--- Log opened Tue Mar 29 00:00:36 2011 | ||
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be back...] | 00:27 | |
-!- blackburn [~qdrgsm@188.168.5.159] has joined #shogun | 00:52 | |
-!- blackburn [~qdrgsm@188.168.5.159] has quit [Quit: Leaving.] | 01:02 | |
-!- seviyor [c1e20418@gateway/web/freenode/ip.193.226.4.24] has quit [Ping timeout: 252 seconds] | 01:21 | |
-!- aifargonos [~aifargono@46.18.27.35] has left #shogun [] | 02:04 | |
-!- serialhex [~serialhex@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 02:38 | |
-!- serialhex [~serialhex@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 02:53 | |
-!- shelhamer [~shelhamer@AMontsouris-152-1-64-127.w83-202.abo.wanadoo.fr] has joined #shogun | 02:55 | |
-!- sploving [~root@124.16.139.196] has left #shogun [] | 02:57 | |
-!- sonney2k [~sonney2k@f053046246.adsl.alicedsl.de] has quit [Ping timeout: 276 seconds] | 03:04 | |
-!- sonney2k [~sonney2k@f053045182.adsl.alicedsl.de] has joined #shogun | 03:17 | |
-!- serialhex [~serialhex@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 03:29 | |
-!- ^knrrrd^ [~knrrrd@bane.ml.tu-berlin.de] has joined #shogun | 05:00 | |
-!- Netsplit *.net <-> *.split quits: cwidmer, yoh, yin, l0uis_, siddharth, sonney2k, meanerelk, @knrrrd, shelhamer | 05:07 | |
-!- Netsplit over, joins: siddharth, meanerelk, l0uis_, sonney2k | 05:20 | |
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has joined #shogun | 05:20 | |
-!- Netsplit over, joins: yoh, shelhamer | 05:20 | |
-!- Netsplit over, joins: cwidmer, yin | 05:20 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 07:16 | |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 07:30 | |
-!- sploving [~root@124.16.139.196] has left #shogun [] | 07:32 | |
-!- eppz [~epps@84.18.157.200] has joined #shogun | 07:39 | |
-!- eppz [~epps@84.18.157.200] has quit [Changing host] | 07:39 | |
-!- eppz [~epps@unaffiliated/epps] has joined #shogun | 07:39 | |
-!- eppz [~epps@unaffiliated/epps] has quit [Client Quit] | 07:44 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 07:47 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 07:48 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Client Quit] | 07:49 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 09:19 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 09:29 | |
sploving | hello sonney2k | 09:53 |
---|---|---|
sonney2k | sploving, good morning | 09:54 |
sploving | good morning. I find that jblas do not support sparse matrix | 09:55 |
sonney2k | sploving, not many packages do - not even R has support for sparse matrices | 09:56 |
sploving | okay. then we should not write these typemaps | 09:57 |
sonney2k | sploving, let me check - I recall that there is one big java matrix package around | 09:57 |
-!- aifargonos [~aifargono@46.18.27.35] has joined #shogun | 09:57 | |
sploving | http://jblas.org/javadoc/index.html | 09:58 |
sonney2k | that one probably http://www.ujmp.org/ | 09:59 |
sonney2k | it can indeed do sparse matrices http://www.ujmp.org/apidocs/ujmp-core/apidocs/ | 10:00 |
sploving | this lib is maturate? | 10:02 |
sploving | I have never heard that | 10:02 |
sploving | it indeed | 10:02 |
sonney2k | sploving, I think it is very mature | 10:02 |
sploving | ok. sonney2k, another question, I took at the typemap.i in the python_modular and know what thatis meaning. But I do not understand the two interface file in pyton dir. "pythoninterface.h" "pythoninterface.cpp" | 10:04 |
sploving | what are they for? | 10:05 |
sploving | are they equal the wrapper file that swig generate? | 10:05 |
sonney2k | sploving, sorry where are these files? | 10:05 |
sonney2k | $ ls python_modular/ | 10:06 |
sonney2k | sg_print_functions.cpp swig_typemaps.i | 10:06 |
sonney2k | is all I have there | 10:06 |
sploving | $ls python/ | 10:06 |
sonney2k | you will only need to modify swig_typemaps.i | 10:06 |
sonney2k | sploving, ahh I understand. That is the legacy static interface | 10:07 |
sonney2k | let me explain | 10:07 |
sonney2k | shogun has two kinds of interfaces: static ones (do not require swig) and modular ones (require swig) | 10:07 |
sploving | this i know. | 10:07 |
sploving | what is the difference? | 10:08 |
sonney2k | the static ones are in the directories octave, r, python, matlab ... | 10:08 |
sonney2k | modular swig based ones in the *_modular | 10:08 |
sploving | the static interface files equal the wrapper file that swig generate? | 10:09 |
sonney2k | the difference is that for static ones one has to write interfaces / for swig they are automatically created *AND* maybe more important the static ones are not object oriented and have interfaces for only a very limit amount of functions | 10:09 |
sonney2k | Did I make myself clear? | 10:11 |
sploving | okay. swig could also generated static wrapper file | 10:11 |
sonney2k | sploving, well no... | 10:11 |
sonney2k | swig exports all of shoguns classes directly into python etc | 10:11 |
sonney2k | so we have e.g. CClassifier -> Classifier in python | 10:12 |
sonney2k | the static interfaces have just some 'sg' command that takes a few options e.g. sg('new_classifier', 'LibSVM') | 10:13 |
sonney2k | so totally different | 10:13 |
sploving | i took at the examples. and clear now. then I also need write the static java file? | 10:13 |
sonney2k | sploving, no | 10:13 |
sonney2k | only java_modular interface | 10:14 |
sploving | i am glad to hear that. but why not support static? | 10:15 |
sonney2k | sploving, because functionality is very limited in static anyways | 10:15 |
sploving | i know~ | 10:15 |
sonney2k | so I expect new users to only ever use a modular interface | 10:15 |
sploving | then for jave support, i need write typemap.i for ujmp. | 10:17 |
sonney2k | sploving, btw, I am just now modifying the configure script etc to enable swig for lua, ruby, java - such that you can get started much easier then | 10:17 |
sonney2k | swig_typemaps.i yes - I think you should start writing wrappers for jblas first and then maybe later for ujmp ... simply because mikio braun can mentor how this works (he is the author of jblas) and also a mentor for this | 10:18 |
sploving | sonney2k, i know. thx. | 10:20 |
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has left #shogun [] | 10:20 | |
sploving | as there are 12 weeks, what do you think better. java coding ->java test,doc; lua coding->lua test,doc; or java coding, lua coding->java test,doc, lua test doc | 10:22 |
sploving | i just need write the proposal plan. although they are not difficult to implement | 10:22 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 252 seconds] | 10:25 | |
sonney2k | sploving, I don't think it will take you that much time but I guess it is better to finish one before attempting the other. | 10:26 |
sonney2k | I guess you will easily be able to also do ruby_modular but I am fine with it not being mentioned in the proposal | 10:27 |
-!- aifargonos [~aifargono@46.18.27.35] has quit [Ping timeout: 240 seconds] | 10:28 | |
sploving | ruby never used:( | 10:29 |
sploving | overestimate me :) | 10:30 |
sonney2k | sploving, doesn't matter ... you know enough languages already to easily read that code | 10:30 |
sonney2k | the difficulty lies only in getting the typemaps to work, not the language - see http://www.ruby-lang.org/en/documentation/quickstart/ | 10:31 |
sploving | then are you familiar with one matrix lib? | 10:32 |
sploving | if support ruby which lib we use? | 10:33 |
sonney2k | sploving, http://www.ruby-doc.org/stdlib/ the matrix package I think. | 10:38 |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has joined #shogun | 10:38 | |
-!- aifargonos [~aifargono@193.206.186.107] has joined #shogun | 10:39 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 246 seconds] | 10:43 | |
-!- ^knrrrd^ is now known as knrrrd | 10:46 | |
-!- mode/#shogun [+o knrrrd] by ChanServ | 10:46 | |
sploving | sonney2k, for a language, the final file i need write is : swig_typemaps.i sg_print_functions.cpp, and testsuite files. the doc and the examples are not need. am i right? | 10:46 |
-!- mode/#shogun [+o sonney2k] by ChanServ | 10:46 | |
@sonney2k | well some examples at least | 10:50 |
-!- knrrrd [~knrrrd@bane.ml.tu-berlin.de] has left #shogun [] | 10:51 | |
@sonney2k | gtg, l8r | 10:53 |
-!- sonney2k [~sonney2k@f053045182.adsl.alicedsl.de] has quit [Quit: Leaving] | 10:53 | |
-!- yin [~chatzilla@207.46.92.18] has quit [Quit: ChatZilla 0.9.86.1 [Firefox 3.6.16/20110319135224]] | 10:55 | |
-!- jabbok_ [8d550074@gateway/web/freenode/ip.141.85.0.116] has joined #shogun | 10:55 | |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has quit [Ping timeout: 252 seconds] | 10:56 | |
-!- jabbok_ [8d550074@gateway/web/freenode/ip.141.85.0.116] has quit [Client Quit] | 10:56 | |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has joined #shogun | 10:56 | |
-!- yin [~chatzilla@207.46.92.18] has joined #shogun | 10:56 | |
-!- knrrrd [~knrrrd@bane.ml.tu-berlin.de] has joined #shogun | 11:20 | |
knrrrd | when did sonne leave? | 11:20 |
-!- mode/#shogun [+o knrrrd] by ChanServ | 11:21 | |
-!- shelhamer [~shelhamer@AMontsouris-152-1-64-127.w83-202.abo.wanadoo.fr] has quit [Quit: Computer has gone to sleep] | 11:42 | |
-!- shelhamer [~shelhamer@AMontsouris-152-1-64-127.w83-202.abo.wanadoo.fr] has joined #shogun | 12:06 | |
-!- sonney2k [~sonney2k@89.204.153.70] has joined #shogun | 12:10 | |
sonney2k | sploving, hmmhh it is actullay not totally straight forward to even get the basic swig thingies to compile | 12:17 |
sonney2k | sploving, I will try a few more hours but then I guess you have to take over... | 12:18 |
sonney2k | there are things happening like: java package names cannot have the same name like a class in the package | 12:18 |
sonney2k | etc | 12:18 |
sonney2k | sploving, it seems like only lua works out of the box | 12:20 |
sploving | sonney2k, you mean lua works well | 12:20 |
sonney2k | lua at least compiles | 12:20 |
sploving | could we use %rename java? | 12:21 |
sonney2k | sploving, yes but it is somehow stupid - I mean in all languages it is named differntly? stupid! | 12:21 |
sonney2k | but yes, for now I will rename Classifier -> ClassifierBase etc | 12:22 |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 12:23 | |
sploving | sonney2k, when you develop with python, r , not have these problems? | 12:26 |
sonney2k | sploving, well with R I had some problems too, but octave & python were easy, but give me a few more moments... | 12:27 |
sploving | sonney2k,okay | 12:28 |
-!- shelhamer [~shelhamer@AMontsouris-152-1-64-127.w83-202.abo.wanadoo.fr] has quit [Quit: Computer has gone to sleep] | 12:28 | |
sonney2k | sploving, regarding lua | 12:31 |
sonney2k | what how are binary extension named? | 12:31 |
sonney2k | I mean do the end with extension.so ? | 12:31 |
sonney2k | and how do I load an extension? | 12:32 |
sploving | yep | 12:32 |
sploving | require ("example") | 12:32 |
sploving | example.so is the binary | 12:32 |
sonney2k | sploving, segfault! | 12:33 |
sonney2k | dammed, this train is crowded with kids calling me a hacker! a hacker *LOL* | 12:33 |
sonney2k | ok so I guess something lua wise is not correct | 12:33 |
sonney2k | but at least it compiles ... | 12:34 |
sploving | sonney2k compiles is a good start. | 12:35 |
sonney2k | sploving, I will commit later today so you can have a look... | 12:35 |
sploving | sonney2k, i am writing the proposal now | 12:35 |
sonney2k | sploving, sure that has priority | 12:36 |
sploving | soney2k, okay | 12:41 |
-!- aifargonos [~aifargono@193.206.186.107] has quit [Ping timeout: 240 seconds] | 12:43 | |
-!- sonney2k [~sonney2k@89.204.153.70] has quit [Ping timeout: 246 seconds] | 12:44 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Read error: Connection reset by peer] | 12:45 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 12:50 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 12:58 | |
-!- sonney2k [~sonney2k@no.ml.tu-berlin.de] has joined #shogun | 13:02 | |
-!- seviyor [c1e20418@gateway/web/freenode/ip.193.226.4.24] has joined #shogun | 13:11 | |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has quit [Ping timeout: 252 seconds] | 13:18 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 14:24 | |
yin | on the second page of the coffin paper, the last formula I(\Phi(x)_k) should be \Phi(z)?? | 14:35 |
sonney2k | yin, ehh could be :) | 14:37 |
yin | :) as x should be the mapped vector and z is the original feature vector. | 14:37 |
sonney2k | true then | 14:39 |
sonney2k | yin, I think I have more papers to be proof red :D | 14:39 |
sonney2k | sploving, java compiles now too... | 14:41 |
sploving | sonney2k, congratulation~ thx for your work | 14:41 |
Tanmoy | @yin...feature map shld be on i/p vectors | 14:41 |
sonney2k | now off to ruby | 14:42 |
-!- aifargonos [~aifargono@46.18.27.35] has joined #shogun | 14:43 | |
sploving | sonney2k, good luck~ | 14:43 |
sploving | waiting for good news. | 14:43 |
sploving | sonney2k, the typemaps.i just map the in/out para, where are the *.i files that include all the functions that pytho,etc calls. for example, in pythin example, svm=LibSVM(..), where is the wrapper file for libsvm function | 14:49 |
sonney2k | sploving, in the modular/ dir we have all the *.i files that do the svm=LibSVM() magic etc | 14:51 |
yin | sonney2k, in section 2.1.2 computing \Phi for a variety of kernels. are the three kernels implemented in shogun now? | 14:52 |
sonney2k | the typemaps just do the conversion e.g. numpy matrix -> set_matrix(float64_t* matrix, int dim1, int dim2) | 14:52 |
sploving | that is the interface file for general language | 14:52 |
sonney2k | yin, spectrum / weightedegree, polynomial ? yes | 14:52 |
yin | sonney2k, yes. these three. | 14:53 |
sploving | sonney2k, i know ant take a look at it | 14:53 |
-!- skydiver [4deac315@gateway/web/freenode/ip.77.234.195.21] has joined #shogun | 15:09 | |
yin | sonney2k, i am reading CFeatures and its children now. These classes should contain all the data structures that hold data samples. Are there any other class outside this family that are relevant to data samples? | 15:40 |
sploving | sonney2k, there are mainly four arrays in python typemap: input output in-place argout. but the octave, r modular just have input argout array. | 15:40 |
sploving | Do r and octave not support output and in-place array? | 15:41 |
-!- siddharth [~siddharth@117.211.88.150] has quit [Remote host closed the connection] | 15:42 | |
sonney2k | sploving, within shogun I've never used inplace argout | 15:52 |
sonney2k | yin, no | 15:52 |
yin | sonney2k, thanks:) | 15:53 |
sonney2k | yin, these classes can be pretty big /messy | 15:53 |
sploving | sonney2k, besides the arrays, there are also CTstring feature, What are they for? | 15:53 |
yin | sonney2k, i am diving into the source now. | 15:53 |
yin | hope i can figure some structure out and i will add to my "some design patterns in shogun" report. | 15:54 |
sonney2k | sploving, regarding features we support strings (that is what these are for), sparse matrices and dense matrices | 15:54 |
yin | i think analyzing this part of code is important as we have a project for designing online feature class. | 15:55 |
yin | sonney2k, i also found that CAttributeFeatures and some others are never used by a classifier or clustering algorithm. | 15:58 |
sonney2k | yin, I hope it does not turn into a 'clean up the mess' project ... | 16:00 |
-!- aifargonos [~aifargono@46.18.27.35] has quit [Ping timeout: 250 seconds] | 16:01 | |
sonney2k | yin, yes they are not used. I was only drafting this - but had no time to finish... basically what I wanted is something like in C structs as features | 16:01 |
yin | sonney2k, i see. btw, there is a "feature name" concept for cdotfeatures? | 16:04 |
sonney2k | no | 16:05 |
yin | e.g. the first dimension of feature vector x means the "length", the 2nd dimension means "age" | 16:05 |
yin | ok. but if we add decision trees, this information is better encoded into its feature class. | 16:06 |
-!- aifargonos [~aifargono@193.206.186.107] has joined #shogun | 16:10 | |
sonney2k | yin, well not really | 16:10 |
sonney2k | it is nice to have I agree | 16:10 |
yin | i've sent a email to the mailing list earlier today about the opencv library | 16:11 |
yin | do you have any comments? | 16:11 |
yin | thanks. | 16:11 |
sonney2k | Yes, I will reply today | 16:11 |
yin | :) | 16:11 |
sonney2k | sploving, OK lua,java,ruby now at least compile with swig interfaces... | 16:31 |
sploving | sonney2k, good news~ | 16:31 |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has joined #shogun | 16:31 | |
sonney2k | sploving, I've pushed the changes to github already ... | 16:32 |
sonney2k | so if you have time try things out... | 16:32 |
sploving | okay. I will | 16:32 |
-!- sonney2k [~sonney2k@no.ml.tu-berlin.de] has quit [Ping timeout: 246 seconds] | 17:04 | |
-!- skydiver [4deac315@gateway/web/freenode/ip.77.234.195.21] has quit [Quit: Page closed] | 17:05 | |
-!- aifargonos [~aifargono@193.206.186.107] has quit [Ping timeout: 248 seconds] | 17:16 | |
sploving | sonney2k, thx. I submitted the proposal. when you have time please take a look at it. | 17:29 |
sploving | bye. good night~ | 17:29 |
-!- aifargonos [~aifargono@46.18.27.35] has joined #shogun | 17:43 | |
-!- fredt [~fredt@sshgate.tu-berlin.de] has joined #shogun | 17:47 | |
* fredt hrrrrr :) | 17:49 | |
fredt | thanks for the reports :) bye | 17:53 |
-!- fredt [~fredt@sshgate.tu-berlin.de] has quit [Quit: thats it :)] | 17:53 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 18:55 | |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Remote host closed the connection] | 19:10 | |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 19:11 | |
-!- epps [~epps@unaffiliated/epps] has quit [Ping timeout: 264 seconds] | 19:20 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 19:48 | |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has quit [Ping timeout: 252 seconds] | 20:13 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has joined #shogun | 20:35 | |
hiteshk | hi | 20:39 |
alesis-novik | hello | 20:40 |
hiteshk | hi , are you a mentor for this project | 20:40 |
hiteshk | or any project | 20:40 |
alesis-novik | No, I'm a student | 20:41 |
hiteshk | Ok , is Vojtech Franc available on irc | 20:42 |
alesis-novik | I don't think so. What idea are you interested in? | 20:43 |
hiteshk | I am interested in new machine learning algorithms , and you | 20:44 |
alesis-novik | Implementing EM | 20:44 |
hiteshk | So the mentor is Vijtech only , did you talk to him | 20:45 |
hiteshk | is any mentor available on the irc | 20:47 |
alesis-novik | I did send him an e-mail today | 20:53 |
alesis-novik | So what are you studying? | 20:56 |
hiteshk | I am a Computer Science undergraduate in India | 21:07 |
hiteshk | what about you | 21:09 |
alesis-novik | Artificial Intelligence masters student from UK | 21:11 |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 21:11 | |
@knrrrd | Hi, Guss | 21:14 |
@knrrrd | guys | 21:14 |
alesis-novik | Hello knrrrd | 21:15 |
hiteshk | alesis-novik , did he reply to your mail | 21:15 |
alesis-novik | I only sent it recently, so no. Hopefully tomorrow. | 21:16 |
@knrrrd | The official deadline is 8th, right? | 21:17 |
hiteshk | have you subscribed to the mailing list | 21:17 |
alesis-novik | yes knrrrd | 21:17 |
alesis-novik | Not yet hiteshk | 21:18 |
@knrrrd | Ah ok. I am looking forward to the applications. | 21:18 |
alesis-novik | If we submit them earlier, will anyone comment on them before the deadline? | 21:19 |
@knrrrd | alesis-novik: I am not sure if this would be fair | 21:20 |
alesis-novik | I meant asking to clarify something or similar comments, but either way is fine | 21:22 |
alesis-novik | Just so I know that I can refine it until the last day and then submit | 21:22 |
@knrrrd | As far as I know, the mentors will have a discussion on all applications after the deadline | 21:23 |
-!- siddharth_ [~siddharth@117.211.88.150] has joined #shogun | 21:45 | |
-!- sonney2k [~sonney2k@g225143240.adsl.alicedsl.de] has joined #shogun | 21:47 | |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be back] | 21:56 | |
sonney2k | *sigh* my router just died | 21:56 |
sonney2k | 4 hours without internet - hard to believe that I survived this nightmare ;-) | 21:57 |
aifargonos | ...I can imagine... | 21:58 |
aifargonos | torture for a programmer ... | 21:58 |
-!- davidcg [d8a57e6f@gateway/web/freenode/ip.216.165.126.111] has joined #shogun | 22:14 | |
davidcg | Not sure if anyone is there, but just wanted to point out that SVMLin is behaving strangely. It generates decent classifiers in very little time, but sometimes it reverses the labels (e.g. generates a classifier with 20% accuracy instead of 80% accuracy). | 22:18 |
sonney2k | davidcg, do you see any pattern in that? | 22:19 |
sonney2k | I mean like the first label is always -1 in such cases? | 22:19 |
davidcg | It happens about 50% of the time, I'd wager it's being assigned randomly. | 22:19 |
davidcg | Hm, not sure if it's dependent on what the first label is. | 22:19 |
sonney2k | davidcg, you could check if liblinear behaves in the same way | 22:21 |
sonney2k | if not then it could be really a bug | 22:22 |
davidcg | It may be due to the first label.... | 22:23 |
davidcg | Just checked and if the first label was -1, it would generate the opposite classifier, when I changed only the first label to 1 it produced the correct result. | 22:23 |
siddharth_ | hello | 22:24 |
sonney2k | davidcg, heh ... that code looks suspicous: | 22:24 |
sonney2k | float64_t sgn=train_labels[0]; | 22:24 |
sonney2k | for (int32_t i=0; i<num_feat+1; i++) | 22:24 |
sonney2k | Weights.vec[i]*=sgn; | 22:24 |
sonney2k | perfectly correlates with your observation | 22:25 |
sonney2k | davidcg, would you mind commenting lines 100-102 in SVMLin.cpp and see if it fixes the problem? | 22:25 |
sonney2k | siddharth_, hi... | 22:25 |
siddharth_ | i would like to work on "Implement Olivier Chapelles fast newton based SVM solver" | 22:27 |
davidcg | Haha, yeah, that looks like it would cause it. | 22:27 |
siddharth_ | so how should i apply through GSOC? | 22:27 |
davidcg | Btw, using the python modular interface - don't seem to be able to use LibLinear (get a "Sorry, not yet implemented." when attempting to train). | 22:28 |
siddharth_ | Btw sonney2k are u Soeren Sonnenburg? | 22:29 |
davidcg | Thanks for the help on SVMLin though - I'm sure that changing those lines will fix it. | 22:31 |
sonney2k | davidcg, so can you confirm it? | 22:31 |
sonney2k | siddharth_, I am about to send out an email with details | 22:31 |
siddharth_ | so should i apply after ur email? | 22:32 |
siddharth_ | There is one more thing | 22:34 |
siddharth_ | i was not able to clearly understand the member function "Calibrate" in SVMSGD | 22:34 |
sonney2k | siddharth_, that is Leon Bottou black magic :) | 22:36 |
davidcg | I'm sorry, don't have the permissions to recompile part of shogun on my server - I've used several other classifiers without seeing this behavior though, and that first line would definitely explain it. | 22:36 |
davidcg | (float64_t sgn=train_labels[0]) | 22:36 |
davidcg | Thanks again! | 22:37 |
-!- davidcg [d8a57e6f@gateway/web/freenode/ip.216.165.126.111] has quit [Quit: Page closed] | 22:37 | |
siddharth_ | I went through the slides of Leon Bottou "learning with large datasets" where he has explained SGD...but from the code its difficult to understand | 22:40 |
-!- davidcg [d8a57e6f@gateway/web/freenode/ip.216.165.126.111] has joined #shogun | 22:48 | |
davidcg | Just wanted to add - probably obvious, but SVMLin inverts the sign for b as well as w in the svm if the first training label is -1. | 22:50 |
-!- davidcg [d8a57e6f@gateway/web/freenode/ip.216.165.126.111] has quit [Client Quit] | 22:50 | |
-!- epps [~epps@unaffiliated/epps] has quit [Ping timeout: 250 seconds] | 22:51 | |
-!- shelhamer [~shelhamer@AMontsouris-152-1-50-9.w83-202.abo.wanadoo.fr] has joined #shogun | 22:53 | |
sonney2k | siddharth_, he simply has a lot of SGD experience | 22:55 |
sonney2k | I think he estimates step sizes for bias and w independently and also has some heuristics how often to update one but not the other | 22:56 |
siddharth_ | yes and he using the sparse features | 22:58 |
sonney2k | siddharth_, yes | 22:59 |
siddharth_ | sonney2k, for the primal SVM solver project,we have to convert matlab code to c++,in the overloaded train() function | 22:59 |
siddharth_ | before jumping to that,can u suggest some SVm code to begin with | 23:00 |
siddharth_ | so that I can get a feel of it | 23:00 |
sonney2k | the liblinear in shogun maybe? | 23:02 |
siddharth_ | sonney2k, ok i will try them...Also for the proposal do we have to give the timeline or just the project name? | 23:05 |
sonney2k | siddharth_, timeline is better. plan conservatively and maybe add additional projects in some later months/outlook... if you accidentally finish early :) | 23:06 |
siddharth_ | sonney2k, thanks...i will keep this in mind...Also i want to ask whether SVMSMO has been implemented in Shogun? | 23:09 |
sonney2k | siddharth_, what is SVMSMO? I mean that sounds like libsvm - no? | 23:10 |
sonney2k | siddharth_, potentially more interesting is a GPU based svm ... | 23:10 |
siddharth_ | sonney2k, Support Vector Machines Using Sequential Minimal Optimization | 23:10 |
sonney2k | siddharth_, but libsvm does that alread... | 23:10 |
sonney2k | http://mklab.iti.gr/project/GPU-LIBSVM | 23:10 |
sonney2k | that would be pretty cool I think ... | 23:11 |
siddharth_ | yeah :) | 23:14 |
siddharth_ | sonney2k, well I am currently going into CUDA thing for my 3D texture modeling project | 23:16 |
sonney2k | siddharth_, well then the plan is clear :D | 23:17 |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 23:17 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has left #shogun [] | 23:26 | |
-!- jabbok [8d550074@gateway/web/freenode/ip.141.85.0.116] has joined #shogun | 23:42 | |
--- Log closed Wed Mar 30 00:00:36 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!