--- Log opened Wed Jan 18 00:00:19 2012 | ||
blackburn | sonney2k_: w/o page on mloss.org? | 01:26 |
---|---|---|
blackburn | every toolkit has one.. | 01:30 |
-!- in3xes [~in3xes@49.249.163.28] has joined #shogun | 05:37 | |
-!- in3xes [~in3xes@49.249.163.28] has quit [Ping timeout: 252 seconds] | 06:10 | |
-!- blackburn [~qdrgsm@188.168.2.4] has quit [Ping timeout: 272 seconds] | 07:45 | |
sonney2k_ | blackburn *with* one | 08:20 |
CIA-27 | shogun: Soeren Sonnenburg master * rbaa7f11 / src/interfaces/modular/SGBase.i : fix hdf5 serialization check mismatch - http://git.io/Cy0tfA | 08:31 |
shogun-buildbot | build #272 of python_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/272 blamelist: sonne@debian.org | 09:29 |
-!- karlnapf [~heiko@144.82.169.103] has joined #shogun | 11:57 | |
-!- karlnapf [~heiko@144.82.169.103] has quit [Ping timeout: 252 seconds] | 12:02 | |
-!- blackburn [~qdrgsm@188.168.3.91] has joined #shogun | 13:18 | |
blackburn | sonne|work: I didn't get you :( how can I have mloss page with shogun code but with 'EDRT' label? | 13:20 |
sonne|work | you do EDRT label and point to your shogun subpage | 13:21 |
blackburn | sonne|work: but what about source? | 13:22 |
blackburn | and binaries | 13:22 |
blackburn | or so | 13:22 |
sonne|work | point to shogun | 13:23 |
blackburn | sonne|work: ok thanks, I thought it would be strange.. | 13:24 |
-!- blackburn [~qdrgsm@188.168.3.91] has quit [Quit: Leaving.] | 13:53 | |
-!- blackburn [~qdrgsm@188.168.14.65] has joined #shogun | 14:05 | |
-!- karlnapf [~heiko@144.82.169.103] has joined #shogun | 15:14 | |
-!- karlnapf [~heiko@144.82.169.103] has quit [Ping timeout: 245 seconds] | 15:50 | |
-!- sonne|work [~sonnenbu@194.78.35.195] has quit [Read error: Connection reset by peer] | 16:10 | |
-!- sonne|work [~sonnenbu@194.78.35.195] has joined #shogun | 16:47 | |
CIA-27 | shogun: Soeren Sonnenburg master * r299493d / src/interfaces/modular/SGBase.i : | 17:18 |
CIA-27 | shogun: fallback to original pickle/serialization routines if objects are not | 17:18 |
CIA-27 | shogun: instances of SGObject - http://git.io/NVGcYQ | 17:18 |
blackburn | sonney2k_: I just got shogun serialized object with size of 7gb :D | 17:42 |
-!- Ram108 [~amma@14.96.88.10] has joined #shogun | 18:07 | |
blackburn | Ram108: hey how are you? | 18:15 |
Ram108 | hey am fine | 18:15 |
Ram108 | working on neural networks nw :) | 18:15 |
Ram108 | hw r u? :) | 18:15 |
blackburn | fine | 18:16 |
blackburn | what do you do with NN? | 18:16 |
Ram108 | classify....... working on backpropagation algorithm | 18:16 |
Ram108 | fairly easy but slow | 18:17 |
blackburn | I see | 18:17 |
Ram108 | hmmm :) i got a faint idea on hw real brain works nw | 18:17 |
Ram108 | lol gona use that on my robot | 18:17 |
Ram108 | built 1 | 18:18 |
Ram108 | lol :P | 18:18 |
blackburn | Ram108: well that's exactly *not* how the brain works :) | 18:19 |
blackburn | too simplified.. | 18:19 |
shogun-buildbot | build #273 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/273 | 18:20 |
Ram108 | oh thats true :) this is a very crude representation | 18:20 |
blackburn | NN is very useful in OCR and visual pattern recognition though | 18:20 |
blackburn | Ram108: sorry I can't remember if you did something with MNIST dataset? | 18:21 |
Ram108 | no that was the other guy | 18:21 |
blackburn | heh okay | 18:21 |
blackburn | then you could check MNIST page | 18:21 |
blackburn | http://yann.lecun.com/exdb/mnist/ | 18:21 |
Ram108 | sure :) looking forward to it :) | 18:22 |
blackburn | best results are computed with convolution neural nets commitee | 18:22 |
Ram108 | god this is huge! | 18:22 |
blackburn | with error of 0.27 | 18:22 |
blackburn | just check results to compare methods | 18:22 |
Ram108 | hmmm i ll try it out :) definitely with shoguns svm methods and the code which i am working on currently lol :P | 18:23 |
blackburn | but in other tasks NNs could act worse.. | 18:24 |
Ram108 | thanks a lot again........ well initially i was overwhelmed but now i seem to have a quaint idea of hw it works :) | 18:24 |
blackburn | nice | 18:24 |
Ram108 | nn is slow and more often than necessary the code keeps running out of memory | 18:24 |
Ram108 | thats 1 thing i have to fix | 18:25 |
Ram108 | bye the way do u mind u could tell me abt urslf? | 18:25 |
blackburn | what exactly you want to know? :) | 18:25 |
blackburn | I guess we are of same age, I'm a student of Samara State Aerospace University in Samara, Russia | 18:26 |
Ram108 | oh hmmm :) | 18:26 |
Ram108 | nice :) | 18:27 |
blackburn | I'm also a software engineer @ netcracker | 18:27 |
Ram108 | aerospace univ? | 18:27 |
Ram108 | cool :) | 18:27 |
blackburn | last year I was a GSoC student, I did impelement 12 dimensionality reduction methods | 18:27 |
blackburn | and am still interested in nonlinear dimensionality reduction :) | 18:28 |
Ram108 | yeah u told me that........ any chance i could try getting in? just asked | 18:28 |
blackburn | as GSoC student? | 18:28 |
Ram108 | yeah | 18:29 |
blackburn | well if we get accepted | 18:29 |
blackburn | and if GSoC 2012 is announced :) | 18:29 |
blackburn | you definitely should try to some organization | 18:29 |
Ram108 | well mostly shogun ll be right? | 18:29 |
blackburn | what do you mean? | 18:30 |
Ram108 | as in ML is one of the frontier fields | 18:30 |
Ram108 | so why wuldnt it? | 18:30 |
blackburn | I meant you can choose one organization you would like to work with | 18:31 |
Ram108 | i guess am no1 to comment upon this....... | 18:31 |
Ram108 | well i really want to learn this out | 18:31 |
Ram108 | the thing is i was into robotics for so long..... and i got saturated with it. and i want to impart intelligence to most of the machines i am building | 18:32 |
Ram108 | lol i feel GOD wen i do that lol :P | 18:33 |
blackburn | we would have to choose some students, I can't promise anything but we would prefer students that know how to work with shogun | 18:33 |
Ram108 | hmmm :) i see | 18:33 |
blackburn | I guess that's one of the reasons you started to learn shogun :) | 18:34 |
Ram108 | well u mean equipping my gadgets with intelligence? | 18:34 |
Ram108 | yeah | 18:35 |
blackburn | okay I have to study optics a little heh | 18:35 |
blackburn | will have exam on saturday | 18:36 |
Ram108 | oh thats a drag.... go on :) | 18:36 |
blackburn | no idea why I have to learn this | 18:36 |
blackburn | :D | 18:36 |
Ram108 | dads a physicist i kno why | 18:37 |
Ram108 | lol :P | 18:37 |
blackburn | btw I don't know anything about 'aerospace' :D | 18:37 |
Ram108 | yeah i know i feel out of the soup as well........ took the wrong branch......... found myself liking robotics more than communication engineering | 18:38 |
Ram108 | in the soup* | 18:39 |
Ram108 | sorry | 18:39 |
Ram108 | actually i was gona type fish out of water | 18:39 |
blackburn | well I'm just major in mathematics and nothing with neither aero nor space ;) | 18:39 |
blackburn | I've never been studying anything related | 18:40 |
Ram108 | math..... thats a pain......... never liked math for itself... fells like it just loses touch with reality | 18:40 |
blackburn | can't agree, there are a lot of branches | 18:41 |
Ram108 | i did make a plane wen i was in my 6th grade though....... i have it with me wana see? | 18:41 |
blackburn | well even machine learning is math 99% :) | 18:41 |
blackburn | plane? | 18:41 |
Ram108 | hate tensors......... and statistics | 18:42 |
Ram108 | yeah agreed but i can really see the touch with the real world | 18:42 |
blackburn | oh tensors :) I guess you had a lot of mechanics | 18:42 |
Ram108 | yeah i did......... fluid mechanics, Bernoullis theorems etc etc | 18:43 |
Ram108 | was way over my head | 18:43 |
Ram108 | just was not cut out for that kind of thing | 18:43 |
Ram108 | http://orange.biolab.si/ | 18:44 |
Ram108 | pretty cool like u guys :) | 18:44 |
blackburn | sure, orange has good gui and is easier | 18:46 |
Ram108 | yeah i saw that | 18:47 |
Ram108 | :) | 18:47 |
blackburn | but we have a lot of features they has not | 18:47 |
Ram108 | thats y i chose shogun over that | 18:47 |
blackburn | scikits-learn has a lot of nice features | 18:47 |
blackburn | I like it pretty much | 18:47 |
Ram108 | lol saw that too :P | 18:48 |
Ram108 | was just wondering | 18:52 |
Ram108 | why cant the algorithms be classified | 18:52 |
Ram108 | based on parameters like datatypes time required and processing requirements | 18:53 |
blackburn | which algos? | 18:54 |
Ram108 | then probably the end user wont have to kno which algorithm to use...... all he has to specify is the parameters and their limits...... | 18:54 |
Ram108 | could act like an AI engine | 18:54 |
blackburn | well nobody knows how good some algorithm will work with some data | 18:55 |
Ram108 | algorithms like SVM NN etc | 18:55 |
Ram108 | oh erm is there no way of predicting that? just curious | 18:55 |
blackburn | well experts do know a little more, but generally you would have to test whether it works well or not | 18:56 |
Ram108 | hmmm so the possibility of automating that is slim right? | 18:57 |
Ram108 | okay :) thanks for ur time :) | 18:57 |
blackburn | there are some estimates of probabilities but they are too much general | 18:57 |
blackburn | it is being researched in COLT | 18:57 |
blackburn | Computation Learning Theory | 18:58 |
Ram108 | oh hmmm | 18:58 |
Ram108 | oh i see :) i ll google that up | 18:58 |
blackburn | I don't know much about COLT, too complex for practice | 18:58 |
Ram108 | lol | 18:59 |
blackburn | anyway choosing right methods is not the way how AI should work | 19:00 |
Ram108 | hw else? | 19:01 |
blackburn | should be unified.. | 19:01 |
blackburn | something of interest is a deep learning | 19:02 |
Ram108 | well cant we use unsupervised learning methods to classify the given data? | 19:02 |
Ram108 | thats an idea..... | 19:02 |
blackburn | unsupervised? well we could but results would be worse | 19:03 |
Ram108 | oh hmmm......... i have to try that someday | 19:04 |
blackburn | well some tasks are supervised by nature | 19:04 |
blackburn | if you want to recognize road signs you know classes, etc | 19:05 |
blackburn | no need to do that unsupervised | 19:05 |
Ram108 | hmmm | 19:05 |
Ram108 | well for now am working on unmanned flights and vehicles | 19:06 |
Ram108 | :) | 19:06 |
blackburn | unfortunately you have not many choices | 19:07 |
Ram108 | as in? | 19:07 |
blackburn | I mean it would be strange to use KNN or Ridge Regression to ride a vehicle :) | 19:07 |
Ram108 | could u be a little more specific pls? | 19:07 |
Ram108 | hmmm lol well gona use NN | 19:07 |
blackburn | usually robots are powered by NN | 19:08 |
Ram108 | yeah lol :P | 19:08 |
Ram108 | read abt em | 19:08 |
blackburn | cause nature of riding a vehicle is not really formalized :) | 19:08 |
Ram108 | thanks :) | 19:08 |
Ram108 | i ll post a youtube video of it soon........ lol but i have to go through a lot of logistics and paper filling to procure the essential funding for that | 19:09 |
blackburn | I see | 19:09 |
Ram108 | i ll get started on the hardware soon :) | 19:10 |
Ram108 | am pretty good with digital circuits | 19:10 |
Ram108 | :) guess that wont be a problem | 19:10 |
blackburn | hah I don't give a fuck about it | 19:10 |
Ram108 | lol | 19:10 |
blackburn | U = IR that's the only thing I know | 19:10 |
Ram108 | hmmm well microcontrollers are not that bad | 19:11 |
Ram108 | besides i firmly believe innovation lies in the merging point between fields.... so lol determined to learn this out | 19:12 |
blackburn | sure | 19:12 |
Ram108 | lol :) | 19:12 |
Ram108 | guess i gotta crash for the night.......... really tired....... | 19:13 |
Ram108 | thanks for ur time :) | 19:13 |
blackburn | ah you have different timezone :) | 19:13 |
blackburn | 1-13? | 19:13 |
blackburn | what time is it? | 19:13 |
Ram108 | yeah apparantly i do lol :P | 19:13 |
Ram108 | its just 11 45 but i woke up at 3 am this morn | 19:14 |
blackburn | 11-45? | 19:14 |
Ram108 | and i had a badminton match which lasted for 3 hrs....... | 19:15 |
blackburn | wow you +1.30 from me? | 19:15 |
Ram108 | +1.30? | 19:15 |
blackburn | yeah I guess so | 19:15 |
blackburn | UTC + 5.30? | 19:15 |
Ram108 | lol yeah :) | 19:15 |
Ram108 | hmmm the time there is 7 15 pm right? | 19:16 |
blackburn | no, I'm in UTC+4 | 19:16 |
blackburn | Moscow timezone | 19:16 |
Ram108 | i can barely open my eyes........ lol :P | 19:16 |
blackburn | then go sleep :) | 19:16 |
Ram108 | good night :) | 19:17 |
blackburn | good night | 19:17 |
blackburn | see you | 19:17 |
Ram108 | u too thanks for ur time :) | 19:17 |
Ram108 | bye :) | 19:17 |
blackburn | bye | 19:17 |
-!- Ram108 [~amma@14.96.88.10] has quit [Quit: Ex-Chat] | 19:17 | |
blackburn | sonney2k_: can one-versus-all ROC be considered as good performance measure? | 20:20 |
CIA-27 | shogun: Soeren Sonnenburg master * rc9e6d08 / src/interfaces/modular/SGBase.i : pickle_ascii should be 1 if pickle version is 0 - http://git.io/VnSEpQ | 20:36 |
blackburn | sonney2k_: did you receive my mail? :) | 21:54 |
sonney2k_ | blackburn, yes | 22:22 |
-!- sonney2k_ is now known as sonney2k | 22:22 | |
blackburn | any thought? | 22:22 |
blackburn | s* | 22:22 |
-!- mode/#shogun [+o sonney2k] by ChanServ | 22:23 | |
@sonney2k | I found some potential mentors... but was/am busy with other stuff | 22:24 |
blackburn | ah I see | 22:24 |
@sonney2k | but yes we need to come up with some more polished list | 22:24 |
blackburn | sonney2k: I'm unavailable for thinking till 21st :) | 22:25 |
blackburn | but I hope to come up with something to the end of January | 22:25 |
@sonney2k | Well then after 21st. it looks like I won't progress before that either... | 22:25 |
@sonney2k | at least the buildbot is happy again :) | 22:26 |
@sonney2k | blackburn, what did you serialize that is *that big* | 22:26 |
@sonney2k | was that with pickle? | 22:26 |
blackburn | sonney2k: it looks like you are going to be busier and busier | 22:26 |
blackburn | sonney2k: yes, pickle | 22:27 |
blackburn | sonney2k: LibSVM with 39202 train vectors and ~12k test ones of dim 81 | 22:27 |
@sonney2k | well kids that don't sleep kill any shogun time... | 22:27 |
blackburn | *multiclass | 22:28 |
@sonney2k | yeah but that should just be 30M | 22:28 |
@sonney2k | how many classes? | 22:28 |
@sonney2k | one-vs-rest? | 22:28 |
blackburn | 43, OVO | 22:28 |
@sonney2k | I suspect features are saved multiple times | 22:29 |
blackburn | I guess so | 22:29 |
blackburn | sonney2k: but libsvm do not support OvA :) | 22:29 |
@sonney2k | svmlight does :) | 22:29 |
blackburn | sonney2k: does svmlight support multiclass? | 22:29 |
@sonney2k | ahh no you need to do it manually of course | 22:30 |
blackburn | oh no I don't like it | 22:30 |
blackburn | :D | 22:30 |
@sonney2k | but yes all that multiclass stuff would be a great gsoc project | 22:30 |
blackburn | sonney2k: do you think it would be cool to have MulticlassMachine (ECOC,OvA,OvO)? | 22:30 |
blackburn | based on linear machines | 22:30 |
@sonney2k | such that any classifier can be turned into * | 22:30 |
@sonney2k | heh | 22:31 |
@sonney2k | exactly | 22:31 |
blackburn | I would say it is not a GSoC project | 22:31 |
blackburn | but week of work.. | 22:31 |
@sonney2k | there is this advanced stuff from nips last year | 22:31 |
blackburn | i'm on vacation starting from 25, Jan to 5, Feb | 22:31 |
@sonney2k | that would certainly make it so | 22:31 |
blackburn | will try to start | 22:31 |
@sonney2k | would be cool | 22:32 |
blackburn | I need it just now, so I would have to implement | 22:32 |
@sonney2k | btw, we don't even have some linear regression methods in shogun | 22:32 |
blackburn | at least simplest concept | 22:32 |
blackburn | sonney2k: we have bad feature list now | 22:32 |
@sonney2k | no LS-regression | 22:32 |
blackburn | I'm getting upset looking on scikit-learn methods | 22:32 |
blackburn | and comparing with us | 22:32 |
@sonney2k | (which I need now) | 22:32 |
@sonney2k | they are much broader sure | 22:33 |
blackburn | they have one billion authors | 22:33 |
blackburn | authors list in jmlr paper do not fit in one page! | 22:33 |
blackburn | and two billion contributors :) | 22:33 |
@sonney2k | yeah but we had the same problem w shogun | 22:34 |
blackburn | sonney2k: aha but where are these guys? :) | 22:34 |
@sonney2k | true all not really active right now. I think shogun does what they needed now and so they are just users | 22:35 |
@sonney2k | time to relax a bit | 22:36 |
blackburn | sonney2k: sleep? :) | 22:36 |
-!- karlnapf [~heiko@host86-174-150-23.range86-174.btcentralplus.com] has joined #shogun | 23:01 | |
-!- karlnapf [~heiko@host86-174-150-23.range86-174.btcentralplus.com] has quit [Ping timeout: 240 seconds] | 23:16 | |
-!- karlnapf [~heiko@host86-174-150-23.range86-174.btcentralplus.com] has joined #shogun | 23:28 | |
@sonney2k | sleep! | 23:49 |
blackburn | sonney2k: http://dl.dropbox.com/u/10139213/shogun/output.pdf that's how my time measure script does output :) | 23:53 |
blackburn | see you | 23:54 |
-!- blackburn [~qdrgsm@188.168.14.65] has quit [Quit: Leaving.] | 23:57 | |
--- Log closed Thu Jan 19 00:00:19 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!