--- Log opened Sat Mar 24 00:00:19 2012 | ||
-!- PhilTillet [~Philippe@tillet-p42154.maisel.int-evry.fr] has joined #shogun | 00:22 | |
-!- Netsplit *.net <-> *.split quits: n4nd0 | 00:55 | |
-!- PhilTillet [~Philippe@tillet-p42154.maisel.int-evry.fr] has quit [Quit: Leaving] | 00:57 | |
-!- PhilTillet [~Philippe@tillet-p42154.maisel.int-evry.fr] has joined #shogun | 00:57 | |
-!- PhilTillet [~Philippe@tillet-p42154.maisel.int-evry.fr] has quit [Quit: Leaving] | 01:42 | |
-!- gsomix [~gsomix@188.168.5.232] has joined #shogun | 04:46 | |
-!- vikram360 [~vikram360@117.192.163.224] has quit [Ping timeout: 260 seconds] | 04:55 | |
-!- gsomix [~gsomix@188.168.5.232] has quit [Ping timeout: 244 seconds] | 04:58 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 05:22 | |
harshit_ | Thanks sonney2k ,figured that out.. i forgot to call train() function .. | 05:41 |
---|---|---|
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 260 seconds] | 06:07 | |
-!- vikram360 [~vikram360@117.192.176.242] has joined #shogun | 06:24 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 06:39 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Quit: Leaving] | 07:17 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 07:19 | |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 246 seconds] | 07:32 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 08:57 | |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 252 seconds] | 09:21 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 09:25 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 09:54 | |
-!- gsomix [~gsomix@188.168.13.208] has joined #shogun | 10:31 | |
gsomix | hi | 10:31 |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 248 seconds] | 10:32 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 10:42 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 11:21 | |
harshit_ | hello people... | 11:22 |
blackburn | hi | 11:22 |
harshit_ | hey blackburn do you knw any function to find inverse of matrix in C++ | 11:23 |
blackburn | we have one in CMath::inverse | 11:23 |
-!- pythonroar [d2198538@gateway/web/freenode/ip.210.25.133.56] has joined #shogun | 11:24 | |
harshit_ | yeah got it, there is one pinv | 11:24 |
harshit_ | i think that would work | 11:24 |
harshit_ | also one more thing | 11:24 |
blackburn | pinv is for pseudoinverse | 11:24 |
blackburn | if you have a square non-singular matrix you may simply use inverse | 11:24 |
harshit_ | actually i want to convert A\B in matlab to C++ | 11:25 |
blackburn | what is B? | 11:25 |
harshit_ | so i should use pinv no ? | 11:25 |
blackburn | a vector? | 11:25 |
harshit_ | A is matrix and B is vector | 11:25 |
blackburn | IIRC A\B is a solution of Ax=b? | 11:26 |
harshit_ | yes | 11:26 |
harshit_ | also in course of converting code from matlab to c++ i have made a lot of new math function, | 11:26 |
harshit_ | like creating a diagonal matrix | 11:27 |
blackburn | well inverse is not the best way to solve Ax=b | 11:27 |
blackburn | worst probably :D | 11:27 |
harshit_ | so what should i do | 11:27 |
blackburn | use lapack dgesv | 11:27 |
harshit_ | find a function in lapack | 11:28 |
harshit_ | ? | 11:28 |
harshit_ | okay | 11:28 |
harshit_ | great, even better | 11:28 |
blackburn | inverse is pretty hard to compute | 11:28 |
harshit_ | yeah .. got your point | 11:28 |
harshit_ | is there a wrapper for dgesv in shogun ? | 11:29 |
blackburn | no, it is not needed | 11:31 |
blackburn | harshit_: is you matrix symmetric? | 11:31 |
harshit_ | no . | 11:32 |
harshit_ | its a matrix created from combination of feature matrix and weight matrix | 11:32 |
harshit_ | so it cannt be symmetric | 11:33 |
blackburn | you need to use wrap_dgetri to wrap_dgetrs | 11:33 |
blackburn | hmm | 11:33 |
blackburn | inverse is computed with dgetri anyway | 11:34 |
blackburn | so probably it is the same for general matrix | 11:34 |
blackburn | harshit_: in this case it is ok to use inverse and matrix-vector multiply then | 11:34 |
harshit_ | i think lapack's dgesv would be better | 11:34 |
blackburn | dgesv does the same | 11:35 |
harshit_ | thanks blackburn for help | 11:35 |
harshit_ | Also my code is about to get over . so could you please guide me though the process of pushing on github | 11:36 |
blackburn | create fork, commit your changes into it and create pull request using github | 11:37 |
harshit_ | so i can commit only cpp and .h file , or i can make changes in some existing files like math.h | 11:38 |
blackburn | in your fork you may change what you need | 11:39 |
harshit_ | okay got it .. thanks again | 11:39 |
blackburn | when you do a pull request you ask admins to merge changes into main repo | 11:39 |
blackburn | if something was done wrong you just would be asked to fix it | 11:40 |
harshit_ | got it.. | 11:40 |
harshit_ | one more thing in gsoc, if we want to submit more than one project to shogun for gsoc,Does all of them need to be in the same application ? | 11:41 |
blackburn | yeah probably | 11:43 |
blackburn | if you want to apply for a few ideas you should write a proposal for each of them | 11:44 |
blackburn | it doesn't mean they should be completely different but anyway should be separated | 11:44 |
blackburn | ok have to go now | 11:45 |
harshit_ | got it | 11:45 |
harshit_ | thanks blackburn | 11:45 |
harshit_ | bye | 11:45 |
blackburn | yw | 11:45 |
-!- blackburn [~qdrgsm@83.234.54.68] has quit [Quit: Leaving.] | 11:45 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 12:02 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit] | 12:07 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 12:10 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit] | 12:13 | |
-!- ekansh [~ekansh@122.162.48.102] has joined #shogun | 12:47 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 12:58 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Read error: Connection reset by peer] | 13:12 | |
-!- vikram360 [~vikram360@117.192.176.242] has quit [Ping timeout: 244 seconds] | 13:48 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Remote host closed the connection] | 13:48 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 13:49 | |
gsomix | sonney2k, hey | 14:07 |
-!- vikram360 [~vikram360@117.192.176.242] has joined #shogun | 14:28 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 14:42 | |
-!- pythonroar [d2198538@gateway/web/freenode/ip.210.25.133.56] has quit [Ping timeout: 245 seconds] | 14:57 | |
-!- ekansh [~ekansh@122.162.48.102] has quit [Quit: Leaving.] | 15:06 | |
-!- ekansh [~ekansh@122.162.48.102] has joined #shogun | 15:09 | |
-!- pythonroar [7b75ac4e@gateway/web/freenode/ip.123.117.172.78] has joined #shogun | 15:27 | |
n4nd0 | I am getting some errors with valgrind and I don't understand at all why they are there | 15:51 |
n4nd0 | there is a large bunch of these guys | 15:51 |
n4nd0 | http://snipt.org/ugTg8 | 15:51 |
n4nd0 | any clue? | 15:51 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Read error: Connection reset by peer] | 16:02 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 16:13 | |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 264 seconds] | 16:20 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 16:20 | |
n4nd0 | sonney2k: hey! it would be great if you take a look to some errors I am getting with valgrind, I don't find any sense on them | 16:23 |
-!- ekansh [~ekansh@122.162.48.102] has quit [Quit: Leaving.] | 16:34 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Remote host closed the connection] | 16:45 | |
-!- vikram360 [~vikram360@117.192.176.242] has quit [Ping timeout: 252 seconds] | 16:59 | |
-!- sahils_ [ca4eafc7@gateway/web/freenode/ip.202.78.175.199] has joined #shogun | 17:11 | |
sahils_ | hello | 17:12 |
sahils_ | my name is Sahil and I am interested in participating in GSoC2012 as a part of shogun | 17:12 |
sahils_ | can anyone please help me with it? | 17:13 |
sahils_ | This is my first time on an IRC channel. | 17:13 |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 244 seconds] | 17:15 | |
n4nd0 | sahils_: hi! | 17:17 |
n4nd0 | welcome ;) | 17:17 |
n4nd0 | so are you interested in any project in particular? | 17:17 |
sahils_ | thanks a lot for replying | 17:17 |
sahils_ | was feeling a bit lost here | 17:17 |
sahils_ | :) | 17:17 |
sahils_ | yeah, I was interested in the third idea stated on shogun ideas page | 17:18 |
sahils_ | implementing new machine learning algorithms | 17:18 |
n4nd0 | sounds good | 17:18 |
n4nd0 | it would be probably useful that you contact the mentor for that project | 17:19 |
sahils_ | i have had previous experience in machine learning implementation, I have tried ID3, regression etc | 17:19 |
n4nd0 | cool | 17:19 |
sahils_ | yeah I shall do that | 17:19 |
sahils_ | but I wanted to know one thing | 17:19 |
sahils_ | does shogun support random forest algorithm | 17:19 |
sahils_ | i was very interested in that one and could not find it explicitly in the source | 17:20 |
n4nd0 | mmmm I don't know with certainty | 17:20 |
sahils_ | oh ok | 17:20 |
sahils_ | I shall contact the mentor then | 17:20 |
n4nd0 | yes | 17:20 |
n4nd0 | and I think you should checkout the project code and get used to it | 17:21 |
sahils_ | may I know if you are also planning to participate or are a mentor | 17:21 |
sahils_ | ? | 17:21 |
n4nd0 | you may have read that it's a requirement to do a small patch together with GSoC application | 17:21 |
n4nd0 | I'm student as well | 17:21 |
sahils_ | yeah, I have read that. I somehow couldn't get the source code working on my machine | 17:22 |
sahils_ | my university has proxy restrictions on downloads so could not get the dependencies | 17:22 |
sahils_ | trying a work around now | 17:22 |
n4nd0 | I think that is the first step you should do | 17:22 |
n4nd0 | ok | 17:22 |
sahils_ | yeah | 17:23 |
n4nd0 | sahils_: I've to go now | 17:23 |
sahils_ | no problem, thank you so much | 17:23 |
-!- serialhex [~serialhex@70-88-36-65-miami-fl.hfc.comcastbusiness.net] has joined #shogun | 17:23 | |
n4nd0 | sahils_: but come around and ask if you need some help with something or use the mailing list, as you prefer | 17:23 |
sahils_ | i shall do that | 17:23 |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 17:25 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 246 seconds] | 17:28 | |
-!- sahils_ [ca4eafc7@gateway/web/freenode/ip.202.78.175.199] has quit [Ping timeout: 245 seconds] | 17:40 | |
-!- pythonroar [7b75ac4e@gateway/web/freenode/ip.123.117.172.78] has quit [Quit: Page closed] | 17:44 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 18:29 | |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 272 seconds] | 18:50 | |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 18:50 | |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 245 seconds] | 19:11 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 252 seconds] | 19:15 | |
-!- karlnapf [~heiko@host86-176-2-96.range86-176.btcentralplus.com] has joined #shogun | 19:26 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 19:30 | |
karlnapf | sonney2k, around? | 19:38 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 20:14 | |
CIA-64 | shogun: Heiko Strathmann master * r902607c / src/NEWS : | 20:15 |
CIA-64 | shogun: added some stuff to news | 20:15 |
CIA-64 | shogun: increased param version - http://git.io/BZ5paw | 20:15 |
CIA-64 | shogun: Soeren Sonnenburg master * rf399095 / (3 files in 3 dirs): | 20:15 |
CIA-64 | shogun: Merge pull request #389 from karlnapf/master | 20:15 |
CIA-64 | shogun: some news updates - http://git.io/0dTtGw | 20:15 |
CIA-64 | shogun: Heiko Strathmann master * r39197dd / src/shogun/base/SGObject.cpp : removed a debug print - http://git.io/LmT24w | 20:15 |
CIA-64 | shogun: Heiko Strathmann master * re5ce2ef / src/shogun/machine/Machine.cpp : added parameter map entry to machine class - http://git.io/yzMI8g | 20:15 |
-!- serialhex [~serialhex@70-88-36-65-miami-fl.hfc.comcastbusiness.net] has quit [Quit: leaving] | 20:15 | |
karlnapf | sonney2k, now there? | 20:16 |
karlnapf | trying to fix these de-serialization things, but got some questions on how the testsuite works | 20:16 |
@sonney2k | karlnapf, hi | 20:16 |
karlnapf | hi | 20:16 |
@sonney2k | karlnapf, the testsuite (aka tester.py and generator.py) do nothing more than run the examples | 20:17 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 20:17 | |
blackburn | hey what's up | 20:17 |
karlnapf | where are the serialised files? | 20:17 |
karlnapf | because the txt files in the tests folder contain pickle serializations | 20:17 |
blackburn | data/testsuite | 20:17 |
@sonney2k | karlnapf, that means that generator.py runs the examples with every setting listed in paramer_lists and pickles the results | 20:18 |
karlnapf | but the problem is when loading serialized files right? | 20:18 |
karlnapf | I want to load such a file by hand and see what goes wrong | 20:19 |
blackburn | karlnapf: what is the problem? | 20:19 |
blackburn | ah | 20:19 |
@sonney2k | karlnapf, then tester.py runs the same example again and loads the serialized file and compares | 20:19 |
@sonney2k | that's it | 20:19 |
karlnapf | yes | 20:19 |
karlnapf | ok, but everythign works with pickle serialization files? | 20:20 |
@sonney2k | karlnapf, the problem can be easily triggered when you try to load e.g. classifier_libsvm_modular0.txt | 20:20 |
karlnapf | from python pickle? | 20:20 |
karlnapf | not possible from c++? | 20:20 |
@sonney2k | just do x=pickle.load(file('classifier_libsvm_modular0.txt')) | 20:21 |
karlnapf | ok will try that | 20:21 |
@sonney2k | and then pickle.dumps(x) | 20:21 |
@sonney2k | you will see that kernel etc is NULL | 20:21 |
karlnapf | is there some way to extract the shogun serialized files from these txt files? | 20:21 |
@sonney2k | well you could do the same from C++ - but you would need an old file | 20:22 |
karlnapf | because form python, working on this is a bit messy | 20:22 |
karlnapf | that why I want to extract such a file from the txt file | 20:22 |
@sonney2k | karlnapf, well then use shogun 1.1.0 to serialize the minimal libsvm example | 20:23 |
@sonney2k | and then load it with shogun 1.2.0 | 20:23 |
karlnapf | mmmh, ok | 20:24 |
karlnapf | thanks for the tips | 20:24 |
karlnapf | will try to fix it now | 20:24 |
blackburn | guys, what are the problems? | 20:24 |
blackburn | is parameter mapping failing? | 20:24 |
@sonney2k | blackburn, one cannot load old serialized files | 20:24 |
karlnapf | blackburn, some files are not properly loaded, subsvms of multiclass svm are NULL | 20:25 |
@sonney2k | as soon as a new parameter is there it fails | 20:25 |
blackburn | sonney2k: oh tht's bad :) | 20:25 |
@sonney2k | but it works ok if one comments the code block showing these WARNINGS | 20:25 |
@sonney2k | about new params | 20:25 |
blackburn | sonney2k: I was always curious, do people really store a lot of shogun's files? | 20:25 |
@sonney2k | no idea | 20:26 |
blackburn | I mean we care on that too much (just my opinion) | 20:26 |
blackburn | pretty late to stop but anyway | 20:26 |
@sonney2k | then we can aswell remove all means to serialize shogun objects | 20:27 |
karlnapf | sonney2k, do you have some names of classes that fail to load? then I will use them for testing first | 20:27 |
@sonney2k | libsvm | 20:27 |
@sonney2k | but as I said any class that has a new parameter | 20:27 |
karlnapf | kk | 20:27 |
blackburn | sonney2k: first spare evening for last N months? | 20:28 |
blackburn | ;) | 20:28 |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 20:32 | |
@sonney2k | n4nd0, what are the errors? | 20:32 |
blackburn | sonney2k http://snipt.org/ugTg8 these one I guess | 20:34 |
@sonney2k | n4nd0, well no wonder - you serialize a SGVector / Matrix but cast it into a SGObject | 20:37 |
blackburn | sonney2k: don't you think some old branches should be removed? | 20:37 |
blackburn | sonney2k: ah yes and some application too :D | 20:37 |
@sonney2k | n4nd0, so fix is just to do m_parameters->add(&vec, "vectors", "The vectors.") | 20:38 |
@sonney2k | blackburn, which? | 20:39 |
blackburn | sonney2k: sailfish | 20:40 |
blackburn | all commented | 20:40 |
blackburn | no idea who will ever fix it | 20:41 |
blackburn | sonney2k: I'm tending to add my road sign recognition application | 20:43 |
blackburn | cause it seems I'm not going to participate in any commercial development of it | 20:44 |
@sonney2k | would be great | 20:52 |
blackburn | the only problem is hog computing thing | 20:52 |
blackburn | I borrowed some code and I have to implement it by myself probably | 20:52 |
blackburn | sonney2k: will my paper be considered by same reviewers? | 20:53 |
blackburn | sonney2k: I've been thinking about more generic thread parameter class | 20:59 |
blackburn | and it seems pretty difficult to implement a map <string,anything> | 21:00 |
blackburn | without boost | 21:00 |
blackburn | but we may copy boost::any thoguh | 21:00 |
-!- harshit_ [~harshit@182.68.130.255] has quit [Ping timeout: 272 seconds] | 21:06 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 21:13 | |
n4nd0 | sonney2k: ok, I will change it to that then, but it is a pointer to a vector, not a vector exactly, I don't know if that makes a difference | 21:13 |
n4nd0 | sonney2k: I found in another file that it was done like that, let me look for it | 21:13 |
@sonney2k | n4nd0, yes that changes everything | 21:15 |
-!- karlnapf1 [~heiko@host86-185-9-214.range86-185.btcentralplus.com] has joined #shogun | 21:15 | |
blackburn | sonney2k: why are you not a fan of unsigned stuff? | 21:16 |
-!- karlnapf [~heiko@host86-176-2-96.range86-176.btcentralplus.com] has quit [Ping timeout: 276 seconds] | 21:16 | |
blackburn | you removed it in liblinear and libocas | 21:16 |
-!- harshit_ [~harshit@182.68.130.255] has joined #shogun | 21:17 | |
@sonney2k | n4nd0, I don't even have an idea how to serialize lists of SGVector etc | 21:18 |
n4nd0 | sonney2k: take a look if you can a moment, these are the types of the members and how I am doing it | 21:18 |
n4nd0 | http://snipt.org/ugUf4 | 21:18 |
n4nd0 | aham | 21:18 |
@sonney2k | n4nd0, there is DynamicArray / DynArray but these cannot serialize the things beneath | 21:19 |
n4nd0 | sonney2k: mmmm I see, so what is the best solution to take then? | 21:19 |
@sonney2k | n4nd0, yeah that cannot work | 21:20 |
@sonney2k | n4nd0, well what you could do is use SGMatrix instead of SGVector (with number of columns being number of classes) | 21:20 |
@sonney2k | and then we need a SGTensor for SGMatrix* | 21:21 |
@sonney2k | n4nd0, actually we have that as SGNDArray | 21:21 |
n4nd0 | sonney2k: tensors :O I got scared of those in a plasma physics course | 21:21 |
@sonney2k | n4nd0, well call it 3 dimensional array then :) | 21:22 |
blackburn | n4nd0: tensors can be viewed just like N dimensional arrays ;) | 21:22 |
n4nd0 | blackburn: yeah, it was nice until that part ... but it got dark later on :) | 21:22 |
@sonney2k | blackburn, unsigned vs signed - it is just dangerous to do any math on unsigned | 21:23 |
n4nd0 | blackburn: I can have a nice reconciliation in any case :P | 21:23 |
@sonney2k | i.e. u-1 can be greater than you | 21:23 |
@sonney2k | so I generally prefer to not use it (if there is no very good reason) | 21:24 |
@sonney2k | s/you/u/ | 21:24 |
blackburn | are you sure there is no performance reasons to use unsigned? | 21:24 |
@sonney2k | yes 100% | 21:24 |
blackburn | sonney2k: you was right - 'you' works too | 21:24 |
blackburn | I mean I am definitely less than unsigned -1 ;) | 21:26 |
@sonney2k | n4nd0, so is it clear how to resolve the problem? | 21:30 |
@sonney2k | karlnapf1, could your reproduce the problem? | 21:30 |
karlnapf1 | sonney2k, yes | 21:30 |
karlnapf1 | there are memory errors | 21:30 |
n4nd0 | sonney2k: not really :( I am looking for SGTensor but I don't manage to find it actually | 21:30 |
n4nd0 | sonney2k: ah ok you say we need, not that we have it | 21:31 |
karlnapf1 | just found the bad line ... I think this may be because new parameters are initialized with NULL, and when migrating I am reading from these | 21:31 |
@sonney2k | n4nd0, use SGNDArray | 21:32 |
n4nd0 | sonney2k: got it, SGMatrix for SGVector* and SGNDArray for SGMatrix* right? | 21:32 |
@sonney2k | n4nd0, it is a bare container only though | 21:33 |
@sonney2k | n4nd0, yes | 21:33 |
blackburn | too much serialization | 21:33 |
@sonney2k | blackburn, did you know that java has no unsigned data types? | 21:35 |
blackburn | sonney2k: yes but java is all about some other shit | 21:36 |
blackburn | they terribly don't care about performance | 21:36 |
n4nd0 | :D | 21:36 |
@sonney2k | well they care - but it is certainly harder with garbage collectors and no control of the exact size of the underlying data type and jit etc | 21:38 |
@sonney2k | blackburn, I don't know about you - but I want to be able to just say 'dump shogun object' and have it on harddisk | 21:39 |
blackburn | sonney2k: what do you mean? | 21:39 |
blackburn | I don't think java provides such easy way :) | 21:39 |
blackburn | rather some complex framework based on 300 library dependencies | 21:40 |
-!- harshit_ [~harshit@182.68.130.255] has quit [Remote host closed the connection] | 21:43 | |
@sonney2k | blackburn, about your paper - it is likely that same reviewers get it but you might also have new ones | 21:44 |
blackburn | I see | 21:44 |
blackburn | thanks | 21:44 |
n4nd0 | sonney2k, blackburn: I don't see any method Parameter::add for SGNDArray, any workaround to do it? | 21:47 |
blackburn | n4nd0: implement it :D | 21:47 |
n4nd0 | blackburn: haha looks legit :P I just found it curious that there had not been necessity of it before and wondered if there was sth alternative | 21:47 |
blackburn | n4nd0: it is pretty rarely when some method uses Nd array :D | 21:48 |
@sonney2k | n4nd0, well don't implement it just yet - just use ndarray in your qda class and we later have to think of how to implement that stuff | 21:50 |
n4nd0 | blackburn: but there is some stuff using SGVector* and SGMatrix*, I think that ConjugateIndex for example uses sth similar | 21:50 |
blackburn | n4nd0: oh what a wonderful example of useless shit - Conjugate Index | 21:50 |
blackburn | :D | 21:50 |
n4nd0 | haha | 21:51 |
n4nd0 | blackburn: are you angry with the world today man? :P | 21:51 |
blackburn | n4nd0: not really | 21:51 |
blackburn | n4nd0: http://www.waset.org/journals/waset/v13/v13-13.pdf | 21:51 |
blackburn | you should try to understand at least one sentence | 21:51 |
blackburn | it looks like they were translating it with some old version of google translate or so | 21:52 |
n4nd0 | :) | 21:52 |
n4nd0 | that pisses me off so much man ... when I see a paper with really bad language it is like wtf!? | 21:52 |
n4nd0 | but I don't really get it, isn't that controlled somehow? | 21:53 |
blackburn | n4nd0: probably it is a non-free journal | 21:53 |
blackburn | so they just publish some shit | 21:53 |
blackburn | just to publish | 21:53 |
n4nd0 | blackburn: aham! so there are also journals where paying a fee one gets it published? | 21:54 |
n4nd0 | more or less, just the general idea | 21:54 |
blackburn | probably they even pay for getting published I do not really know | 21:55 |
n4nd0 | blackburn: ok, are you working on any publication? I think I read sth about that in the conversation | 21:56 |
blackburn | n4nd0: yes, trying to get JMLR paper accepted | 21:56 |
blackburn | about dim reduction stuff | 21:56 |
-!- mudit3774 [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun | 21:57 | |
n4nd0 | blackburn: that's really cool man, I wish you good luck with it! | 21:57 |
blackburn | I already got rejected with first version | 21:57 |
mudit3774 | Hi | 21:57 |
n4nd0 | blackburn: :( | 21:58 |
n4nd0 | mudit3774: hey! | 21:58 |
mudit3774 | I am interested in applying for GSoC- 2012 for shogun | 21:58 |
blackburn | n4nd0: you are the most friendly student ever | 21:59 |
blackburn | :D | 21:59 |
n4nd0 | mudit3774: good! any project you are interested in, in particular? | 21:59 |
mudit3774 | I have gone through the ideas page and I am interested in Built generic multiclass learning framework | 21:59 |
mudit3774 | I am sorry.. for being so formal :) ..but the whole idea and two lenghy publications made me write my intro like this | 22:00 |
mudit3774 | :) | 22:00 |
n4nd0 | mudit3774: good! I think the mentor in charge of it is around here | 22:00 |
n4nd0 | :) | 22:00 |
n4nd0 | mudit3774: anyhow, have you already worked with shogun? | 22:01 |
mudit3774 | Well no.. :( but I have some machine learning experience.. | 22:01 |
mudit3774 | Last year I was selected for University of Michigan (CSCS) | 22:02 |
mudit3774 | and worked with complex adaptive systems | 22:02 |
n4nd0 | mudit3774: that's good | 22:02 |
n4nd0 | mudit3774: so I guess you may have come across it, it looks like this year it's important to have a small patch already done for the project in order to apply for GSoC | 22:03 |
mudit3774 | Oh.. dats cool.. but do we have to give a link to the patch in our proposal | 22:03 |
mudit3774 | ? | 22:03 |
gsomix | sonney2k, there is good news. | 22:03 |
mudit3774 | or we can do in during and after application period | 22:04 |
gsomix | *are | 22:04 |
mudit3774 | I mean till 23rd MAr? | 22:04 |
n4nd0 | mudit3774: I think there's no an strict deadlie for it | 22:04 |
n4nd0 | mudit3774: but it would be good to have it done when you submit your proposal | 22:04 |
n4nd0 | mudit3774: there is a section to write what experience do you have with shogun | 22:05 |
mudit3774 | Well that would be a bit tough, I am still on the documentation.. | 22:05 |
@sonney2k | gsomix, ? | 22:06 |
@sonney2k | gsomix, python3 works? | 22:06 |
n4nd0 | mudit3774: I see ... | 22:06 |
mudit3774 | but doen familiarity with the algorithm won't help? and is there an application template? | 22:06 |
n4nd0 | mudit3774: in any case I would suggest you to fork the project, checkout the code locally and play around with it | 22:07 |
@sonney2k | yes that is what you should do - and contribute (useful!) patches | 22:07 |
@sonney2k | that is a big big plus | 22:07 |
@sonney2k | basically everyone we accepted last year was contributing quite a bit before | 22:08 |
mudit3774 | yeah.. that's what I was thinking, but then I thought of introducing myself first, anyways, will get back if I could contribute a useful patch, this mean having only algorithmic knowledge won't help much literature survey is not that important? | 22:09 |
n4nd0 | mudit3774: http://www.google-melange.com/gsoc/org/google/gsoc2012/shogun scroll quite a bit down to find the application template | 22:09 |
gsomix | sonney2k, yep. Now it can be configured and compiled. Now this requires testing. | 22:10 |
gsomix | Too many 'now'. :( | 22:11 |
mudit3774 | thanks.. | 22:11 |
blackburn | mudit3774: what is your multiclass experience? | 22:11 |
mudit3774 | Well,. I have not gone through the publications yet, but I worked on a project which used | 22:13 |
mudit3774 | Ant Colony Optimization for controlling tree evolution in a RF agorithms | 22:14 |
mudit3774 | I am guessing it is on similar lines | 22:14 |
blackburn | what is RF? | 22:14 |
mudit3774 | Random Forest | 22:14 |
blackburn | ah I see | 22:14 |
mudit3774 | It was used for ranking data according to variable importance | 22:15 |
blackburn | that idea is not really about trees nor forests.. | 22:16 |
blackburn | but anyway any experience is valuable :) | 22:16 |
mudit3774 | well I guess I have to well prepared before my next IRC visit then :) | 22:16 |
blackburn | I meant trees like random forest, C5.0 (different idea) or anything similar | 22:16 |
blackburn | what do you mean? | 22:17 |
mudit3774 | I mean a through literature survey on the topic at least | 22:17 |
blackburn | I think nobody have to prepare visit IRC :) | 22:17 |
mudit3774 | cool.. :) anyways thanks | 22:18 |
blackburn | I'm currently trying to come up with some idea | 22:18 |
blackburn | how can you contribute | 22:18 |
mudit3774 | Well I have experience with Neural Networks, Bio inspired evolutionary algorithms and Random Forest.. | 22:19 |
blackburn | random forest is pretty big project I guess | 22:19 |
mudit3774 | In past GSoC and some projects under faculty members.. | 22:19 |
blackburn | what kind of NNs are you able to implement easy? | 22:20 |
mudit3774 | Well Back propogating.. | 22:20 |
mudit3774 | and some easier versions | 22:20 |
mudit3774 | like | 22:20 |
mudit3774 | hoppfield | 22:20 |
mudit3774 | nns | 22:20 |
mudit3774 | and yeah percetron based | 22:21 |
blackburn | we have averagedperceptron only here | 22:21 |
mudit3774 | oh, well back propogation is quite popular and widely used | 22:21 |
blackburn | feel free to implement any then | 22:22 |
mudit3774 | cool.. so I will try to work on that.. | 22:22 |
@sonney2k | gsomix, great work! | 22:23 |
gsomix | sonney2k, tomorrow(today? :)) I will be correcting our hacks and sending to the github. | 22:27 |
shogun-buildbot | build #381 of ruby_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/381 blamelist: heiko.strathmann@gmail.com | 22:32 |
blackburn | what again | 22:32 |
blackburn | timeout? | 22:32 |
blackburn | :D | 22:33 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 22:56 | |
blackburn | wiking: re your message - just send an email or so I am happy to discuss anything related ;) | 22:56 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 23:03 | |
gsomix | sonney2k, posted. :) | 23:11 |
-!- mudit3774 [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Quit: Page closed] | 23:16 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 23:34 | |
--- Log closed Sun Mar 25 00:00:19 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!