--- Log opened Thu Apr 07 00:00:36 2011 | ||
-!- lionelc [48cb93da@gateway/web/freenode/ip.72.203.147.218] has quit [Ping timeout: 252 seconds] | 00:33 | |
-!- lionelc [48cb93da@gateway/web/freenode/ip.72.203.147.218] has joined #shogun | 00:35 | |
-!- lionelc [48cb93da@gateway/web/freenode/ip.72.203.147.218] has quit [Ping timeout: 252 seconds] | 00:46 | |
-!- sonney2k [~shogun@87.118.92.43] has quit [Ping timeout: 240 seconds] | 00:55 | |
-!- sonney2k [~shogun@87.118.92.43] has joined #shogun | 00:56 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 00:56 | |
-!- argetlahmsource [~argetlahm@99.149.171.230] has joined #shogun | 01:08 | |
-!- siddharth_ [~siddharth@117.211.88.150] has quit [Ping timeout: 240 seconds] | 01:09 | |
-!- aifargonos [~aifargono@46.18.27.35] has left #shogun [] | 01:27 | |
-!- blackburn [~qdrgsm@188.168.5.128] has left #shogun [] | 02:08 | |
-!- epps [~epps@84.18.157.200] has joined #shogun | 02:32 | |
-!- epps [~epps@84.18.157.200] has quit [Changing host] | 02:32 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 02:32 | |
-!- ameerkat [HydraIRC@149-169-243-3.nat.asu.edu] has joined #shogun | 02:36 | |
-!- lionelc [48cb93da@gateway/web/freenode/ip.72.203.147.218] has joined #shogun | 02:53 | |
serialhex | thats sad... i'm gone like 4 hours and theres NO talk on this channel... :'( i cry for the lonly irc channel | 02:55 |
---|---|---|
josip | timezones? | 02:55 |
serialhex | maybe | 02:55 |
josip | it's 3AM in .de | 02:55 |
serialhex | true... | 02:55 |
serialhex | though i'm sure there are a few people here in amerika where it's 6-9pm or so | 02:56 |
serialhex | (i'm on the east coast, so it's 9 here) | 02:56 |
josip | hehe, maaybe :) | 02:57 |
@bettyboo | lol ;D | 02:57 |
serialhex | bettyboo: that time zone are you in?? | 02:57 |
@bettyboo | serialhex: that means for lua it is really only the typemaps that are left to do -> good news | 02:57 |
serialhex | ...lts see how smart this bot is | 02:57 |
josip | so I'm not the only night bird | 02:57 |
serialhex | 9pm is _hardly_ night IMO | 02:57 |
josip | I want to ask a question about gsoc, you ask for plenty of details about us but the limit is 1000 words. What would be a split that you expect stuff_about_us/project_description ? | 02:58 |
* serialhex just had redbull & 5 hour energy coctail | 02:58 | |
serialhex | i'm not part of the team going through it all... but i think it's 1000 words for the desc of the project... not for you, but idk | 02:59 |
josip | it says excluding bio | 02:59 |
serialhex | when it's a *reasonble* hour in germany you should ask sonney2k | 02:59 |
josip | hm, okay | 02:59 |
serialhex | yeah... thats what i though | 02:59 |
serialhex | t | 02:59 |
josip | I'm also in .de, but I guess I hae a different biological clock | 03:00 |
josip | s/hae/have | 03:00 |
serialhex | ahh... ok :P | 03:00 |
@bettyboo | yep | 03:00 |
josip | bettyboo: was that an answer to my question? | 03:00 |
@bettyboo | josip: whats the answer to life, the universe and everything? | 03:00 |
serialhex | so yeah, then you are *definitately* a night bird :P | 03:00 |
josip | 42 | 03:00 |
@bettyboo | lol | 03:00 |
serialhex | ooh, she can ask the good questions... | 03:01 |
josip | do all the "what I've done " stuff counts in those 1000 words? | 03:01 |
serialhex | i think that counts as the bio... so no i dont think so | 03:01 |
josip | I better ask tomorrow again at a more sane time | 03:02 |
serialhex | yeah, but you only have today-ish and tomorrow to get it finished & in... so i'd finish & then make any changes later if needed | 03:03 |
josip | yeah . | 03:03 |
* serialhex wonders how many words his application is | 03:04 | |
serialhex | 960 total for the draft, andi cut out some for the actual proposal! *YAY* | 03:06 |
* serialhex does a little dance | 03:06 | |
josip | including bio? | 03:07 |
brocolli | I didn't even have a draft | 03:07 |
brocolli | I wrote it all in that box. | 03:07 |
argetlahmsource | I'm drafting mine up now. I had to get some resources. | 03:07 |
brocolli | Are you an undergrad as well? | 03:08 |
josip | graduating in june :) | 03:08 |
serialhex | josip: yeah, including bio | 03:08 |
@bettyboo | josip, he? | 03:08 |
josip | bettyboo: ? | 03:08 |
serialhex | brocolli: the 500 word box??? | 03:08 |
@bettyboo | josip: ok | 03:08 |
argetlahmsource | brocolli: Yes, graduate Spring 2012. | 03:09 |
brocolli | There's no 500 word box, the questionaire is right there you can jsut fill it out | 03:09 |
brocolli | no limit from what I saw | 03:09 |
* serialhex feels like the smallest midget... | 03:09 | |
brocolli | Lol I doubt you are my friend, seems like there are quite a few undergrads here. | 03:10 |
josip | serialhex: in highschool? | 03:10 |
josip | :) | 03:10 |
serialhex | yeah, theres the project name, short description (500 werdz) and then the cool formatified box | 03:10 |
serialhex | josip: not quite, but the first year back in college... working on an A.A. & A.S. :P | 03:11 |
josip | AA/AS ? | 03:11 |
-!- ameerkat [HydraIRC@149-169-243-3.nat.asu.edu] has quit [Ping timeout: 240 seconds] | 03:12 | |
serialhex | why on EARTH mathematics is an associates in ARTS IDFK... but whatever! theres not enough math in the computer science A.S. for my liking | 03:12 |
serialhex | yeah, i'm going to a 2 year local college and going to upgrade to a 4 year to get a bachelors/masters later... | 03:13 |
josip | I don't know what AA/AS stands for | 03:13 |
argetlahmsource | Associates Degree of Arts and Science. | 03:13 |
josip | okay, got it | 03:14 |
josip | thanks | 03:14 |
serialhex | ahh, yea sorry :P | 03:14 |
serialhex | josip: it's like the lowest degree type one can get here in the states | 03:16 |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 03:33 | |
-!- ameerkat [HydraIRC@149.169.217.27] has joined #shogun | 03:49 | |
-!- argetlahmsource [~argetlahm@99.149.171.230] has quit [Ping timeout: 264 seconds] | 04:05 | |
-!- argetlahmsource [~argetlahm@99.149.171.230] has joined #shogun | 04:22 | |
-!- argetlahmsource [~argetlahm@99.149.171.230] has left #shogun [] | 04:24 | |
-!- epps [~epps@unaffiliated/epps] has quit [Ping timeout: 246 seconds] | 04:42 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 04:57 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 04:58 | |
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has joined #shogun | 04:58 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has joined #shogun | 05:06 | |
-!- epps [~epps@84.18.157.200] has joined #shogun | 05:08 | |
-!- epps [~epps@84.18.157.200] has quit [Changing host] | 05:08 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 05:08 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has left #shogun [] | 05:13 | |
-!- chlin [8161e1b0@gateway/web/freenode/ip.129.97.225.176] has quit [Ping timeout: 252 seconds] | 05:14 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 05:17 | |
-!- ameerkat [HydraIRC@149.169.217.27] has quit [Ping timeout: 246 seconds] | 05:44 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 06:00 | |
alesis-novik | Any devs around? | 06:10 |
@sonney2k | brocolli, it truely is difficult to find ML'ers that can code too. | 06:33 |
epps | heh | 06:46 |
alesis-novik | Hmm, I installed ATLAS using apt, but shogun doesn't recognise | 06:58 |
alesis-novik | Possibly because the headers are in include/atlas/ | 06:58 |
@sonney2k | alesis-novik, you need blas & atlas | 07:03 |
@sonney2k | alesis-novik, check the configure.log | 07:03 |
@sonney2k | it tries to compile a small example | 07:04 |
* sonney2k checks the melange site and notes that the number of applications went down by 12 (the one from the one student applying for 12 ideas) | 07:05 | |
* sonney2k is relieved | 07:05 | |
alesis-novik | Actually, I changed configure from include <clapack.h> to <atlas/clapack.h> | 07:05 |
alesis-novik | seems to work now | 07:05 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 07:07 | |
alesis-novik | I guess the atlas dir wasn't set up in the include path | 07:09 |
alesis-novik | oh right, a question: | 07:10 |
alesis-novik | I want to implement pdf for multi-variate Gaussian, should I just create a static method or write a general pdf class and then one of the virtual methods to be the actual compute pdf for point? | 07:11 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 07:22 | |
alesis-novik | any suggestions sonney2k? | 07:26 |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Quit: No Ping reply in 180 seconds.] | 07:37 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 07:37 | |
-!- epps [~epps@unaffiliated/epps] has quit [Read error: Connection reset by peer] | 08:12 | |
-!- epps [~epps@84.18.157.200] has joined #shogun | 08:19 | |
-!- epps [~epps@84.18.157.200] has quit [Changing host] | 08:19 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 08:19 | |
-!- Ryaether [~Ryaether@50-80-170-245.client.mchsi.com] has left #shogun [] | 08:45 | |
@sonney2k | alesis-novik, write a general pdf class indeed... have a look at CDistribution though - it might be close to what you need | 08:49 |
epps | sonney2k, would time series analysis be a good idea for GSoC? | 08:54 |
@sonney2k | epps, I am not an expert in this subject and don't have any mentors in mind... | 08:55 |
@sonney2k | epps, so probably not something we could do this year (due to lack of mentors) | 08:57 |
epps | ok | 08:57 |
epps | I'm already working on it for a school project (in java), maybe after GSoC I'll implement it in shogun | 08:59 |
@sonney2k | epps, you are most certainly welcome - there is no code related to this in shogun right now - so you could define the standard even... | 09:01 |
@sonney2k | gtg | 09:03 |
@sonney2k | l8r | 09:03 |
epps | later | 09:03 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 260 seconds] | 09:55 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has joined #shogun | 09:58 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has left #shogun [] | 10:07 | |
-!- epps [~epps@unaffiliated/epps] has quit [Ping timeout: 276 seconds] | 10:58 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 11:11 | |
-!- Ziyuan [~Ziyuan@116.22.39.134] has joined #shogun | 11:58 | |
-!- Bautista [a174f152@gateway/web/freenode/ip.161.116.241.82] has joined #shogun | 12:41 | |
-!- Bautista [a174f152@gateway/web/freenode/ip.161.116.241.82] has quit [Client Quit] | 12:43 | |
serialhex | ...anyone home?? | 12:47 |
serialhex | such a lonly irc channel this is, wouldnt you agree bettyboo?? | 12:50 |
@bettyboo | serialhex: gsoc is a great project | 12:50 |
serialhex | bettyboo: yes it is! | 12:50 |
@bettyboo | serialhex: so you don't have the lua developer files installed? | 12:50 |
@bettyboo | rock on | 12:50 |
serialhex | bettyboo: no i do not | 12:51 |
@bettyboo | serialhex: Getting comfortable with ML concepts most definitely does not happen overnight. | 12:51 |
serialhex | bettyboo: what is your gsoc project? | 12:51 |
@bettyboo | serialhex: i am not sure if this has been the case before you added gsoc | 12:51 |
@bettyboo | what? | 12:51 |
serialhex | bettyboo: i agree with you totally! | 12:51 |
@bettyboo | serialhex: I will send a proposal about it | 12:51 |
serialhex | bettyboo: interesting | 12:51 |
@bettyboo | serialhex: have you read Information Theory, Inference & Learning Algorithms by David MacKay?? i've started it and it's pretty interesting so far | 12:51 |
serialhex | bettyboo: yes i have read that... in fact i SAID that :P | 12:52 |
@bettyboo | serialhex: you have to git clone and compile again yes. | 12:52 |
serialhex | bettyboo: ...been there, done that, got the t-shirt | 12:52 |
@bettyboo | serialhex: Q has been e-mailed... and sleep is on its way (1:20 here... 7:20 there :P ) | 12:52 |
serialhex | bettyboo: i said that too :D | 12:52 |
@bettyboo | serialhex: I just went through the documentation of simple svm | 12:52 |
@bettyboo | serialhex, lol | 12:53 |
serialhex | bettyboo: you remember me!!! :') | 12:53 |
@bettyboo | serialhex: I just wonder how github put a link to me :) | 12:53 |
serialhex | bettyboo: i dont know | 12:53 |
@bettyboo | serialhex: *undergrad... though as far as statistics go, i dont know the ratio... | 12:53 |
@bettyboo | what :> | 12:53 |
siddharth | serialhex, bettyboo is bot :P | 12:57 |
@bettyboo | siddharth: A question: has there been much interest in the implementation of kernel feature analysis? | 12:57 |
serialhex | yeah, i know | 12:57 |
siddharth | I think bettyboo stores all chats | 12:58 |
@bettyboo | siddharth: i think that counts as the bio... so no i dont think so | 12:58 |
siddharth | and randomly answers | 12:58 |
serialhex | it would seem so... and any time you say her name she responds | 12:58 |
siddharth | yeah | 12:58 |
* sonney2k giggles | 12:59 | |
siddharth | also if you make emoticons | 12:59 |
siddharth | :D | 12:59 |
serialhex | true! i've noticed that too | 12:59 |
serialhex | :P | 12:59 |
siddharth | zzz did not answer :P | 12:59 |
serialhex | she's ignoring us | 12:59 |
siddharth | HAHA | 12:59 |
@sonney2k | :-) | 12:59 |
@bettyboo | HA ;> | 12:59 |
siddharth | ROFL | 12:59 |
siddharth | hi sonney2k | 13:00 |
@bettyboo | hiho | 13:00 |
@sonney2k | hi | 13:00 |
siddharth | I am trying to implement wavelet kernel :) | 13:00 |
serialhex | hows the weather in germany? | 13:00 |
@bettyboo | ;D siddharth | 13:00 |
siddharth | will that do good instead of patch? | 13:00 |
@sonney2k | siddharth, yes that would be cool! | 13:01 |
siddharth | ok :) | 13:01 |
serialhex | you are in germany right sonney2k? | 13:02 |
@sonney2k | serialhex, really good 21 C | 13:02 |
serialhex | sweet! | 13:03 |
@sonney2k | blue sky - I am sitting outside looking at the tv-tower well and do some work of course :) | 13:03 |
siddharth | 40 C in India :D | 13:03 |
* serialhex cant do the C to F conversion in his head but is taking sonney2k's word for it | 13:03 | |
serialhex | damn!!! | 13:03 |
siddharth | LOL | 13:03 |
serialhex | it's nice and cool here in florida | 13:04 |
serialhex | (and like 7 am... wtf amd i doing up at 7am???) | 13:04 |
@sonney2k | serialhex, ask knrrrd to put a C -> F conversion module into betty | 13:04 |
@bettyboo | sonney2k: others may know better than me | 13:04 |
siddharth | haha | 13:04 |
@bettyboo | ;> | 13:04 |
siddharth | bettyboo rocks | 13:05 |
@bettyboo | siddharth: machine learning rocks! {;)} | 13:05 |
siddharth | OMG | 13:05 |
@sonney2k | knrrrd must have done something to her ... | 13:05 |
serialhex | and The Great Google Responds: 21 degrees Celsius = 69.8 degrees Fahrenheit | 13:05 |
serialhex | so yeah, pretty nice | 13:06 |
siddharth | try 40 C :P | 13:06 |
serialhex | yeah... oww! 40 degrees Celsius = 104 degrees Fahrenheit | 13:06 |
serialhex | but the big Q i have is: is it humid?? | 13:06 |
siddharth | nope thats a relief | 13:06 |
serialhex | i was in nevada once and it was like 110F outside but felt nice cause there was no humidity! | 13:07 |
serialhex | sweet! | 13:07 |
siddharth | hopefully rain will come soon :) | 13:07 |
@bettyboo | ;D | 13:07 |
serialhex | when i lived in new york during the summer it would get to be onlu ~80-90 and be SWEATING because of the humidity (i lived in the mountains of NY) | 13:08 |
serialhex | that'd be nice, i love rain! | 13:08 |
siddharth | back to work :) | 13:08 |
serialhex | alright... have fun the both of ya! :D | 13:09 |
@bettyboo | serialhex, yeah | 13:09 |
-!- eppz [~epps@unaffiliated/epps] has joined #shogun | 13:09 | |
-!- epps [~epps@unaffiliated/epps] has quit [Read error: Connection reset by peer] | 13:10 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 13:11 | |
* Ziyuan hello | 13:14 | |
@sonney2k | hi... | 13:16 |
* serialhex waves at Ziyuan | 13:20 | |
siddharth | sonney2k, to show changes we have to do make install right? | 13:24 |
@sonney2k | ? | 13:24 |
siddharth | I have made .cpp and .h file for the kernel function...now to use it I have to compile? | 13:25 |
siddharth | files are in src/kernel | 13:26 |
-!- dvevre_ [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 13:33 | |
siddharth | sonney2k, well i did make install and now I can use the kernel function which I made | 13:36 |
@sonney2k | siddharth, ahh now I understand | 13:36 |
@sonney2k | which interface language are you using? | 13:36 |
@sonney2k | python_modular? | 13:36 |
siddharth | matlab | 13:37 |
@sonney2k | then you also have to open modular/Kernel.i | 13:37 |
@sonney2k | and include your kernel there (look at how its fone for GaussianKernel) | 13:37 |
@sonney2k | for static interfaces like matlab you have to modify libshogunui/GUIKernel.cpp / SGInterface.cpp - not as nice but will work :) | 13:38 |
siddharth | sonney2k, I think its working fine...just replaced gaussiankernel with Wavelet kernel in Kernel_gaussian code | 13:40 |
@sonney2k | siddharth, well it still needs to be found in the interfaces right? So you changed that too? | 13:41 |
siddharth | yeah m working on that | 13:41 |
@sonney2k | ok | 13:43 |
josip | hello, any devs around? :) | 14:05 |
josip | I wanted to ask you about the 1000 word limit on the proposal. Does that include the sections about me? (i.e. what I've done, courses, project work etc.) | 14:06 |
-!- aifargonos [~aifargono@46.18.27.4] has joined #shogun | 14:17 | |
-!- josip [~josip@unaffiliated/josip] has quit [Remote host closed the connection] | 14:26 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has joined #shogun | 15:02 | |
hiteshk | sonney2k , hi | 15:02 |
-!- josip [~josip@212.201.44.246] has joined #shogun | 15:10 | |
-!- josip [~josip@212.201.44.246] has quit [Changing host] | 15:10 | |
-!- josip [~josip@unaffiliated/josip] has joined #shogun | 15:10 | |
-!- josip [~josip@unaffiliated/josip] has quit [Remote host closed the connection] | 15:22 | |
-!- josip [~josip@212.201.44.246] has joined #shogun | 15:27 | |
-!- josip [~josip@212.201.44.246] has quit [Changing host] | 15:27 | |
-!- josip [~josip@unaffiliated/josip] has joined #shogun | 15:27 | |
-!- aifargonos [~aifargono@46.18.27.4] has quit [Read error: Connection reset by peer] | 15:30 | |
-!- aifargonos [~aifargono@46.18.27.4] has joined #shogun | 15:36 | |
josip | sorry, did my question got answered? I ahd some problems with the system | 15:48 |
Ziyuan | It says "Please keep your application short. Ideally below 1000 words (excluding bio, publications, references)." | 15:48 |
Ziyuan | http://www.booki.cc/gsocstudentguide/_v/1.0/proposal-example-2/ | 15:49 |
Ziyuan | Here are two examples: | 15:49 |
Ziyuan | http://www.booki.cc/gsocstudentguide/_v/1.0/proposal-examples/ | 15:49 |
Ziyuan | http://www.booki.cc/gsocstudentguide/_v/1.0/proposal-example-2/ | 15:49 |
josip | okay | 15:53 |
josip | well i also did one last year, but they might vary from an organization to an organization | 15:53 |
Ziyuan | Wow, did you get accepted last year? | 15:58 |
josip | yes | 16:00 |
-!- aifargonos [~aifargono@46.18.27.4] has quit [Ping timeout: 248 seconds] | 16:02 | |
josip | you can apply multiples times to gsoc :) | 16:02 |
-!- blackburn [~qdrgsm@188.168.5.43] has joined #shogun | 16:04 | |
-!- dvevre_ [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 252 seconds] | 16:05 | |
-!- aifargonos [~aifargono@46.18.27.4] has joined #shogun | 16:07 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has left #shogun [] | 16:07 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 240 seconds] | 16:15 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 16:29 | |
-!- eppz [~epps@unaffiliated/epps] has quit [Read error: Operation timed out] | 16:41 | |
-!- eppz [~epps@unaffiliated/epps] has joined #shogun | 16:55 | |
siddharth | sonney2k, need help | 17:05 |
siddharth | I am running kernel_sigmoid.m on toy dataset | 17:06 |
siddharth | but getting this error | 17:06 |
siddharth | ??? Undefined function or method 'sg' for input | 17:06 |
siddharth | arguments of type 'char'. | 17:06 |
-!- aifargonos [~aifargono@46.18.27.4] has quit [Ping timeout: 276 seconds] | 17:06 | |
siddharth | dataset is real valued | 17:07 |
-!- hiteshk [~hitesh_na@117.211.88.150] has joined #shogun | 17:09 | |
@sonney2k | siddharth, sorry I need more context. | 17:10 |
@sonney2k | josip, so is your question sufficiently answered (you were already offline when I saw your question) | 17:11 |
hiteshk | sonney2k , hi | 17:11 |
@bettyboo | hiteshk, hey | 17:11 |
hiteshk | hi | 17:11 |
@sonney2k | hi... | 17:11 |
siddharth | sonney2k, well I am running kernel_gaussian on matlab... | 17:12 |
siddharth | error is in this line | 17:12 |
hiteshk | sonney2k , you had to give me some feedback but not received it yet | 17:12 |
siddharth | sg('set_kernel', 'SIGMOID', 'REAL', size_cache, gamma, coef0); | 17:12 |
siddharth | its saying ??? Undefined function or method 'sg' for input | 17:12 |
siddharth | arguments of type 'char'. | 17:12 |
siddharth | This is the standard Sigmoid kernel computed on dense real valued features. | 17:13 |
siddharth | and m using toy data | 17:13 |
-!- aifargonos [~aifargono@46.18.27.35] has joined #shogun | 17:13 | |
siddharth | code is kernel_sigmoid.m in examples/documented/matlab | 17:16 |
@sonney2k | hiteshk, I sent you an email with detailled comments (yesterday at 10:43) | 17:16 |
@sonney2k | siddharth, you were using matlab right? | 17:17 |
siddharth | yes | 17:17 |
josip | sonney2k: yep, i think i got the idea :) I have bio ~ 700 words < 1000 words description | 17:18 |
siddharth | dataset: fm_train_word.dat | 17:18 |
hiteshk | sonney2k , I didn't receive it could you please send it again | 17:18 |
@sonney2k | hiteshk, resent | 17:18 |
@sonney2k | josip, good :) | 17:19 |
@sonney2k | siddharth, well follow the steps in http://www.shogun-toolbox.org/doc/installation.html | 17:19 |
@sonney2k | there is a | 17:19 |
@sonney2k | export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:path_to_shogun/trunk/src/libshogunui:path_to_shogun/trunk/src/libshogun | 17:19 |
@sonney2k | matlab -desktop | 17:19 |
@sonney2k | in the matlab section | 17:20 |
siddharth | sonney2k, actually i have installed correctly and tested on an example given in the installation guide | 17:20 |
@sonney2k | siddharth, then please type sg('help') | 17:20 |
@sonney2k | from within matlab | 17:20 |
siddharth | ok | 17:20 |
@sonney2k | if that works - shogun is indeed installed | 17:21 |
siddharth | hmm not working | 17:21 |
siddharth | though | 17:21 |
siddharth | start matlab in the src/matlab directory | 17:22 |
siddharth | and type | 17:22 |
siddharth | addpath('../examples/documented/matlab/graphical') | 17:22 |
siddharth | svr_regression | 17:22 |
siddharth | i did this and this was working | 17:22 |
@sonney2k | yeah but then always start matlab from the src/matlab directory and it will work or add this path to your matlab path | 17:22 |
@sonney2k | hiteshk, did you receive my email? | 17:26 |
hiteshk | yes I received , thanks | 17:27 |
@sonney2k | ok good | 17:27 |
hiteshk | what are linadd extensions | 17:27 |
@sonney2k | speedups for the special case where kernel is linear or a string kernel | 17:28 |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 17:31 | |
hiteshk | where can I read about it | 17:32 |
brocolli | sonney2k: lol yeah | 17:32 |
brocolli | sonney2k: MLers who can code are few and far betwee | 17:33 |
brocolli | That's why I'm pretty happy being good at systems and ML | 17:33 |
brocolli | It's a rare skill | 17:33 |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 17:33 | |
@sonney2k | hiteshk, did you see the message with the link (I got disconnected...) | 17:37 |
brocolli | oh | 17:42 |
brocolli | I didn't see you get disconnected | 17:42 |
brocolli | speedups for the special case where kernel is linear or a string kernel | 17:42 |
brocolli | serialhex has quit (Remote host closed the connection) | 17:42 |
brocolli | hiteshkwhere can I read about it | 17:42 |
brocolli | brocollisonney2k: lol yeah | 17:42 |
brocolli | sonney2k: MLers who can code are few and far betwee | 17:42 |
brocolli | That's why I'm pretty happy being good at systems and ML | 17:42 |
brocolli | It's a rare skill | 17:42 |
brocolli | is what is aw | 17:42 |
-!- brocolli was kicked from #shogun by bettyboo [flood] | 17:42 | |
-!- brocolli [ylorb@unaffiliated/brocolli] has joined #shogun | 17:42 | |
serialhex | ??? someone asking for me?? | 17:43 |
@sonney2k | hiteshk, http://sonnenburgs.de/soeren/publications/SonRaeRie07.pdf | 17:44 |
serialhex | brocolli: you asking for me? | 17:44 |
@sonney2k | <sonney2k> and there it is algorithm 4.2 | 17:44 |
@sonney2k | <sonney2k> we have implemented that in svmlight? | 17:44 |
@sonney2k | <sonney2k> !! | 17:44 |
serialhex | i guess this is where i say "cool"? :P | 17:44 |
@sonney2k | brocolli, thanks | 17:44 |
* serialhex looks at sonney2k's link | 17:45 | |
serialhex | cool, something for me to read later sonney2k | 17:46 |
@sonney2k | serialhex, there is probably better things to read ... also nowadays we mostly use http://sonnenburgs.de/soeren/publications/SonFra10.pdf | 17:50 |
brocolli | sonney2k: In all seriousness, are there THAT many MLers that are strictly theoretical? | 17:52 |
serialhex | sonney2k: ahh, sweet!!! | 17:52 |
brocolli | I mean, didn't they learn from the whole "Vapnik got jacked by the guy who implemented his alg" event? | 17:52 |
* serialhex follows link and saves it | 17:52 | |
@sonney2k | brocolli, well most can write a few lines of matlab at least | 17:53 |
brocolli | matlab doesn't even count | 17:53 |
brocolli | I noly use it for publication stuff since it makes life easy | 17:53 |
brocolli | but for real performance I wouldn't use it, I'd write my stuff in C >< | 17:53 |
brocolli | or C++ | 17:53 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 17:53 | |
@sonney2k | brocolli, what is this "Vapnik got jacked by the guy who implemented his alg" event? | 17:53 |
brocolli | Vapnik didn't win the prize | 17:53 |
brocolli | The person who implemented slack margin svm did | 17:53 |
brocolli | not the person who created it. | 17:53 |
@sonney2k | which prize? | 17:54 |
brocolli | It's a fundamental illustration of what you just said: Not many MLers can code | 17:54 |
@sonney2k | and who was that person? | 17:54 |
brocolli | I have to find it, sec | 17:54 |
@sonney2k | Corinna? | 17:54 |
@sonney2k | Or Joachims? | 17:54 |
blackburn | Vapnik is pretty old-school research, the fundamentals of SVMs can be found even in his 1970's books | 17:54 |
brocolli | Pretty sure he officially proposed it 97 | 17:55 |
blackburn | researcher* | 17:55 |
brocolli | Let me find it sonney2k, I'm not certain as to which one | 17:55 |
brocolli | joachims | 17:55 |
blackburn | do you mean Cortes? | 17:55 |
brocolli | thats' right | 17:55 |
blackburn | eh | 17:55 |
brocolli | SVMs up until 97 from what I remember dealt with STRICTLY linear seperable data | 17:56 |
brocolli | That awasn't as useful as slack margin loss for penalty | 17:56 |
* sonney2k recalls having done a practical back then at GMD FIRST where that '97 paper was lying around for review 'support vector networks, cortes & vapnik' | 17:56 | |
brocolli | Which was incredibly useful, since bad classification was penalized by the hinge loss function (convex) and allowed the method to be more usable | 17:56 |
@sonney2k | blackburn, yes | 17:56 |
blackburn | sonney2k: yes about that? | 17:57 |
@sonney2k | brocolli, to be useful at all I would say | 17:57 |
brocolli | lol | 17:57 |
@bettyboo | he | 17:57 |
* sonney2k gets lost | 17:57 | |
brocolli | agreed, I was being nice | 17:57 |
blackburn | sorry, about what I mean | 17:57 |
@sonney2k | or have you ever had separable data? | 17:57 |
brocolli | hahaha | 17:58 |
brocolli | iris1 | 17:58 |
brocolli | ;) | 17:58 |
blackburn | brocolli: kernel trick can be found in Aizerman's works | 17:58 |
brocolli | so? | 17:58 |
brocolli | That's not SVM with hinge loss | 17:58 |
blackburn | ah, you mean kernel trick with SVMs | 17:58 |
blackburn | okay | 17:58 |
brocolli | Kernel trick merely reduces the amount of tedious computation for the inner product | 17:58 |
@sonney2k | and kernel trick is a burden with large scale data | 17:59 |
brocolli | word. | 17:59 |
brocolli | Man this online learning course is hard, sonney2k >< | 18:00 |
brocolli | I should have realized that unsupervised learning (THUMBSDOWN) | 18:00 |
brocolli | hehe | 18:00 |
@sonney2k | anyways I will have to leave | 18:00 |
@sonney2k | l8r | 18:00 |
brocolli | later! | 18:01 |
blackburn | brocolli: don't understand why you wonder about Vapnik didn't implement SVMs :) | 18:03 |
@bettyboo | ;> | 18:03 |
brocolli | I didn't wonder. | 18:03 |
brocolli | It just illustrates what sonney2k said. That many people who are in the area of ML do not know how to program a real language like C++/C. Joachims did | 18:03 |
blackburn | I think, there is no need to program in C++ for Vapnik | 18:04 |
blackburn | furthermore, C++ and C isn't the only 'real' languages at all | 18:05 |
brocolli | He's a statistician, he can't even program or program much | 18:05 |
brocolli | Also | 18:05 |
brocolli | Many statisticians from the older times do not use computers much, and if they do it's not very much. There is a fundamental disconnect in the areas many times, and that's why the great learning theorist ideas are rarely implemented by them. | 18:06 |
brocolli | The rigorousness of being a learning theoriest requires an incredible amount of mathematics, the depth and volume of which rarely allows one to simultaneously learn how to program. | 18:07 |
blackburn | I don't think so, programming is not so difficult for them | 18:07 |
blackburn | most likely they simple do not need it very much | 18:08 |
brocolli | Why wouldn't they? Propose a theory and implementing it is much more complete than giving a theory that's implementation might not even be possible | 18:08 |
blackburn | for example, how can Vapnik implement ideas of computational learning theory? | 18:09 |
brocolli | what? | 18:09 |
brocolli | Are you joking? | 18:09 |
blackburn | I mean what should Vapnik implement? VC-dimesion or something like that? | 18:09 |
brocolli | Michael Kearns implements all his stuff. | 18:10 |
brocolli | He's not necessarily a Learning theorist either, he's a good CSer with sufficient math knowledge to transform his theorems into a profit | 18:10 |
brocolli | Michael Kearns is a PhD in CS so you're again barking up the wrong tree man | 18:10 |
blackburn | eh, up to you :) | 18:11 |
@bettyboo | HA ;D | 18:11 |
brocolli | Vapnik could have implemented his own algorithm, Joachims did a year after officially. It's a disconnect that exist, whether you want to accept that or not is up to you | 18:12 |
blackburn | okay | 18:12 |
brocolli | I am not discrediting Vapnik in any way. I believe he should have gotten the award. | 18:13 |
brocolli | But the fact is, your idea is only as good as it is useful. That's how society will distribute awards. | 18:13 |
blackburn | seems like you just arguing with me, I don't like this way of discussion | 18:15 |
brocolli | Pretty sure you're arguing with me as evidenced by your statement of me saying I wonder why Vapnik didn't implement SVM, when I said no such thing. | 18:19 |
blackburn | may be I didn't understand you right, sorry | 18:20 |
brocolli | it's okay, ecause I know what you're saying | 18:23 |
brocolli | I'm just saying I don't disagree that Vapnik's work is very important and he probably could have implemented it himself | 18:23 |
brocolli | but all I'm saying is that he didn't and as a result he didn't get some prize :/ | 18:23 |
blackburn | I think the main reason why he don't implement own algorithms is his age | 18:24 |
brocolli | He must be elder, no doubt. | 18:26 |
blackburn | I mean he wrote many things about learning even in 70s with no heavy computing at all, especially in ussr | 18:26 |
brocolli | Still a legend, and we are all thankful for his great contribution | 18:26 |
brocolli | Because who knows where we would be without him | 18:26 |
brocolli | Just like Valiant (who made PAC learning) | 18:27 |
blackburn | same thing with legends like McCarthy and Knuth | 18:27 |
blackburn | I'm not sure that they did program very much in modern things like C++ | 18:28 |
blackburn | and so on | 18:28 |
blackburn | hope you will understand my idea right :) | 18:28 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 258 seconds] | 18:37 | |
-!- Mengyun [~dorothyk@S0106001a803d4363.vc.shawcable.net] has joined #shogun | 18:57 | |
-!- Mengyun [~dorothyk@S0106001a803d4363.vc.shawcable.net] has quit [Quit: Leaving] | 19:07 | |
Ziyuan | Application submitted. | 19:14 |
josip | the deadline is in 24hrs, right? | 19:16 |
Ziyuan | The calendar | 19:17 |
Ziyuan | http://www.google-melange.com/gsoc/events/google/gsoc2011 | 19:17 |
Ziyuan | Fri. 12:00 | 19:18 |
josip | so I was right | 19:18 |
josip | 12 that's California time | 19:18 |
Ziyuan | But I am not sure about the time zone | 19:18 |
Ziyuan | Ok | 19:18 |
Ziyuan | It's Friday 1:18 of my place :) | 19:19 |
@bettyboo | Ziyuan, rahaha! ;> | 19:19 |
Ziyuan | :> | 19:19 |
@bettyboo | Ziyuan, rahaha | 19:19 |
Ziyuan | My background is not strong enough, but anyway, I'll try. | 19:20 |
josip | do you apply for one of the proposed projects? | 19:20 |
Ziyuan | Yes, I applied for the "Kernel Feature Analysis" | 19:22 |
josip | not the same one, good | 19:22 |
josip | :)) | 19:22 |
Ziyuan | But I "collide" with Andrew Tereskin | 19:24 |
josip | we all compete with each other of course :) | 19:24 |
josip | not all projects are gonna get accepted | 19:24 |
Ziyuan | Yep | 19:24 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 19:25 | |
Ziyuan | Accord.Net is another Machine Learning framework in C# hosted by ETH Zürich. Together with Shogun, they give me an impression that ML research is prosperous in Europe. | 19:32 |
alesis-novik | Don't forget nice people like Chris M. Bishop | 19:36 |
brocolli | you guys can seeo ther people's proposals? | 19:38 |
alesis-novik | I don't think so brocolli | 19:39 |
brocolli | Oh then how did Ziyuan know that he "collides | 19:41 |
brocolli | with Andrew Tereskin | 19:41 |
Ziyuan | I read it from the mail list | 19:41 |
brocolli | oh | 19:42 |
josip | is it binding that I send an e-mail to the mailing list? | 19:46 |
Ziyuan | I didn't and maybe I should send one. | 19:48 |
alesis-novik | See you later | 19:49 |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 19:49 | |
Ziyuan | Time to sleep, bye~ | 19:57 |
-!- Ziyuan [~Ziyuan@116.22.39.134] has quit [] | 19:57 | |
-!- eppz [~epps@unaffiliated/epps] has quit [Ping timeout: 264 seconds] | 20:29 | |
-!- ameerkat [HydraIRC@149-169-253-50.nat.asu.edu] has joined #shogun | 21:12 | |
-!- Pinio [~IceChat7@90-156-17-126.internetia.net.pl] has joined #shogun | 21:16 | |
-!- ameerkat [HydraIRC@149-169-253-50.nat.asu.edu] has quit [Ping timeout: 276 seconds] | 21:33 | |
-!- ameerkat [HydraIRC@149-169-253-238.nat.asu.edu] has joined #shogun | 21:48 | |
@sonney2k | siddharth, would you mind sending the patch to the mailinglist - and please as git format-patch - thanks! | 21:48 |
@sonney2k | I will then apply your patch (and do some minor changes) | 21:51 |
blackburn | sonney2k: I have another one patch for further development of CKNN, but it is small. should I sent it only with major changes later? | 21:51 |
@sonney2k | blackburn, as you like | 21:52 |
blackburn | sonney2k: okay, sent it to ML | 21:56 |
blackburn | sonney2k: I simply hide this distance initialization to another method. it is because I want to introduce new methods for classify with weights | 21:57 |
blackburn | and there is always will be that initialization | 21:57 |
@sonney2k | blackburn, I understand. I am just wondering if the SG_UNREF in the init is ok | 22:15 |
@sonney2k | hmmhh | 22:15 |
blackburn | sonney2k: sorry. it shouldn't be there? | 22:15 |
blackburn | I thought it is equal to that initialization in classify(CFeatures* data) | 22:16 |
@sonney2k | blackburn, it was there before.... | 22:16 |
blackburn | ah, so not my fault, right? | 22:17 |
@sonney2k | no, and it is actually correct too | 22:17 |
@sonney2k | it is there because of the get_lhs() | 22:17 |
@sonney2k | that increaases the ref count | 22:17 |
@sonney2k | so one has to decref | 22:17 |
blackburn | okay. you scared me a bit :) | 22:17 |
blackburn | that concept of lhs and rhs is a bit curious, especially first time | 22:18 |
@sonney2k | blackburn, not my intention - but better to be overly critical :) | 22:18 |
blackburn | sonney2k: one more question. should I use params in classify_weighted_etc() or only data to be classified and getters and setters for params? | 22:20 |
blackburn | shortly, classify_rank_weighted(CFeatures* data, float64_t q) vs. classify_rank_weighted(CFeatures* data) + set_q(float64_t q) + get_q() | 22:21 |
@sonney2k | blackburn, I would prefer setters / getters. this enables that everything works in the modular interfaces | 22:21 |
blackburn | aha, I see now. thank you | 22:21 |
blackburn | sonney2k: and finally, that commit -- should I sent it one more time after implementing weighted classifying? | 22:23 |
CIA-31 | shogun: Sergey Lisitsyn master * r32186a0 / (2 files): Introduced protected method init_distance(CFeatures* data) for avoiding further code duplication - http://bit.ly/ggNuDU | 22:23 |
blackburn | oh, I don't | 22:23 |
blackburn | :) | 22:24 |
CIA-31 | shogun: Soeren Sonnenburg master * r88ee55b / (2 files): Minor formatting changes - http://bit.ly/ffdO11 | 22:24 |
@sonney2k | I just inserted 2 lines with spaces | 22:25 |
blackburn | sonney2k: sorry with formatting | 22:25 |
blackburn | just getting familiar with code style | 22:25 |
blackburn | there is always an empty line before comments, right? | 22:25 |
@sonney2k | blackburn, don't worry it takes time to adjust to someone else's coding style | 22:26 |
blackburn | sonney2k: thank you very much for answers and your time :) | 22:26 |
blackburn | later will send a patch with weighted classifying | 22:27 |
hiteshk | sonney2k , is linadd already implemented in shogun | 22:49 |
@sonney2k | hiteshk, yes in svmlight | 22:49 |
@sonney2k | would be great to port it over | 22:49 |
hiteshk | is it parallelised | 22:50 |
@sonney2k | yes | 22:52 |
blackburn | that editor on google-melange is horrible one.. | 22:54 |
@sonney2k | not optimal true - it somehow does the job though. | 22:54 |
blackburn | wonder with all the things seems to be right in editor and when submitted looked a very different way | 22:55 |
blackburn | if I could, I'd rather send a .pdf with latex beauty :) | 22:57 |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 22:58 | |
-!- ameerkat [HydraIRC@149-169-253-238.nat.asu.edu] has quit [Read error: Connection reset by peer] | 23:01 | |
@sonney2k | good night everyone. | 23:02 |
blackburn | good night, Sören | 23:02 |
@sonney2k | bettyboo, have fun with all the people in the chat | 23:02 |
@sonney2k | l8r | 23:02 |
@bettyboo | sonney2k: start matlab in the src/matlab directory | 23:02 |
@bettyboo | don't care :> | 23:02 |
blackburn | bettyboo: will you have fun with anybody? | 23:03 |
@bettyboo | blackburn: pretty cool host: @unaffiliated/epps | 23:03 |
@bettyboo | what | 23:03 |
blackburn | bettyboo: seriously? | 23:03 |
@bettyboo | blackburn: have you read Information Theory, Inference & Learning Algorithms by David MacKay?? i've started it and it's pretty interesting so far | 23:03 |
blackburn | bettyboo: I did | 23:03 |
@bettyboo | blackburn: you really are working hard - I mean it :) | 23:03 |
-!- ameerkat [HydraIRC@149-169-253-238.nat.asu.edu] has joined #shogun | 23:07 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 23:10 | |
-!- hiteshk [~hitesh_na@117.211.88.150] has left #shogun [] | 23:16 | |
-!- epps [~epps@unaffiliated/epps] has joined #shogun | 23:19 | |
-!- Chen [cab5d6cd@gateway/web/freenode/ip.202.181.214.205] has joined #shogun | 23:22 | |
alesis-novik | good book | 23:23 |
-!- Chen [cab5d6cd@gateway/web/freenode/ip.202.181.214.205] has quit [Ping timeout: 252 seconds] | 23:38 | |
serialhex | alesis-novik: yeah, i found it online and asked brocolli about it, he didnt know anything about it but bettyboo has been singing its praises ever since! | 23:44 |
@bettyboo | serialhex: however, the basis vector of kpca are dense. | 23:44 |
serialhex | bettyboo: i'm sure they are | 23:44 |
@bettyboo | serialhex: though i'm sure there are a few people here in amerika where it's 6-9pm or so | 23:44 |
serialhex | bettyboo: yeah, there are a few people like that, though it's only ~5:45 here on the east coast | 23:45 |
@bettyboo | serialhex: shogun: (and already obsoleted). We therefore attempt to use lua config and if | 23:45 |
serialhex | bettyboo: and if what? | 23:45 |
@bettyboo | serialhex: shogun is what? | 23:45 |
serialhex | bettyboo: is cool! | 23:45 |
@bettyboo | serialhex: computer science is fun. | 23:45 |
serialhex | bettyboo: yes it is! are you a computer scientist? | 23:45 |
@bettyboo | serialhex: so, yes, it is a bit slowly, but there is always a gain | 23:45 |
alesis-novik | serialhex, it was my information theory course book, but we mostly used the information theory bits. Though I did use it in another ML course as well | 23:45 |
serialhex | bettyboo: lots of gain! | 23:45 |
@bettyboo | serialhex: but what are the weights of each | 23:45 |
serialhex | alesis-novik: sweet, yeah i just really started it, but it's on my "must read" rotation once i took a gander inside | 23:46 |
* serialhex is a n00b when it comes to machine learning | 23:46 | |
serialhex | bettyboo: as long as you're not gaining pounds and retain that girlish figure of yours we'll be ok! | 23:47 |
@bettyboo | serialhex: you are in germany right | 23:47 |
serialhex | bettyboo: no, i'm in florida | 23:47 |
@bettyboo | serialhex: no, I think they prefer email | 23:47 |
serialhex | bettyboo: i sure do | 23:47 |
@bettyboo | serialhex: no, it is that when you have a set of vectors to ensure that they all have norm one. | 23:47 |
@bettyboo | sure | 23:47 |
serialhex | bettyboo: i dont think so, i think they tought you wrong | 23:48 |
@bettyboo | serialhex: I think there's an issue about this somewhere | 23:48 |
serialhex | bettyboo: where dod you go to school? | 23:48 |
@bettyboo | serialhex: ooh, she can ask the good questions... | 23:48 |
serialhex | bettyboo: throwing all my statements back at me are ya?? :P | 23:48 |
@bettyboo | serialhex: my guess is that you won't need all the time to do the newtonsvm if you take that one | 23:48 |
-!- aifargonos [~aifargono@46.18.27.35] has left #shogun [] | 23:57 | |
--- Log closed Fri Apr 08 00:00:36 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!