| --- Log opened Thu Jul 14 00:00:18 2011 | ||
| blackburn | alesis-novik: how are you? | 00:03 |
|---|---|---|
| alesis-novik | Hey blackburn, same as always I guess :) | 00:03 |
| alesis-novik | You? | 00:04 |
| blackburn | fine, except of suffering some teethache :) | 00:04 |
| alesis-novik | ouch | 00:06 |
| alesis-novik | That's one of the worst things | 00:07 |
| blackburn | what's up with your EMs? anything cool? :) | 00:07 |
| blackburn | yeah, I was kinda careless again and will go to dentist soon | 00:08 |
| alesis-novik | Well, I'm working on Split-Merge EM now | 00:08 |
| blackburn | what is it? | 00:08 |
| alesis-novik | Also planing to clean up the examples | 00:08 |
| alesis-novik | Essentially it has a step when it splits one of the mixture components into 2 and merges 2 components into 1 (at the same time) | 00:09 |
| blackburn | ah | 00:09 |
| blackburn | I see | 00:09 |
| alesis-novik | Thus the number of components stays the same, but it can produce a better fit | 00:09 |
| alesis-novik | For example with the current EM implementation, it sometimes decides to fit a single gaussian over all of the data, which is wrong | 00:10 |
| alesis-novik | Hopefully Split-Merge EM will fix that | 00:10 |
| blackburn | split-merge isn't a 'classical' EM, right? | 00:11 |
| alesis-novik | No, I guess you could call it an extension | 00:12 |
| blackburn | I see | 00:12 |
| alesis-novik | It shouldn't take long to implement and it can use the same examples as the 'classical' EM | 00:14 |
| blackburn | nice | 00:15 |
| blackburn | it seems there are a lot of EMs, right? | 00:15 |
| alesis-novik | Well, there are many methods to make EM better along different dimensions for GMMs | 00:29 |
| alesis-novik | But EM generally is a method | 00:30 |
| blackburn | ah I see | 00:30 |
| alesis-novik | So it's used with many different things, not just GMMs | 00:30 |
| blackburn | okay late here | 00:32 |
| blackburn | alesis-novik: see you | 00:32 |
| alesis-novik | See you blackburn | 00:32 |
| -!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 00:32 | |
| -!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Read error: Connection reset by peer] | 01:04 | |
| -!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 01:06 | |
| -!- alesis-novik [~alesis@cpat001.wlan.net.ed.ac.uk] has quit [Quit: Leaving] | 02:32 | |
| -!- f-x [~user@117.192.199.83] has joined #shogun | 05:33 | |
| -!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 06:24 | |
| -!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 250 seconds] | 06:26 | |
| -!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 06:30 | |
| -!- in3xes [~in3xes@180.149.49.230] has quit [Client Quit] | 06:32 | |
| -!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Ping timeout: 276 seconds] | 06:41 | |
| -!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 06:42 | |
| -!- f-x [~user@117.192.199.83] has quit [Read error: Connection reset by peer] | 07:18 | |
| CIA-32 | shogun: Soeren Sonnenburg master * r140b139 / examples/undocumented/python_modular/library_fisher2x3_modular.py : add example for fisher test - http://bit.ly/neVWTu | 09:03 |
| -!- vojta [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 09:24 | |
| -!- vojta is now known as VojtechFranc | 09:24 | |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 09:34 | |
| -!- vojta [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 09:34 | |
| -!- vojta is now known as VojtechFranc | 09:34 | |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 09:34 | |
| -!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 10:27 | |
| -!- in3xes [~in3xes@180.149.49.227] has quit [Read error: Connection reset by peer] | 10:54 | |
| -!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 10:55 | |
| -!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 11:06 | |
| -!- blackburn [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 11:51 | |
| -!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 11:59 | |
| -!- in3xes [~in3xes@180.149.49.227] has quit [Remote host closed the connection] | 12:23 | |
| -!- in3xes_ is now known as in3xes | 12:44 | |
| -!- sploving1 [~sploving@124.16.139.134] has joined #shogun | 13:03 | |
| sploving1 | sonney2k, anything? | 13:04 |
| -!- blackburn1 [~blackburn@188.122.238.99] has joined #shogun | 13:15 | |
| -!- blackburn [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 13:17 | |
| @sonney2k | sploving1, yes | 13:19 |
| sploving1 | sonney2k, what is going on? | 13:19 |
| @sonney2k | sploving1, I was wondering what the state regarding the examples is? | 13:20 |
| @sonney2k | which are still missing? | 13:20 |
| sploving1 | most of the exmaples are added. just remaining some kernel examples. | 13:21 |
| sploving1 | I am working on them | 13:21 |
| sploving1 | now | 13:21 |
| sploving1 | maybe no more than ten about | 13:21 |
| @sonney2k | sploving1, ok that sounds good | 13:23 |
| @sonney2k | the only thiing left though is that you check if these examples give the same result like the python ones | 13:24 |
| @sonney2k | otherwise we won't know if they all work | 13:24 |
| @sonney2k | you can do this by printing out the numbers in both python / java and comparing them | 13:24 |
| sploving1 | I know. I will do this after adding these examples. | 13:26 |
| sploving1 | I think the result are the same. as the typemap works well. | 13:26 |
| @sonney2k | sploving1, yeah they should be... you should at least compare the most complex examples (as in that use many algorithms to return something) | 13:28 |
| @sonney2k | if these work for all the different feature types than the rest will likely too | 13:28 |
| @sonney2k | sploving1, when that is done I would suggest to start with lua? | 13:28 |
| @sonney2k | you were very familiar with lua right? | 13:29 |
| sploving1 | lua basic typemap are already work now | 13:29 |
| sploving1 | now just discuss with what library need support | 13:30 |
| sploving1 | I have no idea now. I will discuss it later with my mentor next irc meetin | 13:31 |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 13:33 | |
| @sonney2k | sploving1, I guess you should just use the existing lua tables... | 13:41 |
| sploving1 | maybe. As I found there is no other API suiable to lua matrix | 13:43 |
| @sonney2k | sploving1, and lists of strings work as is... sparse matrix is then just not supported (or one has to load this directly with shogun) | 13:46 |
| @sonney2k | sploving1, I will ask the torch5 guys if they are still developing their toolbox - if so we should try to interface with them. but the baseline is still basic types | 13:47 |
| sploving1 | okay. but its API docs is not well. just a little: http://torch5.sourceforge.net/manual/luaT/index-2.html | 13:52 |
| @sonney2k | sploving1, I sent them an email | 14:01 |
| sploving1 | sonney2k, thx. | 14:01 |
| sploving1 | sonney2k, in kernel_weighted_degree_position_string_modular.py example | 14:07 |
| sploving1 | from numpy import zeros,ones,float64,int32 | 14:07 |
| sploving1 | #kernel.set_shifts(zeros(len(fm_train_dna[0]), dtype=int32)) | 14:07 |
| sploving1 | #kernel.set_position_weights(ones(len(fm_train_dna[0]), dtype=float64)) | 14:07 |
| sploving1 | why use "#"? | 14:08 |
| sploving1 | does that mean it cannot work? | 14:08 |
| @sonney2k | sploving1, just to illustrate that there are more options | 14:19 |
| sploving1 | sonney2k, another thing, from shogun.Distribution import HMM, BW_NORMAL. then in java, ne need write: import ?? | 14:21 |
| @sonney2k | a static import like we have for DNA | 14:21 |
| sploving1 | import static org.shogun.EAlphabet.DNA; I mean import static org.shogun.?.BW_NORMAL; | 14:22 |
| sploving1 | I forgot grep. I will use it | 14:23 |
| -!- heiko [~heiko@main.uni-duisburg.de] has joined #shogun | 14:27 | |
| -!- f-x [~user@117.192.194.58] has quit [Read error: Connection reset by peer] | 14:35 | |
| sploving1 | sonney2k, public boolean set_shifts(SWIGTYPE_p_int shifts, int len) should be public boolean set_shifts(Doublematrix) { | 14:36 |
| sploving1 | I think WeightedDegreeStringKernel need change the set_shifts method | 14:36 |
| sploving1 | it is the same for the set_position_weights method | 14:37 |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 14:41 | |
| -!- f-x [~user@117.192.194.58] has quit [Read error: Connection reset by peer] | 14:50 | |
| @sonney2k | sploving1, this function needs to be converted for SGVector - so just assume that it works I can fix that later | 14:52 |
| sploving1 | sonney2k, ok | 14:52 |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 15:00 | |
| CIA-32 | shogun: Soeren Sonnenburg master * r32a8769 / (6 files in 5 dirs): | 15:19 |
| CIA-32 | shogun: Merge pull request #194 from karlnapf/master | 15:19 |
| CIA-32 | shogun: first working little python_modular example (+5 more commits...) - http://bit.ly/pbhsyk | 15:19 |
| @sonney2k | heiko, I am getting excited :) | 15:22 |
| heiko | yes, me too :) | 15:23 |
| heiko | the example is a little one | 15:23 |
| heiko | but hower | 15:23 |
| heiko | however | 15:23 |
| heiko | but i also got another problem: | 15:23 |
| blackburn1 | hey guys is "bioinformatics the machine learning approach" by baldi a good textbook? | 15:23 |
| heiko | swig/python detected a memory leak of type 'CDynamicObjectArray< shogun::CParameterCombination > *', no destructor found. | 15:24 |
| heiko | blackburn1, i do not know | 15:24 |
| blackburn1 | heiko: bad boy | 15:24 |
| heiko | lol :) | 15:24 |
| blackburn1 | :D | 15:24 |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 15:27 | |
| @sonney2k | why is heiko always causing so much trouble ;-) | 15:27 |
| heiko | hehe google pais me for this ;) | 15:28 |
| @sonney2k | blackburn1, I haven't read this one. Are you seriously interested in doing machine learning and bioinformatics at some point? | 15:28 |
| @sonney2k | heiko, too bad that google doesn't pay me for this ;-) | 15:28 |
| heiko | :D | 15:29 |
| sploving1 | blackburn1, do you know how to write a simple function that equal: fm_hmm_pos=[traindat[i] for i in where([label_traindat==1])[1] ] in kernel_top_modular.py | 15:29 |
| sploving1 | java function | 15:29 |
| heiko | sonney2k, this method, what does it mean? I mean, CDynamicObjectArray has a destructor | 15:30 |
| heiko | ehm message | 15:30 |
| heiko | not method | 15:30 |
| sploving1 | blackburn1, if there is not, I will use the for loop | 15:31 |
| CIA-32 | shogun: Soeren Sonnenburg master * r61647c8 / (2 files): convert set_position_weights and set_shifts to SGVector - http://bit.ly/qjo2VE | 15:33 |
| CIA-32 | shogun: Soeren Sonnenburg master * r4d8af57 / (6 files in 5 dirs): Merge branch 'master' of github.com:shogun-toolbox/shogun - http://bit.ly/pprfQj | 15:33 |
| CIA-32 | shogun: Soeren Sonnenburg master * r0849d96 / examples/undocumented/libshogun/modelselection_model_selection_parameters_test.cpp : | 15:34 |
| CIA-32 | shogun: Merge pull request #195 from karlnapf/master | 15:34 |
| CIA-32 | shogun: more tests - http://bit.ly/rh6fS1 | 15:34 |
| @sonney2k | sploving1, fixed | 15:35 |
| sploving1 | oh | 15:35 |
| @sonney2k | heiko, no idea | 15:36 |
| @sonney2k | when does it occur? | 15:36 |
| heiko | when i call CModelSelectionParameter::get_combinations() | 15:36 |
| @sonney2k | heiko, does that thing return a new object? | 15:36 |
| heiko | which creates a new instance of DynamicParameterCombinationArray | 15:36 |
| @sonney2k | so yes? | 15:36 |
| heiko | yes | 15:36 |
| @sonney2k | heiko, then you have to use the %newobject CModelSelectionParameter::get_combinations; directive | 15:37 |
| heiko | i have %newobject get_combinations(); | 15:37 |
| @sonney2k | this tells swig that this function creates a new object :) | 15:37 |
| heiko | do i need the class before this? | 15:37 |
| @sonney2k | it should work like this - could be that it needs to know the namespace | 15:37 |
| heiko | testing ... | 15:38 |
| heiko | oh yes, the classname fixed it :) | 15:38 |
| heiko | what about pure virtual methods | 15:39 |
| heiko | which classname to use? | 15:39 |
| heiko | the one of the implementing class? | 15:39 |
| blackburn1 | sonney2k: yeah, kinda interested | 15:41 |
| blackburn1 | sploving1: no, can't mind any non-loop solution | 15:42 |
| @sonney2k | heiko, http://www.swig.org/Doc2.0/Customization.html#Customization_ownership | 15:42 |
| heiko | ok will have a look, thanks | 15:44 |
| heiko | sonney2k, blackburn, anyone of you could give me another hint for swig stuff? | 15:56 |
| heiko | get_combinations returns in c++ CDynamicObjectArray<CParameterCombination>* | 15:56 |
| @sonney2k | heiko, regarding what? | 15:56 |
| heiko | a new instance | 15:56 |
| heiko | I can call this from python | 15:56 |
| @sonney2k | doesn't %newobject work now? | 15:56 |
| heiko | yes it does | 15:56 |
| heiko | but i cannot call methods on the object, the type is somehow wrong | 15:57 |
| heiko | <Swig Object of type 'CDynamicObjectArray< shogun::CParameterCombination > *' at 0x936ed58> | 15:57 |
| heiko | whereas temp=DynamicParameterCombinationArray gives | 15:57 |
| @sonney2k | heiko, did you use %template to define it? | 15:57 |
| heiko | yes | 15:57 |
| heiko | <shogun.Modelselection.DynamicParameterCombinationArray; proxy of <Swig Object of type 'shogun::CDynamicObjectArray< shogun::CParameterCombination > *' at 0x936ead0> > | 15:57 |
| @sonney2k | heiko, in the right namespace? | 15:58 |
| heiko | I used shogun | 15:58 |
| @sonney2k | hmmhh it is correct | 15:59 |
| @sonney2k | what code snippet does not work | 16:00 |
| @sonney2k | I cannot see it from what you pasted so far | 16:00 |
| heiko | ok | 16:00 |
| heiko | in the modelselection_parameter_tree_modular example | 16:01 |
| heiko | just add | 16:01 |
| heiko | combinations=root.get_combinations() | 16:01 |
| heiko | combinations.get_num_elements() | 16:01 |
| heiko | error is AttributeError: 'SwigPyObject' object has no attribute 'get_num_elements' | 16:01 |
| heiko | and the type of "combionations" is | 16:02 |
| heiko | <Swig Object of type 'CDynamicObjectArray< shogun::CParameterCombination > *' at 0x936ed58> | 16:02 |
| heiko | but it should be something like i get when i do: | 16:02 |
| heiko | test=DynamicParameterCombinationArray() | 16:02 |
| heiko | >>> test | 16:02 |
| heiko | <shogun.Modelselection.DynamicParameterCombinationArray; proxy of <Swig Object of type 'shogun::CDynamicObjectArray< shogun::CParameterCombination > *' at 0x936ead0> > | 16:02 |
| blackburn1 | uh it was a mistake I ran isomap on 10000 examples | 16:11 |
| heiko | sonney2k, got to go now, will come back later | 16:16 |
| -!- heiko [~heiko@main.uni-duisburg.de] has quit [Ping timeout: 258 seconds] | 16:21 | |
| -!- heiko [~heiko@main.uni-duisburg.de] has joined #shogun | 16:21 | |
| -!- heiko [~heiko@main.uni-duisburg.de] has quit [Client Quit] | 16:21 | |
| * sonney2k listens to blackburn1 harddisk seeking music | 16:27 | |
| -!- blackburn1 [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 16:30 | |
| -!- heiko [~heiko@main.uni-duisburg.de] has joined #shogun | 16:36 | |
| sploving1 | sonney2k, TOPFeatures(int size, SWIGTYPE_p_shogun__CHMM p, SWIGTYPE_p_shogun__CHMM n, boolean neglin, boolean poslin) | 16:44 |
| sploving1 | in TopFeatures.java, it should be TOPFeatures(int size, SWIGTYPE_p_shogun__CHMM p, SWIGTYPE_p_shogun__CHMM n, boolean neglin, boolean poslin) | 16:45 |
| sploving1 | TOPFeatures(int size, HMM p, HMM n, boolean neglin, boolean poslin) | 16:45 |
| @sonney2k | sploving1, features don't know about distributions - so that is the problem | 16:45 |
| @sonney2k | f-x, btw did you manage to run svmsgd2 on the challenge data? | 16:46 |
| sploving1 | sonney2k, can not fix it?? | 16:47 |
| @sonney2k | doing | 16:47 |
| sploving1 | okay | 16:49 |
| -!- sploving1 [~sploving@124.16.139.134] has left #shogun [] | 16:49 | |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 16:50 | |
| CIA-32 | shogun: Baozeng Ding master * r686bc79 / (4 files): added some kernel examples - http://bit.ly/nslH39 | 16:51 |
| CIA-32 | shogun: Soeren Sonnenburg master * re639ed1 / (4 files): | 16:51 |
| CIA-32 | shogun: Merge pull request #196 from sploving/master | 16:51 |
| CIA-32 | shogun: add some kernel example for java_modular - http://bit.ly/nzXcuY | 16:51 |
| heiko | sonney2k, do you have any idea for this type error? i am stuck with it :( | 16:51 |
| heiko | somehow, python does not recognize the return type as the defined template | 16:52 |
| -!- gsomix [~gsomix@85.26.233.160] has joined #shogun | 16:57 | |
| @sonney2k | heiko, I've never seen sth like this before | 17:21 |
| heiko | oh :( | 17:21 |
| heiko | i searched for examples where a templated newobject is returned, but found none | 17:22 |
| @sonney2k | I suspect that the %template definition is wrong | 17:22 |
| heiko | yes, me too, but which one is the correct one? | 17:23 |
| heiko | i tried adding several shogun:: , but does not help | 17:23 |
| CIA-32 | shogun: Soeren Sonnenburg master * r2f5bb6e / (9 files): make distributions known from features - http://bit.ly/nUtcT7 | 17:23 |
| @sonney2k | usually swig spits out warnings | 17:24 |
| heiko | mmh, ok, but however, this method is not SO important for now | 17:25 |
| heiko | i am putting the grid-search example together now, | 17:25 |
| heiko | there, no methods returns an array instance | 17:25 |
| heiko | so it should work :) | 17:25 |
| @sonney2k | http://www.swig.org/Doc1.3/SWIGPlus.html#SWIGPlus_renaming_templated_types_namespaces | 17:26 |
| @sonney2k | heiko, look at this %template(ABCXYZ) Space::ABC<Space::XYZ>; | 17:26 |
| @sonney2k | of course we have namespace around this so Space shouldn't be necessary I thought | 17:27 |
| heiko | I already tried this :( | 17:28 |
| -!- heiko [~heiko@main.uni-duisburg.de] has quit [Ping timeout: 258 seconds] | 17:34 | |
| -!- gsomix [~gsomix@85.26.233.160] has quit [Read error: Connection timed out] | 17:39 | |
| -!- gsomix [~gsomix@85.26.233.160] has joined #shogun | 17:39 | |
| @sonney2k | heiko! fixed! | 17:42 |
| CIA-32 | shogun: Soeren Sonnenburg master * rf87cfdd / (2 files in 2 dirs): | 17:43 |
| CIA-32 | shogun: change order of includes to fix | 17:43 |
| CIA-32 | shogun: <shogun.Modelselection.DynamicParameterCombinationArray; proxy of <Swig Object of type 'shogun::CDynamicObjectArray< shogun::CParameterCombination > *' at 0x936ead0> > | 17:43 |
| CIA-32 | shogun: problem. - http://bit.ly/p6PJoH | 17:43 |
| -!- gsomix [~gsomix@85.26.233.160] has quit [Ping timeout: 255 seconds] | 18:11 | |
| -!- [1]warpy [~warpy@bzq-79-181-19-104.red.bezeqint.net] has joined #shogun | 18:16 | |
| [1]warpy | hey guys anyone here? | 18:16 |
| [1]warpy | trying to solve a compilation error with 0.10. i am not sure why it quit "install -d -m755 //usr/local/lib | 18:16 |
| [1]warpy | install -d -m755 "//usr/local/include/shogun" | 18:16 |
| [1]warpy | install -m755 libshogun.so.9.0 //usr/local/lib | 18:16 |
| [1]warpy | install -m755 libshogun.a //usr/local/lib | 18:16 |
| [1]warpy | install: cannot stat `libshogun.a': No such file or directory | 18:16 |
| [1]warpy | make[1]: *** [install-libshogun] Error 1 | 18:16 |
| [1]warpy | make[1]: Leaving directory `/media/sf_ubunto_shared/shogun-0.10.0/src/libshogun' | 18:16 |
| [1]warpy | make: *** [install] Error 2 | 18:16 |
| [1]warpy | root@ori-VirtualBox:/media/sf_ubunto_shared/shogun-0.10.0/src# | 18:16 |
| [1]warpy | sorry, was going to paste one line.. | 18:16 |
| -!- gsomix [~gsomix@85.26.235.116] has joined #shogun | 18:24 | |
| -!- gsomix [~gsomix@85.26.235.116] has quit [Ping timeout: 255 seconds] | 18:38 | |
| -!- gsomix [~gsomix@85.26.232.212] has joined #shogun | 18:51 | |
| -!- [1]warpy [~warpy@bzq-79-181-19-104.red.bezeqint.net] has quit [Ping timeout: 252 seconds] | 19:13 | |
| -!- gsomix [~gsomix@85.26.232.212] has quit [Ping timeout: 255 seconds] | 19:27 | |
| -!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 19:46 | |
| -!- gsomix [~gsomix@85.26.233.55] has joined #shogun | 19:48 | |
| blackburn | sonney2k: another one include to make things work? :D | 19:51 |
| bettyboo | ha ha. | 19:51 |
| blackburn | sonney2k: meet gsomix | 19:51 |
| * gsomix away | 19:52 | |
| -!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 19:54 | |
| -!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 19:54 | |
| blackburn | okay my wi-fi deserves to die | 19:55 |
| -!- f-x [~user@117.192.194.58] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)] | 20:14 | |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 20:15 | |
| -!- f-x [~user@117.192.194.58] has left #shogun [] | 20:15 | |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 20:15 | |
| f-x | sonney2k: ran sgd2 and sgdqn on 100k test, 100k training examples | 20:19 |
| f-x | they are about 3-4% more successful than the shogun sgd at classification | 20:19 |
| @sonney2k | blackburn, yes | 20:22 |
| @sonney2k | even more includes :/ | 20:22 |
| @sonney2k | gsomix, welcome... | 20:28 |
| gsomix | hi | 20:28 |
| @sonney2k | f-x, even the batch SGD variant? | 20:29 |
| @sonney2k | f-x, if so we need to figure out why. | 20:29 |
| f-x | sonney2k: yes | 20:29 |
| @sonney2k | gsomix, so you are working with blackburn? | 20:29 |
| f-x | i'm looking into it now | 20:29 |
| gsomix | sonney2k, yep | 20:30 |
| @sonney2k | f-x, thanks | 20:30 |
| blackburn | sonney2k: if you have more 'tasks' let me or him know ;) | 20:31 |
| -!- [1]warpy [~warpy@bzq-79-181-19-104.red.bezeqint.net] has joined #shogun | 20:33 | |
| @sonney2k | gsomix, well c5.0 integration, the multiclass svms, lots of SGVector transitioning, nice visual examples for python modular (that generate figures or are interactive) | 20:33 |
| @sonney2k | blackburn, ^ | 20:33 |
| @sonney2k | or kdtree | 20:33 |
| @sonney2k | etc etc :) | 20:33 |
| blackburn | sonney2k: cover tree | 20:33 |
| @sonney2k | err yes | 20:33 |
| blackburn | sonney2k: now he is working on fibonaccy heap hehe | 20:33 |
| [1]warpy | hey soeren | 20:34 |
| [1]warpy | hey blackburn | 20:34 |
| blackburn | hey warpy ;) | 20:34 |
| @sonney2k | gsomix, just ask blackburn when you have questions. he knows everything :D | 20:34 |
| @sonney2k | hey | 20:34 |
| [1]warpy | i dove into github | 20:34 |
| bettyboo | you here? | 20:34 |
| [1]warpy | i am trying to add an upstream | 20:34 |
| [1]warpy | for upstream: url = git://github.com/octocat/shogun-toolbox.git or url = git://github.com/octocat/shogun.git doesnt work | 20:35 |
| gsomix | sonney2k, ok | 20:35 |
| [1]warpy | do you know the correct address? | 20:35 |
| blackburn | why are you trying to add upstream? | 20:35 |
| [1]warpy | it says so on the help page | 20:35 |
| blackburn | just clone your fork and you will have your upstream in git remotes | 20:35 |
| [1]warpy | you realise i have no idea what you just said | 20:36 |
| [1]warpy | i am like only a day into this | 20:36 |
| blackburn | :D | 20:36 |
| [1]warpy | i cloned my fork. sounds nasty though | 20:36 |
| [1]warpy | what is git remotes | 20:36 |
| [1]warpy | ah yes | 20:37 |
| blackburn | okay you have your local git repo | 20:37 |
| [1]warpy | look here at 2. | 20:37 |
| [1]warpy | http://help.github.com/fork-a-repo/ | 20:37 |
| [1]warpy | it says remotes (also upstream at the same time) | 20:37 |
| blackburn | ah | 20:37 |
| blackburn | okay | 20:37 |
| [1]warpy | so what is the correct address | 20:37 |
| blackburn | I'm sure your upstream is set already | 20:38 |
| blackburn | try | 20:38 |
| blackburn | git pull upstream master | 20:38 |
| [1]warpy | what does that do | 20:39 |
| [1]warpy | before i try it | 20:39 |
| [1]warpy | i dont want to screw things | 20:39 |
| [1]warpy | it has taken me a whole day to setup this mess | 20:39 |
| blackburn | right address should be git://github.com/shogun-toolbox/shogun.git | 20:40 |
| blackburn | with these command you just update your git repo with up-to-date upstream | 20:40 |
| blackburn | i.e. if sth is changed in shogun-toolbox/shogun - git pull upstream master will _pull_ these changes from _master_ branch | 20:41 |
| [1]warpy | i thought pull is a request to insert something to the master branch | 20:41 |
| [1]warpy | okay works | 20:42 |
| [1]warpy | i did git fetch upstream | 20:42 |
| blackburn | so upstream is set already | 20:42 |
| [1]warpy | no | 20:42 |
| [1]warpy | i manually set it | 20:42 |
| blackburn | ah | 20:42 |
| blackburn | okay | 20:42 |
| blackburn | strange btw, it should be set | 20:42 |
| [1]warpy | okay so now i am going to copy some files into it | 20:42 |
| blackburn | yes | 20:42 |
| [1]warpy | and then | 20:42 |
| blackburn | then just do git add for these files | 20:42 |
| blackburn | git commit -m 'Some message' | 20:43 |
| blackburn | and git push origin | 20:43 |
| [1]warpy | so commit is a limbo state ? | 20:43 |
| [1]warpy | and push is uploading ? | 20:43 |
| blackburn | yes | 20:43 |
| blackburn | with push you will upload your repo state to github | 20:43 |
| blackburn | with commit you just save it or so | 20:44 |
| @sonney2k | f-x, my guess is that I didn't get the learning rate or so right in shogun | 20:44 |
| @sonney2k | f-x, so I would look there | 20:44 |
| f-x | sonney2k: hmm. i'll begin from there | 20:44 |
| [1]warpy | hey soeren, i sent you two emails. the second one is accidental. did you see the first though ? | 20:45 |
| f-x | btw do you think the implicit normalization in the convert.py script can have any effect? | 20:45 |
| f-x | since the calibrate() function waits till the sum of features exceeds 1000, so it calibrates using many more vectors in the normalized form than the raw form | 20:46 |
| @sonney2k | [1]warpy, does that bug still exist? | 20:48 |
| @sonney2k | I just built and it is compiling fine | 20:49 |
| @sonney2k | ahh that was c# | 20:49 |
| @sonney2k | [1]warpy, yes the problem is the typemaps are just not correct | 20:52 |
| @sonney2k | they cause build failure | 20:52 |
| @sonney2k | f-x, well you can use the libsvm format and read it with shogun | 20:53 |
| @sonney2k | f-x, could very well bee | 20:53 |
| @sonney2k | be | 20:53 |
| -!- gsomix [~gsomix@85.26.233.55] has quit [Ping timeout: 252 seconds] | 20:53 | |
| f-x | sonney2k: okay, i'll try that too.. after some inspection of the code | 20:54 |
| blackburn | sonney2k: you forgot to remind we have castle siege in world of warcraft | 20:56 |
| @sonney2k | blackburn, wft? | 20:56 |
| [1]warpy | damnit | 20:57 |
| blackburn | sonney2k: I was denied to joke during meetings so have to joke now :D | 20:57 |
| [1]warpy | i realy need some expert help, we are driving full gas in neutral :) | 20:57 |
| bettyboo | ho | 20:57 |
| [1]warpy | soeren what about the c# warnings | 20:57 |
| [1]warpy | about ref -> _ref | 20:58 |
| [1]warpy | and string -> _string | 20:58 |
| [1]warpy | etc | 20:58 |
| @sonney2k | these are harmless | 20:59 |
| @sonney2k | [1]warpy, the problem really is that typemaps are just not there yet and it is more difficult to do them then for other scripting languages | 21:00 |
| @sonney2k | I haven't found any how to extend c# from C# native interface docu | 21:00 |
| -!- heiko [~heiko@134.91.10.200] has joined #shogun | 21:01 | |
| [1]warpy | does the java thing compile completely ? | 21:02 |
| @sonney2k | yes | 21:03 |
| [1]warpy | ouch | 21:04 |
| @sonney2k | not all examples work but 90% of them by now | 21:04 |
| [1]warpy | i thought the conversion will go smoothly | 21:04 |
| @sonney2k | what is missing is things like how to access a csharp array from C | 21:04 |
| @sonney2k | how to get the pointer to all those values ... | 21:04 |
| [1]warpy | in my typemap ? or in swig ? | 21:05 |
| [1]warpy | btw do we need the data repository to test the examples ? | 21:05 |
| @sonney2k | in csharp_modular/swig_typemaps.i | 21:06 |
| [1]warpy | also if we are on the subject, i didnt see any sparse matrix support in the java code, is it something that needs to be added or a normal matrix can be used instead? | 21:07 |
| @sonney2k | [1]warpy, I had a closer look on these typemaps and from what I see there is that these all are not yet even close to work | 21:08 |
| @sonney2k | [1]warpy, well sparse matrix is not so important I would say | 21:08 |
| @sonney2k | in the worst case you convert a dense into a sparse one | 21:08 |
| @sonney2k | or load it directly with shogun's functions into shogun objects | 21:08 |
| [1]warpy | but libsvm and svmlight natively use sparse matrices in their input files. | 21:09 |
| [1]warpy | unless its because that shogun can handle those billion examples ? | 21:10 |
| blackburn | haha guys from scikit-learn suffering LLE instability just like me, what a bad algo to work with | 21:10 |
| @sonney2k | [1]warpy, well shogun has i/o support for such files | 21:11 |
| [1]warpy | so shogun reads sparse then converts to dense and sends them to the cpp code ? | 21:11 |
| [1]warpy | blackburn, i dont think it updated the github. how can i tell | 21:11 |
| [1]warpy | https://github.com/orico/shogun | 21:12 |
| -!- f-x [~user@117.192.194.58] has quit [Read error: Connection reset by peer] | 21:12 | |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 21:12 | |
| blackburn | [1]warpy: have you made commit and then git push? | 21:12 |
| @sonney2k | [1]warpy, no shogun reads sparse and it stays sparse | 21:13 |
| @sonney2k | but you can send dense and shogun has a function to convert to sparse | 21:13 |
| [1]warpy | but it will know how to use normal matrices from the type map as sparse ? (sorry if i sound too newbie) | 21:14 |
| [1]warpy | blackburn, i did a commit when i was root | 21:14 |
| [1]warpy | then i went to my own user and did push because it didnt let me | 21:14 |
| [1]warpy | but i think the commit was only good for good and it didnt stick to the user | 21:14 |
| blackburn | strange | 21:14 |
| [1]warpy | if i commit on the user i will get this | 21:14 |
| [1]warpy | fatal: could not open '.git/COMMIT_EDITMSG': Permission denied | 21:15 |
| [1]warpy | bad learning curve | 21:15 |
| blackburn | you would rather work with git with either root or your user ;) | 21:16 |
| [1]warpy | okay so | 21:17 |
| [1]warpy | with root | 21:17 |
| [1]warpy | i get this | 21:17 |
| [1]warpy | no changes added to commit (use "git add" and/or "git commit -a") | 21:17 |
| [1]warpy | any ideas? | 21:18 |
| [1]warpy | if i push origin | 21:18 |
| [1]warpy | i get this | 21:18 |
| [1]warpy | Permission denied (publickey). | 21:18 |
| [1]warpy | fatal: The remote end hung up unexpectedly | 21:18 |
| [1]warpy | ah yes | 21:18 |
| [1]warpy | because i am root and i didnt add the root ssh key | 21:19 |
| [1]warpy | okay so i am returning to the user now. | 21:19 |
| blackburn | uff | 21:19 |
| [1]warpy | still it cant access .git/commit editmsg | 21:19 |
| [1]warpy | any ideas? | 21:19 |
| blackburn | remove it | 21:19 |
| blackburn | with sudo | 21:19 |
| blackburn | or change rights | 21:20 |
| blackburn | whatever you want | 21:20 |
| [1]warpy | removed. committed successfuly. pushed. | 21:20 |
| [1]warpy | it says everything upto data | 21:21 |
| [1]warpy | date* | 21:21 |
| blackburn | your fork is not up-to-date, last commit - 3 days ago | 21:21 |
| blackburn | did you made git push origin? | 21:21 |
| blackburn | ah nevermind | 21:22 |
| blackburn | git push should work too | 21:22 |
| [1]warpy | both just now | 21:22 |
| [1]warpy | everything up to date | 21:22 |
| blackburn | git pull upstream master | 21:22 |
| blackburn | git push master | 21:22 |
| blackburn | ? | 21:22 |
| [1]warpy | master is not a git repository for the second command | 21:23 |
| [1]warpy | i dont think i want to pull | 21:23 |
| blackburn | git push origin master | 21:23 |
| [1]warpy | yeah i did that | 21:23 |
| [1]warpy | first thing | 21:23 |
| [1]warpy | cd .. | 21:23 |
| [1]warpy | :) | 21:23 |
| blackburn | i have no idea what is wrong | 21:24 |
| [1]warpy | why cant they make it friendly | 21:24 |
| [1]warpy | okay look at this | 21:25 |
| [1]warpy | #examples/undocumented/csharp_modular/tests_check_commwordkernel_memleak_modular.cs | 21:25 |
| [1]warpy | no changes added to commit (use "git add" and/or "git commit -a") | 21:25 |
| -!- gsomix [~gsomix@85.26.233.55] has joined #shogun | 21:26 | |
| heiko | blackburn, just to be sure: how do i handle this in swig if i get a method that returns a SGVector<> with a newly created array in it? | 21:32 |
| blackburn | heiko: I'm not sure but afaik you should set free flag to true | 21:32 |
| blackburn | if you want it to be cleared after usage in swig | 21:32 |
| -!- f-x` [~user@117.192.194.58] has joined #shogun | 21:33 | |
| heiko | ok alright thanks | 21:33 |
| @sonney2k | gsomix, is this code you added released under the GPL? | 21:34 |
| @sonney2k | ahh heiko good evening! | 21:35 |
| -!- f-x [~user@117.192.194.58] has quit [Ping timeout: 260 seconds] | 21:35 | |
| heiko | sonney2k, hi! | 21:35 |
| @sonney2k | heiko, it was just a matter of flipping order of includes | 21:35 |
| -!- f-x` is now known as f-x | 21:35 | |
| heiko | oh no -.- | 21:35 |
| @sonney2k | the %template and include were after the class was %included | 21:35 |
| @sonney2k | so it was too late | 21:35 |
| -!- f-x is now known as Guest10512 | 21:36 | |
| heiko | ok i did not know that the order is important | 21:36 |
| heiko | but well | 21:36 |
| heiko | did you already change it? | 21:36 |
| -!- Guest10512 is now known as f-x` | 21:37 | |
| @sonney2k | heiko, yes | 21:37 |
| blackburn | sonney2k: code from here http://www.cosc.canterbury.ac.nz/tad.takaoka/alg/spalgs/spalgs.html | 21:38 |
| blackburn | sonney2k: no license, are we able to use it? | 21:38 |
| heiko | sonney2k, ok cool :) | 21:38 |
| gsomix | sonney2k, there is no license for original code. | 21:39 |
| * gsomix so slow. =___= | 21:39 | |
| @sonney2k | you should in this case ask the original author | 21:40 |
| blackburn | :D | 21:40 |
| blackburn | as I thought | 21:41 |
| -!- in3xes_ [~in3xes@180.149.49.230] has joined #shogun | 21:44 | |
| [1]warpy | sonney2k, so now i have my own fork and its updated with the cs files. | 21:48 |
| -!- in3xes_ [~in3xes@180.149.49.230] has quit [Quit: Leaving] | 21:53 | |
| @sonney2k | [1]warpy, the problem of course is that typemaps are totally not working :) | 21:54 |
| [1]warpy | yeah, to be honest i am not entirely sure we can make them work | 21:57 |
| [1]warpy | we just dont have enough swig experience | 21:58 |
| heiko | sonney2k, just sent another pull request | 21:58 |
| heiko | @all i am going to bed, good night! see you tomorrow! | 21:58 |
| blackburn | heiko: do you sleep at university? | 21:59 |
| blackburn | :D | 21:59 |
| heiko | yes, going to the roof now :) | 21:59 |
| blackburn | seriously you have no internet access at home ;) | 21:59 |
| heiko | i go to a café in cases of emergency :) | 22:00 |
| bettyboo | ;) | 22:00 |
| heiko | bye! | 22:00 |
| blackburn | see you | 22:00 |
| @sonney2k | heiko, thanks and cu tomorrow! | 22:04 |
| -!- heiko [~heiko@134.91.10.200] has quit [Ping timeout: 258 seconds] | 22:04 | |
| @sonney2k | [1]warpy, you don't need swig experience | 22:04 |
| @sonney2k | [1]warpy, you just need to figure out how to extend c# with C code | 22:04 |
| @sonney2k | [1]warpy, the native interface extensions - like the java JNI docs | 22:05 |
| @sonney2k | [1]warpy, if you find them things are trivial | 22:05 |
| CIA-32 | shogun: Heiko Strathmann master * rd5988ef / (10 files in 5 dirs): (log message trimmed) | 22:07 |
| CIA-32 | shogun: Squashed commit of the following: | 22:07 |
| CIA-32 | shogun: commit 116d110c03a9a4c394ecfbf52f11bb4eaac0d31e | 22:07 |
| CIA-32 | shogun: Author: Heiko Strathmann <heiko.strathmann@gmail.com> | 22:07 |
| CIA-32 | shogun: Date: Thu Jul 14 21:40:53 2011 +0200 | 22:07 |
| CIA-32 | shogun: -removed wrong SG_REF | 22:07 |
| CIA-32 | shogun: -usage of SGVector constructor to create array | 22:07 |
| CIA-32 | shogun: Soeren Sonnenburg master * rb29e944 / (10 files in 5 dirs): | 22:07 |
| CIA-32 | shogun: Merge pull request #198 from karlnapf/master | 22:07 |
| CIA-32 | shogun: adding classes to Evaluation swig interface and doing some DynArray cleanups - http://bit.ly/omJkUk | 22:07 |
| [1]warpy | the problem is that daniel has started working and he can only give a few hours a week to this. maybe less. and i am running several projects in parallel. | 22:08 |
| [1]warpy | i simply dont have the time to do it al | 22:09 |
| @sonney2k | [1]warpy, it is not about time it is about documentation w/o pointers I cannot help - not at all. | 22:13 |
| [1]warpy | can you elaborate on the "the native interface extensions - like the java JNI docs" is it in one of the chapters in the swig documentation ? | 22:15 |
| @sonney2k | [1]warpy, I am talking about c# | 22:16 |
| @sonney2k | has nothing to do with swig | 22:16 |
| @sonney2k | I need the JNI docs equivalent for c# | 22:17 |
| [1]warpy | okay will look into it | 22:18 |
| [1]warpy | so the java one is this : http://download.oracle.com/javase/1.4.2/docs/guide/jni/ | 22:19 |
| [1]warpy | or similar for whatever version they are at now ? | 22:19 |
| @sonney2k | yes | 22:23 |
| [1]warpy | soeren i believe daniel was using this page to do the conversions : http://www.swig.org/Doc1.3/CSharp.html#csharp_differences_java | 22:26 |
| -!- gsomix [~gsomix@85.26.233.55] has quit [Ping timeout: 255 seconds] | 22:30 | |
| [1]warpy | sonney2k, i believe this is what we need http://en.wikipedia.org/wiki/Platform_Invocation_Services | 22:37 |
| -!- f-x` [~user@117.192.194.58] has quit [Read error: Connection reset by peer] | 22:38 | |
| [1]warpy | examples of usage here : http://www.codeproject.com/KB/cs/marshalCPPclass.aspx | 22:38 |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 22:38 | |
| -!- f-x [~user@117.192.194.58] has quit [Remote host closed the connection] | 22:43 | |
| -!- warpy2 [~androirc@bzq-79-181-19-104.red.bezeqint.net] has joined #shogun | 22:44 | |
| warpy2 | anyone still alive | 22:45 |
| -!- f-x [~user@117.192.194.58] has joined #shogun | 22:45 | |
| --- Log closed Fri Jul 15 00:00:21 2011 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!