--- Log opened Wed May 18 00:00:09 2011 | ||
-!- blackburn [~qdrgsm@188.168.4.130] has quit [Quit: Leaving.] | 00:48 | |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 03:02 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 258 seconds] | 05:59 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 06:32 | |
-!- blackburn [~qdrgsm@188.168.4.186] has joined #shogun | 10:11 | |
-!- blackburn [~qdrgsm@188.168.4.186] has quit [Quit: Leaving.] | 10:26 | |
@sonney2k | sploving, around? | 10:28 |
---|---|---|
sploving | yeap | 10:28 |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 10:28 | |
sploving | I pushed the request | 10:28 |
sploving | sonney2k, I am around | 10:28 |
@sonney2k | sploving, great | 10:31 |
@sonney2k | sploving, the vector code already works for all the different types right? | 10:31 |
@sonney2k | I mean bool, double,byte,float etc? | 10:31 |
sploving | sonney2k, I have no such test examples | 10:32 |
sploving | So I am not sure | 10:33 |
sploving | It could support all of these. | 10:33 |
sploving | if there is no special | 10:33 |
@sonney2k | sploving, ok | 10:33 |
@sonney2k | there is not | 10:34 |
sploving | for the matrix | 10:34 |
sploving | My typemap does not work well | 10:34 |
@sonney2k | uint8_t should map to byte btw | 10:34 |
@sonney2k | sploving, what is the problem with matrix? | 10:35 |
sploving | sonney2k, uint8_t map byte, then what about char | 10:36 |
@sonney2k | char too | 10:36 |
@sonney2k | at least they are both 1 byte in size | 10:36 |
sploving | I will correct it | 10:37 |
@sonney2k | char is signed on some platforms though... | 10:37 |
@sonney2k | anyways how does one create a jblas matrix and how does it look underneath | 10:38 |
sploving | sonney2k, jblas matrix is use array to construct it | 10:38 |
sploving | one dim array | 10:39 |
@sonney2k | sploving, perfect then | 10:39 |
@sonney2k | so it is just double[] ? | 10:39 |
sploving | so it would not confict with double[][] | 10:39 |
sploving | sonney2k, yeap | 10:39 |
@sonney2k | sploving, so you only need to test if the class is indeed a jblas matrix class and then get get the pointer to the matrix | 10:40 |
sploving | I know. but for now the matrix typemap does not work well. sonney2k | 10:41 |
@sonney2k | sploving, what does that mean? | 10:41 |
@sonney2k | what is the problem? | 10:41 |
sploving | sonney2k, when I run "ant run", it shows the following error: | 10:41 |
sploving | \ | 10:41 |
sploving | java_modular/shogun/IntFeatures.java:96 could not find the simbol | 10:42 |
sploving | simbole: constructer: IntMatrix(int[][],boolean) | 10:43 |
sploving | location: shogun.IntMatrix | 10:43 |
sploving | [javac] return new IntMatrix(FeaturesJNI.IntFeatures_get_feature_matrix__SWIG_1(swigCPtr, this), true); | 10:43 |
sploving | If you have time, you just have a try(need add the TYPEMAP_SGMATRIX macros) | 10:44 |
@sonney2k | sploving, but it is reight there is no constructor with int[][] | 10:44 |
@sonney2k | only int[] | 10:44 |
sploving | not int[], all of the types, double, byte[] and so no | 10:45 |
@sonney2k | sure | 10:45 |
sploving | sonney2k, you mean Ipathelement location="../../../jblas/ | 10:46 |
sploving | not pathelement location="../../../jblas/jblas-1.2.1.jar" | 10:46 |
@sonney2k | sploving, yeah you shouldn't add this | 10:47 |
@sonney2k | sploving, I mean you should assume that jblas.jar is in the CLASSPATH | 10:47 |
sploving | just cancel this line? | 10:47 |
@sonney2k | sploving, yes | 10:48 |
@sonney2k | remove it | 10:48 |
sploving | ok | 10:48 |
@sonney2k | in your MatrixTest example you use double y[][] etc - and then x.set_feature_matrix(y) | 10:49 |
@sonney2k | is this what does not work? | 10:49 |
sploving | yeap. | 10:49 |
@sonney2k | sploving, but you have a typemap for this | 10:50 |
@sonney2k | ? | 10:50 |
sploving | sonney2k, yeap. | 10:51 |
sploving | the typemap is map double[][] in the SGMatrix<float_64_t> | 10:52 |
@sonney2k | yeah but look at return new ByteMatrix(FeaturesJNI.ByteFeatures_get_feature_matrix__SWIG_1(swigCPtr, this), true); | 10:54 |
@sonney2k | this really does not exist | 10:54 |
sploving | sonney2k, removed the line and push it | 10:54 |
sploving | sonney2k, yeap. this confuse me | 10:54 |
sploving | I donot know why swig genereated such function | 10:54 |
@sonney2k | it is in the out typemap | 10:55 |
@sonney2k | I mean public short[][] get_feature_matrix() { | 10:55 |
sploving | sonney2k,if FeaturesJNI.ByteFeatures_get_feature_matrix__SWIG_1(swigCPtr, this) return a long, then it will be OK | 10:57 |
sploving | Maybe my %typemap(jni) shogun::SGMatrix<SGTYPE>%{jobjectArray%} | 10:57 |
sploving | %typemap(jtype) shogun::SGMatrix<SGTYPE>%{JTYPE[][]%} | 10:57 |
sploving | %typemap(jstype) shogun::SGMatrix<SGTYPE> %{JTYPE[][]%} | 10:57 |
sploving | maybe there is somemistake | 10:58 |
sploving | sonney2k, not discuss the typemap, just the examples, do you think my examples is OK? | 10:59 |
sploving | or we juse RealMatrix in Java instead of double[][] | 11:00 |
@sonney2k | sploving, DoubleMatrix is from jblas right? | 11:00 |
sploving | sonney2k, yeap | 11:00 |
@sonney2k | then it looks all good | 11:01 |
@sonney2k | indeed the in typemap causes this thing to be generated | 11:01 |
@sonney2k | sploving, yes I think %typemap(jtype) shogun::SGMatrix<SGTYPE> %{JTYPE[][]%} is the problem | 11:02 |
sploving | sonney2k, I will try to correct it | 11:03 |
@sonney2k | sploving, what is this for? do you really need it? | 11:04 |
sploving | sonney2k, jtype provides the default mapping of types from C/C++ to Java for use in the native functions in the intermediary JNI class | 11:06 |
@sonney2k | still to abstract to me | 11:09 |
@sonney2k | too | 11:09 |
sploving | sonney2k, for exmaple, a c++ function, we call A(arg1), after wrapper, we generated the following function: | 11:12 |
CIA-31 | shogun: Baozeng Ding master * rc88eb49 / (5 files in 3 dirs): put the examples into the undocumented/java_modular - http://bit.ly/klaW6p | 11:13 |
CIA-31 | shogun: Baozeng Ding master * rda5ef18 / src/java_modular/build.xml : add jblas classpath - http://bit.ly/kM0iNA | 11:13 |
CIA-31 | shogun: Baozeng Ding master * rcac23b3 / src/java_modular/build.xml : remove the jblas classpth - http://bit.ly/mTycKs | 11:13 |
CIA-31 | shogun: Soeren Sonnenburg master * r96dc253 / (5 files in 3 dirs): Merge branch 'master' of git://github.com/sploving/shogun - http://bit.ly/ieXt8O | 11:13 |
sploving | wrapper function: B(argB), JNI funcion C(argC), Java funcion D(argD) | 11:13 |
sploving | the jni typemap is to map arg1 to arg B; jtype is to map arg1 to arg C ; and jstyp map arg1 to argD | 11:14 |
sploving | when we call the C++ function in Java, it calles like this: D call C, C call B, B call A | 11:15 |
@sonney2k | I understand | 11:16 |
sploving | sonney2k, maybe I should use RealMatrix in jtype(I am not sure now) | 11:17 |
@sonney2k | but RealMatrix is the same as SGMatrix<float64_t> | 11:18 |
@sonney2k | (which you use) | 11:18 |
sploving | sonney2k, http://www.swig.org/Doc2.0/SWIGDocumentation.html#Java_templates | 11:22 |
sploving | RealMatrix could be used in Java but SGMatrix<float64_t> could not be used in Java | 11:23 |
sploving | sonney2k, so they are a bit different | 11:23 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 240 seconds] | 11:24 | |
@sonney2k | sploving, yes but we are not in java but the typemap still... | 11:24 |
@sonney2k | sploving, anyway you can try... | 11:24 |
sploving | sonney2k, I will dive it later. see you I need to have supper | 11:25 |
@sonney2k | sploving, any progress? | 13:04 |
-!- blackburn [~qdrgsm@188.168.4.250] has joined #shogun | 18:24 | |
serialhex | oi blackburn, how are you today? | 18:31 |
blackburn | hi, fine, thanks, what about you? | 18:31 |
serialhex | pretty good, finishing my logic homework for tonight | 18:31 |
blackburn | I have had some test on control theory today | 18:32 |
blackburn | as part of eh... pass-fail exam? | 18:32 |
serialhex | cool, how do you think you did? | 18:32 |
blackburn | think it will be ok :) | 18:33 |
serialhex | ("i took a test on control theory today" btw :P ) | 18:33 |
serialhex | cool | 18:33 |
blackburn | is past perfect not right in this case? | 18:34 |
serialhex | i want to start taking some courses in control & information theory .... | 18:34 |
blackburn | I don't like it but I don't have any choice :) we don't have any optional courses | 18:35 |
serialhex | erm, asking about "past perfect participle presentations purposefully pissing on passerbys" is rediculous, i just know what sounds good and how the language should flow, and while i understood you, it didnt flow right :D | 18:35 |
blackburn | have to leave now for ~ an half a hour (to eat sth) :) | 18:36 |
blackburn | I see | 18:36 |
serialhex | ok, later | 18:36 |
serialhex | hey sonney2k, are you on?? | 18:37 |
blackburn | I would think this way too if I have it as everyday language, but I haven't :) | 18:37 |
blackburn | so I have to think sometimes about tenses and other shit | 18:37 |
serialhex | it's ok... m,aybe you'll visit the US or england sometime and be able to practice more :D | 18:37 |
blackburn | aha | 18:37 |
blackburn | here | 19:22 |
serialhex | how was dinner? | 19:43 |
serialhex | fscking mosquito bites!!!! | 19:43 |
blackburn | serialhex: it was gorgeous because I haven't had any food for a long :D | 19:47 |
serialhex | really? wtf are you doing not eating?? food rocks! :P | 19:48 |
@bettyboo | 8) | 19:48 |
serialhex | is it poor-college-student syndrome (aka being dead broke) | 19:48 |
blackburn | serialhex: well I was at my university and just hadn't time :) | 19:48 |
@bettyboo | ^_^ | 19:48 |
serialhex | ahh, i see... i've had that happen before | 19:49 |
serialhex | learning new smilieys eh bettyboo? | 19:49 |
@bettyboo | serialhex: I think at first I will implement some algo in python to get better comprehension | 19:49 |
serialhex | cool, good for you! | 19:49 |
blackburn | bettyboo: will you? | 19:49 |
@bettyboo | blackburn: eh, how old are you? though ~35? | 19:49 |
serialhex | lolz | 19:50 |
blackburn | bettyboo: ehhh I'm 20 :D | 19:50 |
@bettyboo | blackburn: with 4320 additions :D | 19:50 |
blackburn | ahahahahah | 19:50 |
serialhex | hahahahaha!!!! | 19:50 |
blackburn | that rocks | 19:50 |
serialhex | yeah, she is something else! | 19:50 |
blackburn | bettyboo: what kind of additions I have? | 19:51 |
@bettyboo | blackburn: what exams you will have? | 19:51 |
@bettyboo | *oink* | 19:51 |
blackburn | bettyboo: I will have exams on equation of math.ph., parallel computing and numerical methods, what about you? :D | 19:52 |
@bettyboo | blackburn: I will have no problem with method scoring tp,tn,..., but the number of classes will be >15 :) | 19:52 |
serialhex | heh, talking with her is very interesting sometimes! | 19:52 |
blackburn | exactly :D | 19:53 |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 20:37 | |
@sonney2k | back | 20:46 |
serialhex | sonney2k: i've been having trouble compiling shogun, the last few lines of the make are here: http://pastebin.com/utxzKJRi | 20:58 |
@sonney2k | serialhex, which versio n of gcc? | 21:01 |
serialhex | gcc version 4.6.0 20110429 (prerelease) (GCC) | 21:02 |
@sonney2k | aha! | 21:02 |
serialhex | ooh!!! something wrong with that version?? | 21:02 |
@sonney2k | it is just newer than mine and newer versions are usually more strict | 21:05 |
serialhex | ahh, i see, yeah i never saw the "cc1plus: warning: command line option ‘-Wimplicit’ is valid for C/ObjC but not for C++ [enabled by default]" thing before, and it says that almost every line when i make | 21:06 |
serialhex | yeah, every other line, it has that warning then stuff it's actually making | 21:07 |
@sonney2k | I don't understand the error though | 21:12 |
serialhex | nor do i, it confuses me! | 21:13 |
serialhex | and i've tried it a bunch of times, it always stops there, with the same error (even with different configure opts) | 21:14 |
@sonney2k | serialhex, could you please try the following: | 21:15 |
@sonney2k | in base/SGObject.h change the SG_REF/UNREF lines to | 21:16 |
@sonney2k | #define SG_REF(x) { if (x) x->ref(); } |U | 21:16 |
@sonney2k | #define SG_UNREF(x) { if (x) { if (x->unref()==0) x=NULL; } } |N | 21:16 |
serialhex | are the |U && |N newlines? | 21:18 |
@sonney2k | ignore these - copy & paste error | 21:19 |
serialhex | i figured so... ok, making now | 21:19 |
serialhex | hmm, new error | 21:19 |
serialhex | sonney2k: http://pastebin.com/nPERwbgU | 21:20 |
serialhex | make clean-ing and trying again | 21:20 |
* serialhex gets some tea as this will take a minute or 30 :P | 21:21 | |
@sonney2k | hmmhh I am tryitng to compile again too | 21:21 |
CIA-31 | shogun: Soeren Sonnenburg master * r0fe924e / src/libshogun/base/SGObject.h : potentially fix a gcc 4.6 compilation error - http://bit.ly/mpIhB2 | 21:25 |
@sonney2k | serialhex, please try again | 21:25 |
serialhex | pull upstream then try sonney2k? | 21:26 |
@sonney2k | serialhex, yes | 21:26 |
serialhex | done! make clean && make-ing now | 21:27 |
blackburn | what's up here | 21:31 |
blackburn | hehe | 21:31 |
@bettyboo | *g* :) | 21:32 |
blackburn | () caused an error? | 21:32 |
serialhex | trying to fix shogun so it works w/ gcc 4.6 (i think thats the prob) | 21:32 |
serialhex | nope sonney2k, it gives the same error as before, about "features/AttributeFeatures.cpp: In destructor ‘virtual shogun::CAttributeFeatures::~CAttributeFeatures()’: | 21:36 |
serialhex | features/AttributeFeatures.cpp:25:3: error: using temporary as lvalue [-fpermissive]" | 21:36 |
@sonney2k | blackburn, yes I think it is some ( ) problem | 21:45 |
* serialhex waits patiently for solutions to try | 21:51 | |
blackburn | fk! I forgot what fix I should do for shogun | 21:52 |
blackburn | *facepalm* | 21:52 |
serialhex | heh, did you say it here sometime?? you can probably look through the archives if that is the case | 21:54 |
blackburn | nope! | 21:55 |
blackburn | just have thinked about it this morning | 21:55 |
blackburn | damn I wonder with myself :D | 21:55 |
serialhex | paper + pen + thought + writing = note taking: very effectve for not forgetting things :P | 21:58 |
blackburn | yeaap | 21:58 |
blackburn | oh! it is roc examples | 22:07 |
blackburn | sonney2k: have a minute? | 22:07 |
@sonney2k | serialhex and if you use #define SG_UNREF(x) { if (x) { if (((x)->unref())==0) x=NULL; } } ? | 22:08 |
@sonney2k | serialhex, if that does not work could you please write out { if (x) { if (((x)->unref())==0) x=NULL; } } in the attribute features code | 22:09 |
@sonney2k | blackburn, yes? | 22:09 |
blackburn | sonney2k: can example for python_modular use matplotlib and so on? | 22:09 |
@sonney2k | blackburn, I would prefer one without and one with in the graphical/ subdir | 22:10 |
serialhex | trying now sonney2k | 22:11 |
@sonney2k | blackburn, note that the graphical ones are not tested in the test suite... | 22:11 |
blackburn | sonney2k: aha I see now, thank you | 22:11 |
serialhex | fscking shit! trying the 2nd thing... | 22:20 |
serialhex | sonney2k: so do you want me to do the #define SG_UNREF... stuff? or { if (features[i].attr_obj) { if (((features[i].attr_obj)... ??? | 22:22 |
serialhex | ...in the attribute features file | 22:23 |
@sonney2k | serialhex, no just write out the code | 22:23 |
serialhex | so just the { if (x) { if (((x)->unref())==0) x=NULL; } }? stick it in there... | 22:25 |
serialhex | i'm kinda confused with what you're asking me to so sonney2k :-/ i just tried putting the "#define ..." and i get the same error, and just the stuff in {} gives me a "features/AttributeFeatures.cpp:14:2: error: expected unqualified-id before ‘{’ token" error | 22:29 |
@sonney2k | serialhex, yes the { if (... stuff | 22:30 |
@sonney2k | but each bracket / clause on a separate line such that we see what is the cause | 22:31 |
@sonney2k | I mean the compiler will hopefully tell you then | 22:31 |
serialhex | https://gist.github.com/979494 | 22:38 |
serialhex | it says that the (features[i].attr_obj)=NULL; line is the one that isnt working right | 22:38 |
serialhex | (i think you might be able to edit that sonney2k, which should make life easier for both of us :P ) | 22:39 |
serialhex | btw, it throws the same error with & w/o the () around features[i].attr_obj, so i dont think that is the problem | 22:44 |
@sonney2k | serialhex, ok I understand the problem now but don't know how to fix it *sigh*. for now comment that part out and you can at least compile. | 22:59 |
serialhex | ok, will do! | 22:59 |
serialhex | make-ing now, lets see if there are any other errors :-/ | 23:01 |
serialhex | so what is the problem sonney2k?? | 23:01 |
@sonney2k | it is some pointer in dynarray that is read-only | 23:15 |
@sonney2k | no idea why gcc 4.4.5 here does not complain | 23:15 |
blackburn | sonney2k: have you any idea how to return probabilities (real!) while multiclass classifying? | 23:22 |
serialhex | hmm... well let me know what i can do to help, it seems to compile cleanly so far, so we'll see | 23:23 |
serialhex | sploving: while i'm compiling i get this error - Unable to open file shogun/ShogunException.java: No such file or directory | 23:56 |
serialhex | make[1]: *** [Distance_wrap.cxx] Error 1 | 23:56 |
serialhex | just fyi... | 23:56 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 23:58 | |
--- Log closed Thu May 19 00:00:11 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!