--- Log opened Sun Apr 17 00:00:36 2011 | ||
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has joined #shogun | 00:09 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 00:16 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 00:25 | |
Tanmoy | @alesis | 00:26 |
---|---|---|
Tanmoy | the make is thrwoing some error | 00:26 |
Tanmoy | clapack_dpotri was not declared | 00:27 |
blackburn | just do git fetch upstream & git merge upstream/master | 00:27 |
blackburn | it is fixed already | 00:27 |
Tanmoy | oh okie | 00:31 |
Tanmoy | hmm says remote end hung up unexpectedly | 00:32 |
Tanmoy | ok done | 00:35 |
Tanmoy | thnx again blackburn | 00:36 |
blackburn | np | 00:36 |
Tanmoy | need to get some sleep | 00:37 |
Tanmoy | cya all | 00:37 |
blackburn | see you | 00:37 |
Tanmoy | ok bad stuff | 00:38 |
Tanmoy | the git fetch did not work | 00:38 |
Tanmoy | same error | 00:38 |
Tanmoy | @blackburn did u try to check it | 00:41 |
blackburn | yeap, it's working for me | 00:42 |
Tanmoy | hmmm what modifications were made ? | 00:42 |
blackburn | on the last commit, AveragedPerceptron stuff | 00:42 |
Tanmoy | k ill probably chekc it later | 00:44 |
alesis-novik | blackburn, I don't think it's properly fixed | 00:58 |
alesis-novik | I submitted a pull with a fix | 00:58 |
blackburn | alesis-novik: at least it's compiling now | 00:58 |
alesis-novik | blackburn, do you have ATLAS? | 00:59 |
blackburn | I have libatlas installed on my ubuntu, but I don't sure how sufficient it is | 01:00 |
alesis-novik | That might be it. If you have atlas, you would have compiled without any fixes | 01:01 |
alesis-novik | I forgot to put checks for LAPACK and ATLAS in the first version | 01:01 |
blackburn | aha | 01:02 |
Tanmoy | did u just submit it now | 01:03 |
alesis-novik | A few hours ago. sonney2k will probably check it tomorrow | 01:03 |
Tanmoy | oh okie | 01:04 |
alesis-novik | speaking of which, I'll probably go to bed insanely early by my standards and get up as early tomorrow. I feel I won't get any work done tonight... | 01:04 |
alesis-novik | Tanmoy, you can always install LAPACK :D | 01:05 |
alesis-novik | or rather ATLAS | 01:05 |
alesis-novik | anyway, goodnight folks | 01:05 |
Tanmoy | yup i did tht | 01:05 |
Tanmoy | yup goodnight | 01:05 |
Tanmoy | nned to get some sleep | 01:05 |
Tanmoy | and work on other thngs | 01:05 |
-!- blackburn [~qdrgsm@109.226.117.183] has quit [Quit: Leaving.] | 01:39 | |
-!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun | 01:39 | |
-!- blackburn [~qdrgsm@109.226.117.183] has quit [Client Quit] | 01:41 | |
CIA-110 | shogun: Sergey Lisitsyn master * rbe871a8 / src/libshogun/classifier/GaussianNaiveBayes.cpp : | 05:43 |
CIA-110 | shogun: Fixed destructor of GaussianNaiveBayes, fixed mistype in loop leading | 05:43 |
CIA-110 | shogun: to segfault in case of labels: k,k+1,...,n - http://bit.ly/hxqD6P | 05:43 |
CIA-110 | shogun: Soeren Sonnenburg master * r3cea610 / src/libshogun/classifier/GaussianNaiveBayes.cpp : | 05:43 |
CIA-110 | shogun: A number of coding style fixes like | 05:43 |
CIA-110 | shogun: - space between for and ( | 05:43 |
CIA-110 | shogun: - for loops with single statement don't need brackets | 05:43 |
CIA-110 | shogun: etc - http://bit.ly/f3U3pD | 05:43 |
serialhex | *YAY!!! it's SUNDAY!!!* | 06:03 |
@sonney2k | serialhex, for several hours already :) | 06:03 |
serialhex | for me it's only been sunday for 4 mins! | 06:04 |
serialhex | and only ~3,380,944.1533027566 km until 8am monday :D | 06:05 |
@bettyboo | yep | 06:06 |
serialhex | see!! bettyboo knows :P | 06:06 |
@bettyboo | serialhex: i was thinking of compiling it to a c++ dll and using that as the base. but i'd need to consult someone who knows c++ better | 06:06 |
serialhex | well dont ask me bettyboo, i dont know c++ that well | 06:06 |
@bettyboo | serialhex: well you can only be accepted for 1 for the summer... | 06:06 |
serialhex | sonney2k: did you just wake up? or have you been up?? (the former would be better than the latter, but then what are you doing waking up at 6am on a sunday??) | 06:08 |
@sonney2k | serialhex, our baby woke ab at 2:33 am ... I just managed to get her back to sleep | 06:08 |
@bettyboo | hi | 06:09 |
@sonney2k | so I though before going to bed I review a few pull requests | 06:09 |
serialhex | ahh, nice. i'm glad i dont have kids yet! | 06:09 |
serialhex | i noticed the push, and i updated myself while i was still on | 06:09 |
serialhex | well hopefully you can get some rest soon | 06:10 |
@sonney2k | serialhex, yeah will try now - 2 nights in a row with very little sleep are not good.... ahh and in 3 weeks we will have another child. So the situation can only improve ;-) | 06:12 |
serialhex | oh niiice!!! well g'nite sonney2k, and sleep well! | 06:13 |
serialhex | i hope your newborn is strong and healthy when s/he arrives :D | 06:13 |
@sonney2k | yeah cu | 06:13 |
* sonney2k is now in serialhex's timezone | 06:13 | |
@sonney2k | he indeed - but 3 weeks still | 06:14 |
@sonney2k | anyway off to bed | 06:14 |
@sonney2k | cu | 06:14 |
serialhex | nite | 06:14 |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 06:29 | |
siddharth | hi all | 06:29 |
* siddharth hmm no one is here | 06:31 | |
serialhex | nope... no one is here at all siddharth! | 06:40 |
siddharth | what you up to serialhex | 06:41 |
serialhex | reading some blogs, learning about ruby's narray and thinking about how to use it to speed up my mendelbrodt explorer program... | 06:41 |
serialhex | and i think i butchered his name... :( | 06:42 |
serialhex | what about yourself? | 06:42 |
siddharth | cool...working on integrating SGD-QN into shogun | 06:42 |
serialhex | pardon my ignorance, but what is SGD-QN?? | 06:44 |
serialhex | <3,629,764 km until 5pm monday in berlin!! | 06:45 |
siddharth | its an SVM classifier with with stochastic gradient | 06:46 |
serialhex | ooh, sounds interesting! | 06:48 |
* serialhex likes stochastic things, though chaos is better :P | 06:49 | |
siddharth | serialhex,chaos is a new TV series :P | 06:50 |
@bettyboo | funny!?! | 06:50 |
siddharth | hola bettyboo | 06:50 |
serialhex | really??? i didnt know | 06:50 |
@bettyboo | siddharth: How many applicants have you got so far? | 06:51 |
* serialhex dosnt watch too much tv | 06:51 | |
siddharth | same here ... have watched 24 though | 06:52 |
serialhex | ehh, i didnt like 24 too much, though i watch ncis every so often | 06:53 |
siddharth | there are many fans of Jack Bauer :P | 06:55 |
@bettyboo | rahaha? | 06:55 |
serialhex | i know, there are a BUNCH of people who loved that show and hated to see it go off the air. | 07:00 |
siddharth | yeah | 07:01 |
serialhex | i have ~730,280,756 km to travel before my next birthday :( | 07:15 |
serialhex | monday is getting closer :D 3,574,047 km!!! | 07:16 |
@bettyboo | ;D | 07:16 |
siddharth | you going to berlin? | 07:18 |
serialhex | no, just travelling though space :D | 07:24 |
serialhex | i made a spiffy spaceship earth calculator in ruby and i've been playing with it ever since | 07:24 |
serialhex | code is here: https://github.com/serialhex/fun-stuff/blob/master/spaceship_earth.rb | 07:25 |
siddharth | ohh i see | 07:30 |
Tanmoy | hi all | 07:30 |
serialhex | hey Tanmoy, how are you this... morning? | 07:31 |
@bettyboo | ho serialhex | 07:31 |
Tanmoy | im doing good | 07:32 |
Tanmoy | yeah its morning here | 07:32 |
siddharth | yeah 11 am | 07:33 |
serialhex | ahh, 1:30 am here :D | 07:33 |
Tanmoy | which module r u handling | 07:35 |
serialhex | are you talking to me or sidd?? | 07:37 |
siddharth | well I am currently working on svm classifier | 07:38 |
Tanmoy | oh i meant both :D | 07:38 |
serialhex | ahh... i'm (hopefully) going to be working on the ruby bindings... much to learn before then! | 07:40 |
Tanmoy | the lapack issue is still not resolved right ? | 07:42 |
* serialhex is probably the oldest "student" but also the one with the least skill... he weeps because of his lack of knowlege! ;( | 07:42 | |
Tanmoy | how old r u serialhex | 07:43 |
serialhex | 28 | 07:43 |
serialhex | and i'm barely an undergrad :P | 07:43 |
Tanmoy | ahh beat u | 07:43 |
Tanmoy | 29 | 07:43 |
serialhex | really?? | 07:43 |
Tanmoy | nd postgrad | 07:43 |
serialhex | dammit!!! | 07:43 |
Tanmoy | u at school ? | 07:44 |
serialhex | yeah, i've just started back at school, after many years of goofing off and generally beeing a leech! | 07:44 |
* serialhex is technically still a leech, just a leech on the gov'ment and not his parents | 07:45 | |
siddharth | m doing MS | 07:46 |
Tanmoy | me 2 | 07:46 |
siddharth | ya I know | 07:46 |
serialhex | ooh!!! competition!!! | 07:46 |
* serialhex gets som popcorn to watch the fight :P | 07:47 | |
siddharth | u also from IIIT :P | 07:47 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 07:47 | |
serialhex | ...even more interesting! | 07:47 |
siddharth | serialhex having fun :D | 07:47 |
serialhex | always siddharth!!!! | 07:48 |
Tanmoy | yup :D | 07:48 |
serialhex | except when my cat decids to walk all over my stuff!!! GRR!!! | 07:48 |
siddharth | another IIITian has come... | 07:48 |
siddharth | akhil_ wassup | 07:48 |
Tanmoy | siddharth which module u handling | 07:49 |
siddharth | Integrating SVNSGD-QN, and ollivier Primal SVM solver | 07:50 |
siddharth | SVMSGDQN* | 07:50 |
siddharth | what about u Tanmoy | 07:52 |
Tanmoy | SVM-SO with CRF | 07:52 |
* serialhex is lost with the acronyms being thrown around... | 07:53 | |
Tanmoy | SO Structured output | 07:53 |
Tanmoy | CRF Conditional Random Fields | 07:53 |
serialhex | ahh, i see!! | 07:54 |
* serialhex is a little less confused | 07:54 | |
serialhex | allright, i'm off to bed all | 07:57 |
@bettyboo | cu | 07:57 |
siddharth | good night | 07:57 |
serialhex | g'nite! and lets hope sonney2k gets a few good nights rest also! | 07:57 |
siddharth | clever bettyboo | 07:57 |
@bettyboo | siddharth: is there is 13 mentors | 07:57 |
Tanmoy | night | 07:58 |
serialhex | she is getting clever... a little too clever... | 07:58 |
siddharth | yeah | 07:58 |
serialhex | i leave you all with this: http://xkcd.com/329/ | 07:59 |
* serialhex goes to sleep, a little more afraid of bettyboo than he was before | 07:59 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 08:07 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 251 seconds] | 08:11 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 08:34 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 08:47 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 09:09 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 258 seconds] | 09:31 | |
-!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun | 10:27 | |
blackburn | sonney2k: thanks for style fixes | 10:29 |
blackburn | that was an awful mistype in loop | 10:30 |
-!- Ziyuan [~Ziyuan@116.23.212.45] has joined #shogun | 10:31 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 10:33 | |
blackburn | sonney2k: btw, should I make a pull request with three classes: CEvaluation and derivates CAccuracy, CMeanSquaredError? | 10:37 |
blackburn | or better to implement all of these things an then pull it? | 10:38 |
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has left #shogun [] | 11:15 | |
Tanmoy | @sonney2k does GHMM needs to be reimplemented | 11:34 |
Tanmoy | btw the Gaussian.cpp is still throwing error | 12:42 |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 13:59 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 14:00 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 14:23 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 14:39 | |
alesis-novik | Tanmoy, it's because the patch hasn't been applied yet | 14:42 |
alesis-novik | And good morning | 14:42 |
blackburn | alesis-novik: just woked up? ;) | 14:43 |
alesis-novik | More or less | 14:43 |
alesis-novik | have been feeling broken these last few days | 14:43 |
blackburn | I don't know what is the error you are talking about | 14:44 |
blackburn | alesis-novik: will it appear in case of atlas is installed? | 14:45 |
alesis-novik | It will appear in case of ATLAS not being installed | 14:45 |
blackburn | hmm but it compiles on my | 14:45 |
alesis-novik | when you run configure, what do you get for the LAPACK stuff? | 14:46 |
blackburn | minute.. | 14:46 |
blackburn | Checking for Vanilla lapack and blas support ... no (missing some features: LDA, liblinear, ...) | 14:47 |
alesis-novik | ah | 14:47 |
blackburn | alesis-novik: this? | 14:47 |
alesis-novik | so you don't have LAPACK at all | 14:47 |
blackburn | aha :) | 14:48 |
alesis-novik | I think it breaks for people with Vanilla, but not with ATLAS or none at all | 14:48 |
blackburn | btw, will install it :D | 14:48 |
blackburn | have installed it, now will make | 14:51 |
alesis-novik | don't forget to reconfigure :D | 14:51 |
blackburn | yeap | 14:51 |
blackburn | oh, I think I see now why KRR wasn't working on my shogun :D | 14:52 |
blackburn | Oh! it doesn't compiling now :D | 14:55 |
* blackburn said some nonsense | 14:56 | |
blackburn | it isn't | 14:56 |
blackburn | alesis-novik: yeah, now I see the problem you talking about :D | 14:57 |
@bettyboo | HA | 14:57 |
alesis-novik | You could always install ATLAS :D | 14:58 |
blackburn | aha, already did. | 14:58 |
alesis-novik | It does work with atlas though | 14:59 |
alesis-novik | It shouldn't produce any compilation errors if you have ATLAS installed | 15:00 |
blackburn | ehh | 15:00 |
blackburn | ah, I see | 15:00 |
blackburn | sorry | 15:00 |
serialhex | thats interesting... i just installed atlas & checked for blas but when i configure it tells me that neither are there | 15:01 |
alesis-novik | ah | 15:01 |
alesis-novik | how did you install it? | 15:01 |
blackburn | btw what is the package of atlas on ubuntu? | 15:01 |
alesis-novik | libatlas with different endings | 15:02 |
serialhex | through synaptic... | 15:02 |
blackburn | will install all the things :D | 15:02 |
serialhex | i'll tell you what packages in a min... but they were the sse2 versions | 15:02 |
blackburn | serialhex: no need, I'm installing all of these *atlas now | 15:03 |
blackburn | but atlas support - no | 15:03 |
blackburn | hm! | 15:03 |
alesis-novik | The thing is when you install with synaptic it adds the headers to *whatever*/include/atlas/* | 15:03 |
alesis-novik | so you need to add that to the compiler include path | 15:03 |
alesis-novik | otherwise it tries using #include "lapack.h" and it doesn't work, because it should be #include "atlas/lapack.h" | 15:04 |
blackburn | grhh | 15:04 |
blackburn | alesis-novik: seems to be as good as натягивать глаз на жопу | 15:06 |
serialhex | ..thats interesting, i font have a lapack.h in /usr/include or /usr/include/atlas | 15:06 |
* serialhex thinks blackburn is sending secret messages in russian :D | 15:07 | |
blackburn | serialhex: I don't have one too | 15:07 |
serialhex | :-/ | 15:08 |
blackburn | serialhex: just said it is similar to pull someone's eye to his ass | 15:08 |
alesis-novik | the header file is clapack.h, if that makes any difference | 15:09 |
serialhex | very interesting imagery, i'm sure it flows better in russian, and i cant think of an equivalent we might have in english | 15:09 |
serialhex | yeah, i've got that in /usr/lib/atlas | 15:09 |
serialhex | & cblas.h ... so where do i tell this funky thing where to look??? | 15:10 |
blackburn | what should we do to add this folder for include? | 15:10 |
alesis-novik | Well, personally since I use bash, I added this to .bashrc | 15:14 |
alesis-novik | CPLUS_INCLUDE_PATH=/usr/include/atlas/ | 15:14 |
alesis-novik | export CPLUS_INCLUDE_PATH | 15:14 |
alesis-novik | and | 15:14 |
alesis-novik | C_INCLUDE_PATH=/usr/include/atlas/ | 15:14 |
alesis-novik | export C_INCLUDE_PATH | 15:14 |
serialhex | :-/ maybe the configure script should be set up to look for this? | 15:15 |
alesis-novik | serialhex, I prefer changing env variables, but I guess you could do that | 15:17 |
-!- dvevre [b49531e5@gateway/web/freenode/ip.180.149.49.229] has joined #shogun | 15:17 | |
serialhex | well if it works this way when people use synaptic to install atlas, then a dev cant expect a user to change environment variables to use their software | 15:18 |
blackburn | serialhex: ./configure --enable-debug --interfaces=libshogun,python_modular --includes=/usr/include/atlas | 15:18 |
blackburn | serialhex: it helped | 15:18 |
serialhex | if we know there is a problem & can fix it for future people, why not just do that? | 15:18 |
blackburn | but I think somebody should fix it | 15:19 |
blackburn | serialhex: yeap | 15:19 |
serialhex | yeah, doing the environment variable thing worked too, compiling now... | 15:19 |
alesis-novik | I see your point serialhex | 15:19 |
blackburn | serialhex: I don't like this way :) | 15:19 |
Tanmoy | but yeah i need to get it now | 15:19 |
Tanmoy | well yeah initially i tht too i had atlas | 15:19 |
serialhex | (though it never gave me the msg for vanilla lapack & blas) | 15:20 |
Tanmoy | k it seems to work fine | 15:35 |
* serialhex is still compiling - 1 core @ 3ghz isnt all that fast anymore :( | 15:40 | |
* blackburn too | 15:45 | |
alesis-novik | any idea when sonney2k will be around? | 15:48 |
-!- dvevre [b49531e5@gateway/web/freenode/ip.180.149.49.229] has quit [Quit: Page closed] | 16:08 | |
blackburn | alesis-novik: I have no idea but waiting for him too :) | 16:18 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 16:56 | |
Ziyuan | Why CGaussianKernel is a subclass of CDotKernel but not of CDistanceKernel? | 17:15 |
-!- siddharth [~siddharth@117.211.88.150] has quit [Remote host closed the connection] | 17:37 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 17:58 | |
Ziyuan | And why the "distance" and "width" in DistanceKernel.h is private? | 18:58 |
Ziyuan | Shouldn't it be protected? | 18:59 |
-!- blackburn [~qdrgsm@109.226.117.183] has quit [Quit: Leaving.] | 19:13 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 19:38 | |
-!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun | 20:12 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 20:36 | |
@sonney2k | back | 20:45 |
blackburn | oh! nice | 20:46 |
blackburn | sonney2k: if will have time please take a look on 'new' evaluation. if everything is all right will implement some more 'evaluators' | 20:47 |
@sonney2k | blackburn, reading your diff | 20:50 |
blackburn | sonney2k: then you could just comment it right here ;) | 20:51 |
blackburn | I will have a trouble with implementing ROC and etc | 20:52 |
blackburn | evaluate returns float64_t.. | 20:53 |
blackburn | see.. | 20:54 |
blackburn | will fix it right now | 20:54 |
blackburn | sonney2k: yesterday, when testing accuracy noticed that error with gaussiannaivebayes, very sorry for missed it | 20:57 |
@sonney2k | blackburn, ok kommented | 21:01 |
@sonney2k | commented that is | 21:02 |
@sonney2k | blackburn, what error with GNB? | 21:02 |
blackburn | sonney2k: eh.. I fixed it last night | 21:02 |
blackburn | if labels was e.g. 9,10,11,.. | 21:02 |
blackburn | it will fail | 21:02 |
blackburn | it was because for(int i= >1< instead of 0.. | 21:04 |
@sonney2k | ok | 21:04 |
@sonney2k | ahh yes | 21:04 |
@sonney2k | alesis-novik, pong | 21:05 |
blackburn | anyway, it's working now | 21:05 |
blackburn | sonney2k: using sign there won't be capability of multiclass accuracy estimation | 21:08 |
blackburn | is it ok? | 21:09 |
@sonney2k | blackburn, yes - that is why we need another MultiClassAccuracy ... | 21:09 |
blackburn | sonney2k: okay, thank you | 21:09 |
@sonney2k | if you have a better idea - let me know | 21:09 |
@sonney2k | ( I mean fuzzyly one could check if labels are all integers) | 21:09 |
blackburn | sonney2k: SVM outputs 1.0 and -1.0? | 21:10 |
blackburn | but how I can check it? | 21:11 |
alesis-novik | Hello sonney2k | 21:11 |
@sonney2k | blackburn, they are not +/-1 | 21:11 |
@sonney2k | alesis-novik, hi - what's up? | 21:11 |
@sonney2k | blackburn, they are in R | 21:11 |
blackburn | sonney2k: ok. | 21:11 |
alesis-novik | I've submitted the changes to the pull request | 21:11 |
@sonney2k | but for multiclass we know that they are alll integers >= 0 | 21:12 |
@sonney2k | alesis-novik, ok looking into it | 21:12 |
blackburn | sonney2k: what way do you prefer? | 21:12 |
alesis-novik | Now working on the simple EM version. I was wondering, should I name it GMM or EMGMM and should it be under distributions or clustering? Thoughts? | 21:13 |
@sonney2k | blackburn, I think having MultiClassAccuracy to is easiest | 21:14 |
@sonney2k | (then your code is ok already) | 21:14 |
blackburn | sonney2k: ok, will implement it | 21:14 |
@sonney2k | alesis-novik, is it more clustering or distribution? | 21:17 |
blackburn | sonney2k: may be sth like set_multiclass(true)? | 21:17 |
blackburn | ah, nevermind :) | 21:17 |
alesis-novik | sonney2k, I think it's quite similar to K-Means, so clustering I guess | 21:18 |
* blackburn thinks EM is more clustering too | 21:18 | |
@sonney2k | ok then... | 21:19 |
dvevre | sonney2k: hey! did you get the online features class i mailed yesterday? | 21:21 |
@sonney2k | dvevre, I didn't see the pull request?! | 21:21 |
dvevre | sonney2k: wasn't a pull request... i sent it through email :) | 21:22 |
dvevre | can't send a pull request until i think the design's final | 21:22 |
@sonney2k | alesis-novik, looks like great work | 21:25 |
alesis-novik | thanks sonney2k | 21:26 |
CIA-110 | shogun: Alesis Novik master * r68a9604 / (3 files in 2 dirs): Changed constructor and init for PCACut, changed GUIPreProc so it would compile - http://bit.ly/frs8pL | 21:26 |
CIA-110 | shogun: Soeren Sonnenburg master * r47f74d3 / (6 files in 4 dirs): Merge branch 'master' of https://github.com/alesis/shogun - http://bit.ly/gm6Fez | 21:26 |
alesis-novik | I was thinking of slightly changing the PDF computation to actually first compute the log-PDF | 21:27 |
blackburn | sonney2k: done. | 21:31 |
blackburn | sonney2k: ah, sorry, there is no '\sign' in latex | 21:34 |
blackburn | sonney2k: now done truly | 21:37 |
alesis-novik | sonney2k, any suggestions on what should be the GMM superclass? | 21:52 |
blackburn | alesis-novik: is EM your project? | 21:53 |
alesis-novik | blackburn, yes | 21:53 |
blackburn | nice | 21:53 |
alesis-novik | I mean hopefully :D | 21:54 |
blackburn | alesis-novik: btw, you may design some hierarchy for clustering, now it seems to be not so good | 21:54 |
blackburn | the only problem - K means use CDistanceMachine just like KNN.. | 21:55 |
blackburn | alesis-novik: seems you have more chances than that guy ;) | 21:57 |
alesis-novik | Не дели шкуру неубитого медведя. :D | 21:58 |
blackburn | Я просто хочу чтобы был ты, а не представитель одной прекрасной страны :) | 21:58 |
@bettyboo | yeah! | 21:59 |
alesis-novik | :D /high5 bettyboo | 22:00 |
@bettyboo | alesis-novik: samara, or moscow, same time | 22:00 |
Ziyuan | sonney: I submitted a patch for Bessel kernel | 22:01 |
Ziyuan | And how use git send-email to send patches as attachments (I don't find the option) | 22:02 |
alesis-novik | Ziyuan, did you submit a pull? | 22:04 |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 22:04 | |
Ziyuan | submitted to sonney, actually... | 22:08 |
Ziyuan | via mail | 22:08 |
alesis-novik | Ziyuan, I think it's easier for him to comment/apply/not-loose-in-all-the-mail the patches if you submit them via pull request (unless you had some sort of arrangement) | 22:09 |
@sonney2k | Ziyuan, coulnd' you use jn() instead of the cyl_bessel_j ? | 22:11 |
@sonney2k | then we don't need boost | 22:11 |
@sonney2k | Ziyuan, btw, Tanmoy submitted a similar thing - not sure if it is correct https://github.com/shogun-toolbox/shogun/pull/48/files | 22:13 |
@sonney2k | your compute functions looks totally different | 22:13 |
Ziyuan | I follow this http://crsouza.blogspot.com/2010/03/kernel-functions-for-machine-learning.html#bessel | 22:14 |
@sonney2k | alesis-novik, KernelMachine or Distribution (utilizing kernel machine internally) | 22:15 |
Ziyuan | and use width=1/sigma | 22:15 |
@sonney2k | Ziyuan, it lists 2 functions for bessel - the one that is in kernlab and another one | 22:15 |
Ziyuan | to fit the intuition of width... | 22:15 |
Ziyuan | So I should point out which one I used? | 22:16 |
@sonney2k | I simply don't know which one should use... | 22:17 |
@sonney2k | never heard of the bessel kernel before | 22:17 |
Ziyuan | I want to raise my score so picked up a kernel not yet implemented... | 22:20 |
Ziyuan | I've found a paper having something to do with Bessel kernel | 22:26 |
Ziyuan | http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.62.7887 | 22:27 |
alesis-novik | sonney2k, why do you say I should use KernelMachine? | 22:31 |
@sonney2k | alesis-novik, sorry forget it - doesn't make sense | 22:36 |
@sonney2k | Ziyuan, I think your formula makes more sense (at least from the url) | 22:42 |
@sonney2k | can you pleae send a git pull request from github? | 22:43 |
@sonney2k | dvevre, could you please also do it via github? | 22:43 |
dvevre | okay, but now way it can be merged in this condition! | 22:43 |
@sonney2k | Ziyuan, ahh and please add copyright headers | 22:43 |
@sonney2k | dvevre, yes but still it is easier for me | 22:44 |
Ziyuan | Ok | 22:44 |
blackburn | sonney2k: please let me know to continue implement things in 'evaluation' or not | 22:44 |
@sonney2k | blackburn, yes - it is all good now! | 22:44 |
blackburn | sonney2k: should I do it in this pull? | 22:44 |
@sonney2k | blackburn, I would love to replace your new code with the preivous eval functions | 22:45 |
@sonney2k | in one go I am afraid. | 22:45 |
blackburn | eh? | 22:45 |
blackburn | sorry, don't understand about replacement | 22:46 |
@sonney2k | blackburn, the old PerformanceMeasures.* | 22:46 |
@sonney2k | I would replace them with what you have written | 22:47 |
blackburn | sonney2k: ah, you wrote vice versa :) | 22:47 |
@sonney2k | which looks much more readable and extendible this way | 22:47 |
@sonney2k | blackburn, I guess I need some sleep tonight | 22:47 |
alesis-novik | sonney2k, it would kind of make sense to derive GMM from Distribution, though it still would be under clustering | 22:47 |
blackburn | sonney2k: you should ;) | 22:47 |
blackburn | sonney2k: so, I will make it in that pull request, right? | 22:48 |
@sonney2k | alesis-novik, do as you wish - I don't have any strong preference here | 22:48 |
@sonney2k | blackburn, yes | 22:48 |
blackburn | sonney2k: okay, will write to you when it will be done | 22:48 |
* sonney2k yawns | 22:48 | |
blackburn | sonney2k: ah, I still don't know what to do with ROC | 22:48 |
@sonney2k | sorry but I really need sleep | 22:48 |
blackburn | sonney2k: yeap, you should | 22:49 |
blackburn | will ask you later | 22:49 |
@sonney2k | blackburn, copy the code from perfmeasures | 22:49 |
@sonney2k | for now at least | 22:49 |
@sonney2k | anyway | 22:49 |
* sonney2k is off to bed | 22:49 | |
@sonney2k | l8r | 22:49 |
blackburn | sonney2k: I mean design.. | 22:49 |
blackburn | see you | 22:49 |
blackburn | because evaluate is float64_t | 22:49 |
blackburn | whatever :) | 22:49 |
alesis-novik | See you | 22:50 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 23:39 | |
-!- blackburn [~qdrgsm@109.226.117.183] has quit [Quit: Leaving.] | 23:59 | |
--- Log closed Mon Apr 18 00:00:36 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!