but I suspect that I should really have used the %import stuff in the first place to make other types known | 14:35 |
@sonney2k | I didn't know sth like import exists... | 14:35 |
@sonney2k | did only ever use include | 14:35 |
sploving | sonney2k, I still did not undersand why import could solve this. the function need (CDotFeatures), we transfer it with RealFeatures | 14:36 |
@sonney2k | sploving, look at the generated *.java file | 14:36 |
sploving | what is the relationship between the two types | 14:36 |
@sonney2k | CDotFeatures is a 'base' of RealFeatures | 14:37 |
@sonney2k | and since this is not included in the e.g. module Kernel (defined in modular/Kernel.i) swig for java generates some dummy type | 14:37 |
@sonney2k | SWIG_* | 14:37 |
@sonney2k | and so does not know that the RealFeatures coming from a separate module are in fact the ones needed | 14:38 |
@sonney2k | at least that is what I think happens | 14:39 |
sploving | I saw RealFeatures extends DotFeatures | 14:39 |
@sonney2k | ohh strong thunderstorm here... | 14:39 |
@sonney2k | yes | 14:39 |
sploving | if we transefer DotFeatures, does it work well for that function | 14:40 |
@sonney2k | transfer? | 14:40 |
sploving | I mean as argument | 14:40 |
@sonney2k | if we in Kernel.i include Features.i it would work yes | 14:40 |
-!- VojtechFranc [~quassel@r2.inovacentrum.cvut.cz] has joined #shogun | 14:42 | |
sploving | I think when you after pushed the code, then maybe I understand clearly | 14:42 |
sploving | I will do the vector typemap | 14:42 |
sploving | hope good news~ | 14:42 |
@sonney2k | sploving, I don't know if one can really understand this | 14:42 |
@sonney2k | but look at the compile error you are getting | 14:43 |
@sonney2k | I mean look in the java file of GaussianKernel.java | 14:43 |
@sonney2k | you will see that its constuctor has some weird SWIG_*shogun*CDotFeatures arguments | 14:43 |
@sonney2k | even though these should just be dotfeatures | 14:43 |
sploving | sonney2k, you mean: | 14:44 |
sploving | public GaussianKernel(SWIGTYPE_p_shogun__CDotFeatures l, SWIGTYPE_p_shogun__CDotFeatures r, double width) | 14:44 |
@sonney2k | yes | 14:44 |
sploving | what it should like? | 14:44 |
@sonney2k | just DotFeatures | 14:44 |
sploving | public GaussianKernel(DotFeatures l, CDotFeatures r, double width) | 14:44 |
@sonney2k | I think so yes | 14:44 |
sploving | after that if we use RealFeature it would work OK? | 14:44 |
@sonney2k | not sure if the JNI underneath is also messed up because of this | 14:44 |
sploving | why it generated SWIG* not DotFeatures? | 14:47 |
sploving | just not import? | 14:48 |
sploving | sonney2k, how strange | 14:48 |
@sonney2k | sploving, it just doesn't know | 14:48 |
@sonney2k | I mean look at modular/Kernel.i | 14:48 |
@sonney2k | there is no include or whatever of DotFeatures | 14:48 |
@sonney2k | so it just doesn't know this type | 14:48 |
@sonney2k | that is why I am hoping that %import will fix this properly and potentially some side effects that we had in python_modular | 14:49 |
@sonney2k | sploving, does that make sense? | 14:51 |
sploving | I am looking at the java doc about import | 14:52 |
sploving | that make sense, but I do not know import could fix this | 14:52 |
-!- f-x [~shashwat@117.192.210.115] has quit [Ping timeout: 240 seconds] | 15:01 | |
alesis-novik | Hello VojtechFranc | 15:05 |
VojtechFranc | helo alesis, how it goes? | 15:06 |
alesis-novik | Good I believe | 15:06 |
VojtechFranc | great, can I alrady pull some code from you? | 15:06 |
alesis-novik | I should finish most of the current plan for EM itself this week. | 15:06 |
alesis-novik | Well, the currently pushed code is only a change in Gaussian to work with different covariances in log domain, but you can look at it | 15:07 |
VojtechFranc | you mean to finish the translation of all matlab codes to C? | 15:07 |
alesis-novik | VojtechFranc, most of it anyway | 15:07 |
VojtechFranc | as I told you, I woun't be connected the next week (till Sunday). | 15:08 |
alesis-novik | Since you said you'll be gone next week, I was wondering what kind of representative tests should I create? | 15:08 |
alesis-novik | Since that was the plan after finishing the matlab port | 15:08 |
VojtechFranc | did I send you the example application which uses EM-GMM for USPS classifiction? | 15:09 |
sploving | sonney2k, are you familiar with directorin and directorout typemap? | 15:10 |
VojtechFranc | it would be nice to have 2 examples using EM-GMM on synthetic 1D and 2D data with visualization of the solution | 15:10 |
VojtechFranc | in addition one application on real data, e.g. the USPS classification | 15:10 |
alesis-novik | VojtechFranc, was it with the matlab code? | 15:11 |
VojtechFranc | I don't remember ... I'm going to check | 15:11 |
alesis-novik | yes | 15:12 |
alesis-novik | it's there | 15:12 |
VojtechFranc | it should be "gmm_classifier/example1.m" | 15:13 |
VojtechFranc | so, you have it? | 15:14 |
alesis-novik | Yes | 15:14 |
alesis-novik | I think I'll write the examples in octave, if that's ok? | 15:14 |
VojtechFranc | octave is ok | 15:15 |
VojtechFranc | I guess it should run in Matlab as welll , right? | 15:15 |
alesis-novik | Because I can't use Matlab from where I do my shogun coding | 15:15 |
alesis-novik | It should either run, or be very easily portable | 15:15 |
@sonney2k | sploving, not javadoc import | 15:16 |
@sonney2k | sploving, %import in swig | 15:16 |
VojtechFranc | if you have the latest version (svn up), then you have also 2 example script using em-gmm on synthetic data | 15:16 |
@sonney2k | we don't use directors | 15:16 |
VojtechFranc | namelly, gmm_em_example1, gmm_em_example2 | 15:16 |
@sonney2k | sploving, woudl be too slow | 15:16 |
sploving | oh. sonney2k, how is going for python? | 15:17 |
alesis-novik | VojtechFranc, yes, those are with the gmm_em code itself | 15:17 |
@sonney2k | sploving, didn't have time to start yet... kids... | 15:17 |
sploving | take care of them | 15:18 |
VojtechFranc | ok, so you think you can finish implemnetaion of em-gmm and these examples by the end of the next week? | 15:18 |
VojtechFranc | alesis-novik, I think you can just modify these matlab examples to work with shogun | 15:19 |
alesis-novik | VojtechFranc, unless I have some problems with initialization (namely 1NN), I should be able to do it | 15:19 |
@sonney2k | sploving, I usually work through the nights on shogun | 15:19 |
@sonney2k | sploving, anyway have a look at the python_modular/swig_typemaps.i | 15:20 |
@sonney2k | I did the strings, sparse matrices nd-matrices support | 15:20 |
@sonney2k | using the new datatypes | 15:20 |
VojtechFranc | alesis-novik, it would be perfect. Then, the week after, we could do some testing and then trying to implement split/merge strategies | 15:20 |
-!- sploving [~root@124.16.139.196] has quit [Read error: Operation timed out] | 15:20 | |
alesis-novik | VojtechFranc, ok, that's the plan then | 15:21 |
VojtechFranc | alesis-novik, what is exactly the current state. I guess you have the object to represent GMM with different cov types. and what else | 15:21 |
alesis-novik | Well, it essentially also computes the pdf's of Gaussians, and I've done the rest for EM as well so expectation part is kind of done. | 15:23 |
alesis-novik | but since I haven't updated maximization yet, I can't test if it really works yet | 15:24 |
alesis-novik | I know pdf's work, since I tested then separately | 15:24 |
VojtechFranc | do you have already the implemenation of gmm_ml.m ? | 15:24 |
alesis-novik | not yet, I'm working on it now | 15:24 |
VojtechFranc | ok, you can test (write test scripts) also for this function. This function ois | 15:25 |
VojtechFranc | this function is actually very useful per se | 15:25 |
VojtechFranc | if you have annotated data, then it can be used to ML estimate of GMM | 15:25 |
alesis-novik | Right, so I'll separate that out into another method | 15:26 |
VojtechFranc | yes, it would be defintelly useful | 15:26 |
VojtechFranc | actually, there is also a simple example of use if you type "help gmm_ml" in matlab | 15:27 |
VojtechFranc | alesis-novik, I'll be leaving on Saturday (11.6) and return (18.6). So, if any problem appears don't hassitate to ask till Friday night | 15:30 |
alesis-novik | Will do VojtechFranc | 15:31 |
VojtechFranc | ok, and, please try to write some short progress report for each week | 15:32 |
alesis-novik | I will. I'll be writing them on Saturdays | 15:32 |
VojtechFranc | thanks | 15:33 |
-!- sploving [~root@124.16.139.196] has joined #shogun | 15:37 | |
alesis-novik | I'll be off now. I'll write if I have any questions VojtechFranc. | 16:01 |
sploving | ok done. see you | 16:11 |
-!- sploving [~root@124.16.139.196] has left #shogun [] | 16:11 | |
-!- VojtechFranc [~quassel@r2.inovacentrum.cvut.cz] has quit [Remote host closed the connection] | 16:24 | |
-!- f-x [~shashwat@117.192.192.195] has joined #shogun | 16:48 | |
CIA-18 | shogun: Soeren Sonnenburg master * r046c6bf / (22 files): (log message trimmed) | 16:50 |
CIA-18 | shogun: use %import to make other data types known to swig | 16:50 |
CIA-18 | shogun: To this end we need to introduce *_includes.i for the respective | 16:50 |
CIA-18 | shogun: includes that can then be re-used and %included. | 16:50 |
CIA-18 | shogun: For example for Features we put includes in Features_includes.i | 16:50 |
CIA-18 | shogun: and issue in Features.i a %include "Features_includes.i". | 16:50 |
CIA-18 | shogun: The Kernel module needs features and thus make them known. So we need to | 16:50 |
CIA-18 | shogun: Soeren Sonnenburg master * r911a4db / (2 files): fix the java examples - http://bit.ly/kEMZNs | 16:50 |
blackburn | huh LLE is slooow | 16:58 |
blackburn | sonney2k: will make pull request with LLE very soon | 16:59 |
@sonney2k | :) | 17:00 |
blackburn | oh damn it is slower than MDS 4-5 times | 17:00 |
blackburn | sonney2k: I've got an idea of how to organize multidimensional scaling in preprocessor | 17:01 |
CIA-18 | shogun: Soeren Sonnenburg master * r8100b1c / src/java_modular/swig_typemaps.i : Merge branch 'master' of git://github.com/sploving/shogun - http://bit.ly/lBa4iN | 17:01 |
blackburn | as all of MDSs uses classic MDS (eigendecomposition) on some stage - it should be main class | 17:02 |
blackburn | so Isomap and Landmark MDS will be derived | 17:02 |
CIA-18 | shogun: Soeren Sonnenburg master * r4008878 / (3 files): Merge branch 'streaming' of git://github.com/frx/shogun - http://bit.ly/m7D2b2 | 17:03 |
blackburn | moreover, MDS' eigenvalue problem is same as in PCA | 17:03 |
blackburn | so it could be 'grouped' too | 17:03 |
blackburn | I wonder how can anybody use LLE for some genomic data ahaha | 17:04 |
blackburn | five minutes already for N=1000 | 17:04 |
@sonney2k | blackburn, sounds great! | 17:04 |
@sonney2k | some ML algorithms were never meant to be useful ;-) | 17:04 |
blackburn | Hessian LLE works great as I know, but it would be implemented after mid-term | 17:05 |
@sonney2k | well at least not on real-world data sets | 17:05 |
blackburn | btw it is ever slower than LLE | 17:05 |
blackburn | :D ohsh it will work forever | 17:05 |
blackburn | sonney2k: I don't know any real-world dataset for dimreduction | 17:05 |
blackburn | in most articles they study 'swiss roll's, helixes and so on | 17:06 |
* sonney2k sigh | 17:07 | |
@sonney2k | I generated a makefile dependency problem | 17:07 |
blackburn | it would be nice if somebody suggest me a data set to compare algos will be available in shogun and write some article ;) | 17:07 |
blackburn | what kind of? | 17:07 |
@sonney2k | blackburn, we need to use %import in the module files like Kernel.i | 17:08 |
@sonney2k | to make Features known | 17:08 |
blackburn | aha | 17:09 |
@sonney2k | that means though that the generated Preprocessor_doxygen.i now have weird deps | 17:09 |
blackburn | I see | 17:09 |
@sonney2k | and so now if one compiles with make -j 4 twice it will compile but otherwise won't | 17:10 |
blackburn | hehe | 17:10 |
blackburn | just like latex | 17:11 |
blackburn | with some cross-references | 17:11 |
blackburn | sonney2k: hmm for some reason git is getting mad with some classifier/svm/ files | 17:25 |
blackburn | fatal: CRLF would be replaced by LF in src/libshogun/classifier/svm/GNPPLib.cpp. | 17:27 |
blackburn | is it a some issue with my OS or so? | 17:27 |
blackburn | will try to get another git clone | 17:29 |
blackburn | the same in another clone | 17:31 |
-!- f-x [~shashwat@117.192.192.195] has quit [Ping timeout: 276 seconds] | 17:33 | |
blackburn | aha. the problem was with my git config | 17:34 |
@sonney2k | ? | 17:37 |
blackburn | autocrlf | 17:39 |
blackburn | option | 17:39 |
-!- in3xes [~in3xes@210.212.58.111] has joined #shogun | 17:41 | |
serialhex | oi, whats up blackburn? | 17:53 |
blackburn | serialhex: hi | 17:53 |
blackburn | arrh resolving merge conflicts with LLE and MDS branches :) | 17:53 |
blackburn | how are you? | 17:54 |
serialhex | pretty good, i'm trying to figure out how to get ruby & shogun to play nicely together :D | 17:55 |
@bettyboo | funny | 17:55 |
serialhex | the amount of code i'm having to write to just get a libsvm minimal example up and running is sad! | 17:56 |
blackburn | serialhex: do you know how to group all commits into one? | 17:56 |
blackburn | with rebase or so | 17:56 |
serialhex | what do you mean? like a bunch of small commits that you've made? | 17:56 |
blackburn | aha | 17:57 |
blackburn | git rebase get a conflict | 17:59 |
blackburn | because there was some Preproc -> preprocessor and so on | 17:59 |
serialhex | ooh, nice! | 17:59 |
serialhex | (that you know WTF went wrong :P ) | 18:00 |
blackburn | ERROR: src/modular/PreProc.i: Not handling case 87b96cad515ec530dc680c6897519efb3716e70e -> -> 5304b43d23ff91e03343a49b28abb08e9886bf7e | 18:00 |
blackburn | fatal: merge program failed | 18:00 |
blackburn | I even don't know how to sort it out manually | 18:00 |
serialhex | hmm, that's odd... it usually gives you a way to sort it out manually | 18:02 |
serialhex | maybe manually change the name of PreProc.i to Preprocessor.i ?? | 18:03 |
blackburn | ok when what should I do if I have LLE.cpp in preproc dir now | 18:05 |
blackburn | he wants me to resolve it | 18:05 |
serialhex | erm... i dunno, move it?? | 18:07 |
blackburn | it is a crazy shit | 18:08 |
blackburn | oh | 18:10 |
blackburn | done | 18:10 |
blackburn | a bunch of commits but no merging with upstream | 18:13 |
serialhex | cool cool | 18:16 |
blackburn | damn it is raining! and I have to go to met my girlfriend on bus station | 18:17 |
serialhex | ok, later!! | 18:17 |
blackburn | seems i'll get wet hehe | 18:17 |
serialhex | oh, and hopefully you wont get too wet | 18:17 |
blackburn | that summer is rainy | 18:19 |
blackburn | oh how cold was this winter hehe | 18:20 |
blackburn | later | 18:21 |
serialhex | sonney2k are you available? | 18:27 |
serialhex | ok, well i have a qick Q, for you... | 18:30 |
serialhex | the 2nd to last line in the python example you asked me to convert is: testerr=mean(sign(out)!=testlab) | 18:30 |
serialhex | and i'm kind of confused, as != in most langs i know of will simply return either true or false, though from the code it seems to want to be an arry one can find the average of... | 18:32 |
serialhex | so is that what you want? which is easy enough to code (i've added a lot of stuff to the example so it works) let me know, and i'll probably have it done by tonight (or sometime tomorrow if i let myself sleep) | 18:33 |
@sonney2k | serialhex, that will compute how often sign(out) deviates from testlab | 18:58 |
@sonney2k | (both are vectors) | 18:58 |
@sonney2k | and mean will copute the frequency | 18:58 |
blackburn | serialhex: there are some operations for vector '==' | 19:30 |
blackburn | all, mean, etc | 19:30 |
blackburn | numpy vector, of course* | 19:30 |
blackburn | sonney2k: your applet rocks! | 19:31 |
blackburn | i've got wet до ниточки | 19:31 |
blackburn | sonney2k: LLE is in pull requests | 19:33 |
-!- cwidmer_ [~quassel@connect.tuebingen.mpg.de] has joined #shogun | 19:42 | |
-!- cwidmer_ [~quassel@connect.tuebingen.mpg.de] has quit [Client Quit] | 19:44 | |
-!- blackburn [~blackburn@188.168.4.201] has quit [Quit: Leaving.] | 19:59 | |
-!- blackburn [~blackburn@188.168.5.44] has joined #shogun | 21:04 | |
blackburn | huh | 21:05 |
blackburn | sonney2k: sorry will remove Preproc | 21:06 |
blackburn | mistake while merging | 21:06 |
blackburn | sonney2k: fixed | 21:20 |
@sonney2k | blackburn, please rebase against master... | 21:23 |
blackburn | sonney2k: isn't it rebased already? | 21:24 |
@sonney2k | From git://github.com/SergeyLisitsyn/shogun | 21:24 |
@sonney2k | * branch lle -> FETCH_HEAD | 21:24 |
@sonney2k | Auto-merging src/modular/Preprocessor.i | 21:24 |
@sonney2k | CONFLICT (content): Merge conflict in src/modular/Preprocessor.i | 21:24 |
@sonney2k | Automatic merge failed; fix conflicts and then commit the result. | 21:24 |
blackburn | damn | 21:25 |
blackburn | sonney2k: ok now? rebased | 21:25 |
@sonney2k | nope | 21:26 |
blackburn | sorry | 21:28 |
blackburn | sonney2k: i've git pulled upstream master | 21:29 |
blackburn | then git checkout lle and git rebase master | 21:30 |
blackburn | got Current branch lle is up to date. | 21:30 |
blackburn | just don't know what I did wrong :/ | 21:31 |
@sonney2k | hmmhhh let me check | 21:32 |
blackburn | sonney2k: my fault | 21:34 |
blackburn | now will resolve conflict | 21:34 |
@sonney2k | all up-to-date | 21:35 |
blackburn | sonney2k: I just haven't noticed you added _includes file | 21:35 |
blackburn | sorry for my awkwardness | 21:36 |
@sonney2k | so shall I fix it here now? | 21:37 |
@sonney2k | blackburn, ^ | 21:38 |
blackburn | sonney2k: I will fix in a minute | 21:38 |
@sonney2k | kk | 21:38 |
blackburn | I have got broken commit scheme here, too many small commits | 21:39 |
blackburn | and manual merging because of preproc->preprocessor and so on | 21:39 |
blackburn | sonney2k: everything should be good now | 21:40 |
CIA-18 | shogun: Sergey Lisitsyn master * r2520983 / (6 files in 2 dirs): Cleaned and renamed LLE to LocallyLinearEmbedding (+14 more commits...) - http://bit.ly/lbMwPH | 21:42 |
@sonney2k | blackburn, thanks! | 21:43 |
blackburn | sonney2k: thanks for your patience ;) | 21:43 |
blackburn | now shogun has 2 more dimreduction algos | 21:44 |
blackburn | sonney2k: we should integrate your applet into shogun | 21:52 |
@sonney2k | cpufire? | 21:53 |
blackburn | aha | 21:53 |
blackburn | it is awesome | 21:53 |
blackburn | sonney2k: btw two days ago I've installed debian, then enabled testing repos and.. | 21:57 |
blackburn | it became as buggy as ubuntu 11.04 | 21:57 |
blackburn | so now I'm using 10.04 hehe | 21:57 |
serialhex | anyone here?? | 22:16 |
serialhex | blackburn?? sonney2k?? | 22:16 |
serialhex | hmm... guess not ;( | 22:17 |
serialhex | it's ok, i'll live (maybe :P ) | 22:17 |
blackburn | serialhex: here but going to sleep :) | 22:19 |
serialhex | ahh, ok.... well good night then! | 22:19 |
blackburn | tomorrow have to know many new things about math. physics equations | 22:20 |
blackburn | exam on 10th, kernel panic hehe | 22:20 |
serialhex | lolz! | 22:20 |
serialhex | i hope you do well! | 22:20 |
serialhex | tho i'm sure you will | 22:20 |
blackburn | I hope to get 4/5 and it would be good for me | 22:20 |
blackburn | these guys and girls here struggling for 5/5 - they aren't in gsoc :D | 22:21 |
blackburn | so marks is not sufficient or necessary :) | 22:21 |
serialhex | nope :P | 22:21 |
blackburn | well | 22:22 |
blackburn | serialhex: have a good evening | 22:22 |
serialhex | and you have a good night blackburn | 22:22 |
blackburn | thanks | 22:22 |
blackburn | see ya | 22:22 |
-!- blackburn [~blackburn@188.168.5.44] has quit [Quit: Leaving.] | 22:22 | |
@sonney2k | Re | 23:26 |
* sonney2k thinks blackburn should strive for 5/5 too - much more important than GSoC | 23:26 | |
CIA-18 | shogun: Soeren Sonnenburg master * r95b4248 / (5 files in 2 dirs): Merge branch 'master' of git://github.com/karlnapf/shogun - http://bit.ly/kHhEGi | 23:29 |
CIA-18 | shogun: Soeren Sonnenburg master * r5c3d0b9 / src/libshogun/lib/AsciiFile.cpp : Merge branch 'streaming' of git://github.com/frx/shogun - http://bit.ly/mr3RkQ | 23:31 |
@sonney2k | serialhex, wassup? | 23:50 |
serialhex | hey sonney2k, i just have a quick Q about the SVM example | 23:59 |
@sonney2k | shoot | 23:59 |
serialhex | for some reason the python example dosn't want to run on my computer so i can't check against that | 23:59 |
@sonney2k | ehh what does it do? | 23:59 |
--- Log closed Thu Jun 09 00:00:11 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!