--- Log opened Thu May 05 00:00:44 2011 | ||
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 00:20 | |
blackburn | serialhex: hadn't been here for a long! | 00:24 |
---|---|---|
serialhex | yeah i know blackburn, my computer flipped out on me and i've been trying to fix it | 00:24 |
blackburn | serialhex: Soeren got a new baby :) | 00:25 |
serialhex | every so often the windows refuse to redraw unless i _manually_ move them or something! | 00:25 |
serialhex | nice!! is s/he healthy? | 00:25 |
blackburn | I think so, his name is Jan | 00:25 |
serialhex | cool | 00:26 |
* serialhex must congratulate sonney2k when i see him next | 00:26 | |
serialhex | i'm currently setting up a backup system - a laptop wth a broken lcd & slow processor - so if shit wants to be all funky i have something i can work on (in a console only, but still...) | 00:28 |
blackburn | I have one notebook with broken screen :) | 00:29 |
blackburn | he became RAM donor for my new notebook :D | 00:29 |
serialhex | sweet! | 00:32 |
serialhex | yeah, this is a celeron m with 512 meg ram i think... it's pretty crappy, but it'll run a console session just fine (even through ssh) but it'll take FOR FUCKING EVAR to compile stuff! | 00:33 |
serialhex | ...in fact, trying to run 2 ssh sessions (root & user) is being slow :P | 00:35 |
blackburn | err, such a crap :D | 00:37 |
blackburn | damn ARMA models! AEGHEFW | 00:40 |
alesis-novik | blackburn, ARMA models are fun | 01:18 |
blackburn | alesis-novik: oh, may be you could help me? | 01:18 |
alesis-novik | with? | 01:18 |
blackburn | look, why it could be possible that ARMA(2,3) is better than ARMA(3,3)? | 01:19 |
blackburn | why ARMA(3,3) didn't become ARMA(2,3)? | 01:19 |
alesis-novik | Depends, are we talking about training or specifying the model? | 01:19 |
blackburn | training. if we find models, why it could be possible? | 01:20 |
blackburn | I mean if we are finding best models | 01:20 |
blackburn | why ARMA(3,3) not have one zero coef. for AR what makes it ARMA(2,3) | 01:20 |
blackburn | alesis-novik: do you see a 'paradox'? | 01:22 |
alesis-novik | Ah, what do you mean by better? | 01:23 |
blackburn | alesis-novik: eh. yeah, forgot to tell you what I mean | 01:23 |
blackburn | I meant less deviation from sample covariation | 01:24 |
blackburn | alesis-novik: 'better' is more similar covariation of both model and sample | 01:25 |
blackburn | sorry for curvy language I'm getting afraid when talking about math :D | 01:25 |
alesis-novik | so we train models on the sample and the question is why can ARMA(2, 3) be better than ARMA(3, 3) for the same sample? | 01:27 |
blackburn | yeap | 01:28 |
blackburn | alesis-novik: ARMA(2,3) is just a special case of ARMA(3,3), right? | 01:29 |
alesis-novik | I'd say both of them are just different ARMA(p, q) models | 01:30 |
alesis-novik | and it doesn't really make sense for ARMA(3, 3) to perform worse on *training* data than ARMA(2, 3), if we do train the models | 01:31 |
blackburn | alesis-novik: yeah, I'd say they are different but my prof. don't agree :) | 01:32 |
alesis-novik | what do they say? | 01:33 |
blackburn | only one zero coef. get ARMA(3,3) to ARMA(3,2) | 01:33 |
alesis-novik | that's true | 01:33 |
blackburn | alesis-novik: I was asked by my lecturer | 01:33 |
blackburn | with that question, why it could be better | 01:33 |
blackburn | I think the answer is somewhere around word 'better' | 01:34 |
alesis-novik | Well, I'd say the problem is the same as with most things in ML: more parameters - better training fit - might lead to overfitting the data | 01:34 |
blackburn | yeah but he said the answer isn't in ML or econometrics field | 01:35 |
blackburn | I'm really stucked with it :) | 01:35 |
alesis-novik | you could always go for the silly, yet obvious answer: if the data was generated by an ARMA(2, 3) process, then it will be a better fit :D | 01:36 |
blackburn | exactly, but the answer should contain not data reasoning like that | 01:37 |
blackburn | anyway, thank you :) | 01:38 |
blackburn | how are your exams? | 01:38 |
alesis-novik | 4 out of 8 done, next one on Monday | 01:39 |
blackburn | I see | 01:42 |
blackburn | good luck to you | 01:42 |
blackburn | 3-42 here and I'm a bit tired :) | 01:42 |
blackburn | see you | 01:42 |
-!- blackburn [~qdrgsm@188.168.2.211] has quit [Quit: Leaving.] | 01:43 | |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 01:45 | |
-!- sploving [~root@124.16.139.196] has quit [Read error: Operation timed out] | 02:57 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 02:59 | |
-!- sploving [~root@124.16.139.196] has left #shogun [] | 08:39 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Ping timeout: 246 seconds] | 09:19 | |
-!- blackburn [~qdrgsm@188.168.4.36] has joined #shogun | 09:20 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 09:21 | |
-!- blackburn [~qdrgsm@188.168.4.36] has quit [Quit: Leaving.] | 10:47 | |
-!- heiko [~heiko@infole-06.uni-duisburg.de] has joined #shogun | 10:51 | |
-!- heiko [~heiko@infole-06.uni-duisburg.de] has left #shogun [] | 11:37 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 252 seconds] | 12:21 | |
-!- [1]warpy [~warpy@bzq-79-180-56-86.red.bezeqint.net] has joined #shogun | 12:35 | |
-!- [1]warpy [~warpy@bzq-79-180-56-86.red.bezeqint.net] has quit [Read error: Connection reset by peer] | 12:46 | |
-!- [1]warpy [~warpy@bzq-79-180-56-86.red.bezeqint.net] has joined #shogun | 12:47 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Read error: Connection reset by peer] | 14:27 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 14:44 | |
-!- [1]warpy [~warpy@bzq-79-180-56-86.red.bezeqint.net] has quit [Read error: Connection reset by peer] | 16:31 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Ping timeout: 276 seconds] | 17:23 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 17:25 | |
-!- warpy [~warpy@bzq-79-182-12-108.red.bezeqint.net] has joined #shogun | 17:35 | |
-!- warpy [~warpy@bzq-79-182-12-108.red.bezeqint.net] has quit [Quit: HydraIRC -> http://www.hydrairc.com <- Wibbly Wobbly IRC] | 17:45 | |
-!- warpy [~warpy@bzq-79-182-12-108.red.bezeqint.net] has joined #shogun | 17:45 | |
-!- blackburn [~qdrgsm@188.168.2.26] has joined #shogun | 18:18 | |
* sonney2k takes a deep breath | 19:49 | |
@sonney2k | Blackburn, thanks for your help! | 19:54 |
blackburn | sonney2k: I have done nothing :) | 19:55 |
@bettyboo | :^) | 19:55 |
blackburn | sonney2k: how are you? | 19:55 |
blackburn | the only thing - some guy asked example of kernel usage, but he didn't solved his problem yet. anyway I won't answer him now, have to study more now | 19:57 |
@sonney2k | We are all fine | 19:58 |
blackburn | nice to hear that | 19:58 |
@sonney2k | Little sleep tonight though the baby kept us awake every full hour being hungry :-) | 19:59 |
blackburn | :D | 20:01 |
@sonney2k | Internet connection and mobile reception sucks though in this hospital here. | 20:12 |
@sonney2k | And even worse I don't have Shogun on my Android mobile. | 20:12 |
blackburn | why you want to have shogun? :) | 20:13 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 21:05 | |
serialhex | sonney2k: you dont have ssh set up? that what i've done and (hopefully) will eliminate most of my computer woes! things are 90% set back up so i should be back in action soon! | 21:51 |
@sonney2k | serialhex: this phone needs a keyboard. Typing really is not effective. | 22:26 |
serialhex | sonney2k: ahh, i see! and i really hate those on-screen keyboards too, most of them suck! so i take it everything is good over there? | 22:31 |
@sonney2k | Yes indeed all good. | 22:35 |
@sonney2k | Now for some sleep... | 22:35 |
-!- warpy [~warpy@bzq-79-182-12-108.red.bezeqint.net] has quit [Quit: HydraIRC -> http://www.hydrairc.com <- The alternative IRC client] | 22:59 | |
blackburn | good night ^) | 23:24 |
--- Log closed Fri May 06 00:00:45 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!