--- Log opened Thu Dec 13 00:00:38 2012 | ||
shogun-buildbot | build #173 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/173 | 03:06 |
---|---|---|
shogun-buildbot | build #216 of nightly_default is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/216 | 03:12 |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 03:12 | |
-!- agas [~admin@185.6.154.4] has joined #shogun | 03:35 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Disconnected by services] | 03:35 | |
-!- agas is now known as Hausas | 03:35 | |
-!- Hausas [~admin@185.6.154.4] has quit [Changing host] | 03:35 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 03:35 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 246 seconds] | 04:47 | |
-!- Hausas [~admin@185.6.154.4] has joined #shogun | 04:52 | |
-!- Hausas [~admin@185.6.154.4] has quit [Changing host] | 04:52 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 04:52 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 248 seconds] | 06:34 | |
-!- Hausas [~admin@185.6.154.4] has joined #shogun | 06:36 | |
-!- Hausas [~admin@185.6.154.4] has quit [Changing host] | 06:36 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 06:36 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 250 seconds] | 08:01 | |
-!- n4nd0 [~fjig@grey-02.csc.kth.se] has joined #shogun | 08:03 | |
-!- blackburn [~blackburn@188.168.128.156] has quit [Quit: Leaving.] | 08:39 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 09:38 | |
-!- n4nd0 [~fjig@grey-02.csc.kth.se] has quit [Quit: leaving] | 09:53 | |
-!- zxtx [~zv@82.207.176.18] has quit [Ping timeout: 252 seconds] | 10:08 | |
-!- zxtx [~zv@46.231.183.163] has joined #shogun | 10:59 | |
-!- zxtx [~zv@46.231.183.163] has quit [Ping timeout: 245 seconds] | 11:04 | |
-!- zxtx [~zv@46.231.183.163] has joined #shogun | 11:10 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 260 seconds] | 12:07 | |
-!- zxtx [~zv@46.231.183.163] has quit [Ping timeout: 245 seconds] | 12:10 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 12:12 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 12:51 | |
wiking | sonne|work: here? | 13:06 |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 245 seconds] | 13:06 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 13:06 | |
sonne|work | wiking: ye | 13:22 |
sonne|work | s | 13:22 |
wiking | sonne|work: i wonder what i was about to ask | 13:26 |
wiking | ah yeah one thing | 13:26 |
wiking | how's our release schedule | 13:26 |
wiking | and 2) how do i apply for write permission into shoung | 13:26 |
wiking | shogun repo | 13:26 |
wiking | :> | 13:26 |
sonne|work | wiking: we have no schedule... if sb wants a release he should say so and we should then try to get things more stable / fix issues that appear (not that I know any...) | 13:27 |
sonne|work | wiking: you get write permissions when you continue to stay active like you are now :) | 13:27 |
wiking | sonne|work: but the question is when | 13:28 |
wiking | i'm like almost a year now on the row | 13:28 |
wiking | it'll be in february | 13:28 |
wiking | that i've submitted my 1st patch :) | 13:28 |
sonne|work | ohh indeed | 13:29 |
sonne|work | then *very* sooon | 13:29 |
wiking | hahahah | 13:29 |
wiking | lolface | 13:29 |
wiking | ok | 13:29 |
wiking | i think we should do a release still this year | 13:29 |
wiking | but there's certain things needs to be fix i think | 13:30 |
sonne|work | nothing against it | 13:30 |
sonne|work | please start to fix things you come across then | 13:30 |
sonne|work | I am not so sure what needs fixing though | 13:30 |
sonne|work | I would have liked to attack sparse/stringfeatures to work with SG* datatypes | 13:30 |
sonne|work | but hey that can also wait for another release | 13:31 |
wiking | i need to fix cc | 13:31 |
wiking | with dynamicarray | 13:31 |
wiking | i know that | 13:31 |
sonne|work | cc ? | 13:31 |
wiking | ccsosvm | 13:31 |
sonne|work | we don't have an example for ccsosvm right? | 13:31 |
wiking | no not at all atm | 13:32 |
wiking | i mean i've added it for | 13:32 |
wiking | so_multiclass.cpp | 13:32 |
wiking | sonne|work: hehe yeah | 13:40 |
wiking | you've closed this PR | 13:40 |
wiking | https://github.com/shogun-toolbox/shogun/pull/839 | 13:41 |
sonne|work | sure | 13:41 |
wiking | there was the addition to the so_multiclass example | 13:41 |
wiking | i'll make the changes though | 13:41 |
sonne|work | then fix it and reopen | 13:41 |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has joined #shogun | 13:45 | |
yoo | hi all | 13:45 |
wiking | hey | 13:45 |
sonne|work | so wiking if you want a relase start to test & fix things and some plan what should be stable by then and in NEWS would be nice too | 13:53 |
wiking | ok | 13:56 |
wiking | i can start doing that next week | 13:56 |
wiking | this week is very busy for me estill | 13:56 |
sonne|work | np | 13:57 |
sonne|work | yoo please ask in the channel | 14:10 |
yoo | I can read in the code that OvRStrategy is used. Is CGMNPSVM OneVsRest or OneInAll ? | 14:13 |
n4nd0 | yoo: it should be one vs. rest as it says in the header and because they are using one vs rest in the implementation file | 14:20 |
n4nd0 | yoo: why do you wonder if you saw OvR was used? | 14:21 |
sonne|work | yoo, n4n0 - it is true multiclass | 14:21 |
yoo | because it says "true multiclass svm fashion" | 14:21 |
sonne|work | so all hyperplanes are learned simultaneously | 14:22 |
yoo | and true multiclass mean AllInOne , | 14:22 |
yoo | ? | 14:22 |
n4nd0 | yoo: yeah ok, that is because in shogun you can do multiclass based on binary classifiers | 14:22 |
sonne|work | but when applying you have nr classes many classifiers | 14:22 |
n4nd0 | yoo: but there are also some classifiers that are natively multiclass, like this one | 14:23 |
yoo | allright because I used MKLMulticlass that use CGMNPSVM then for N classes I get global optimiaztion and one set of kernels weights | 14:23 |
yoo | what if I want differents set of kernels weights for each class ? | 14:24 |
yoo | is there a framework for that in shogun ? | 14:24 |
n4nd0 | yoo: I've never used it but it looks like MKLOneClass could make sense | 14:25 |
yoo | Sure. | 14:26 |
yoo | And what Multiclass machine resolve the OvR multiclass problem ? | 14:27 |
n4nd0 | yoo: CGMNPSVM? :) | 14:28 |
yoo | I mean c(x) = argmax(f_i) and not c(x) = argmax(w_i x + b_i) as CGMNPSVM | 14:28 |
n4nd0 | ok | 14:28 |
yoo | or maybe I misunderstood smthg | 14:28 |
n4nd0 | I didn't really get the question to be honest | 14:29 |
n4nd0 | so an SVM intuitively puts a line (hyperplane) somewhere in the space | 14:30 |
n4nd0 | that can only separate two things | 14:30 |
n4nd0 | to do multiclass, you have to put several lines or use several SVMs | 14:30 |
n4nd0 | depending on how this lines are put you get one vs. rest or one vs. one | 14:31 |
n4nd0 | yoo: when you were saying one in all I guess you meant one vs. all? | 14:31 |
n4nd0 | in that case, one vs. all and one vs. rest is the same concept. It just makes more sense to say one vs. rest, it is more semantically correct I think | 14:32 |
yoo | OvR : N binary classfiers are trained to separate one class from the rest and winning class is the one that corresponds to the SVM with highest output value c(x) = argmax(f_i) | 14:34 |
yoo | All in One : N two class rules the ith function separates training vectors of the class i from the other vectors | 14:35 |
yoo | N decisions but all are obtained by solving one problem | 14:35 |
n4nd0 | aham I see | 14:35 |
yoo | All in One is more a generalization of the binary classification algorithm | 14:37 |
yoo | then it is called "true multiclass" | 14:37 |
n4nd0 | yoo: I am not sure about this, but I think that CGMNPSVM solve OneVsRest, since they are using that MulticlassStrategy | 14:37 |
yoo | that's I was saying ^^ | 14:37 |
n4nd0 | hehe ok :) | 14:38 |
yoo | just wanteed to be sure what it is exaclty | 14:38 |
n4nd0 | let me check a moment just in case | 14:38 |
yoo | because you can read in the header "true multiclass" | 14:38 |
yoo | The question is: the N binary machine are learned independently or not ? | 14:39 |
n4nd0 | yoo: I think that they are not learnt independently | 14:41 |
yoo | then it is not OvR | 14:41 |
yoo | what make you say that ? | 14:41 |
n4nd0 | I ignored this distinction between OvR and all in one | 14:42 |
n4nd0 | I didn't know it had that name | 14:42 |
sonne|work | n4nd0: no there is a true multiclass svm | 14:43 |
sonne|work | n4nd0: that simultaneously learns all hyperplanes - so in the end you get a OvR classifier but hyperplanes are not based on learning serveral OvR binary SVMs but learnt in one go | 14:43 |
yoo | allright, then if you wanted to resolve a multiclass pb with OvR strategy what shogun framework will you use ? | 14:43 |
n4nd0 | sonne|work: I see, but even if learnt silmultaneously one could do that with OvR or OvOne I gues, right? | 14:44 |
yoo | I mean I want to optimize mkl weights for each class and MKLMulticlass use CGMNPSVM .. | 14:44 |
sonne|work | n4nd0: I don't understand what you are asking | 14:45 |
n4nd0 | sonne|work: I understand the difference between learning all hyperplanes at the same time or not, we refer to this as TRUE multiclass when all of them are learnt together | 14:46 |
n4nd0 | sonne|work: ok? | 14:46 |
yoo | ok | 14:46 |
n4nd0 | sonne|work: but in true multiclass I suppose that one can do it either using an OvR strategy or a OvO strategy | 14:47 |
yoo | no I dont think so | 14:47 |
sonne|work | n4nd0: true multiclass is what I mean with learning all hyperplanes together | 14:49 |
n4nd0 | sonne|work: ok, then we are speaking the same language :) | 14:50 |
sonne|work | n4nd0: OvR OvO learning strategies are just for binary classifiers | 14:50 |
n4nd0 | yoo: anyway, you can use any of the binary classifiers in shogun to build a multiclass classifier using one vs. rest | 14:51 |
yoo | sonne|work: then if I want to do MKL and optimize mkl weights for each class ? | 14:51 |
n4nd0 | yoo: see the KernelMulticlassMachine | 14:52 |
yoo | n4nd0: ok | 14:52 |
n4nd0 | sonne|work: I understand now, thanks! although I think that doesn't why is MulticlassOneVsRestStrategy used in GMNPSVM | 14:53 |
n4nd0 | doesn't explain* | 14:53 |
n4nd0 | I am guessing that it appears to follow the MulticlassSVM interface | 14:54 |
sonne|work | n4nd0: well they use the same code when applying the classifier... argmax f_c(x) | 14:54 |
n4nd0 | got it | 14:57 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 244 seconds] | 15:26 | |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has quit [Quit: Page closed] | 15:43 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Read error: Connection reset by peer] | 15:56 | |
-!- Hausas [~admin@185.6.154.4] has joined #shogun | 15:57 | |
-!- Hausas [~admin@185.6.154.4] has quit [Changing host] | 15:57 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 15:57 | |
-!- zxtx [~zv@n80-237-234-136.cnet.hosteurope.de] has joined #shogun | 16:13 | |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has joined #shogun | 16:24 | |
yoo | hi again | 16:25 |
yoo | sonne|work: I tried the following CKernelMulticlassMachine* mkl = new CKernelMulticlassMachine(new CMulticlassOneVsRestStrategy(), cker, new CMKLOneClass(), train_labels); but it raises the error: Interleaved MKL optimization is currently only supported with SVMlight | 16:27 |
yoo | cker for CCombinedKernels | 16:27 |
sonne|work | well you should not use oneclass | 16:28 |
sonne|work | but twoclass | 16:28 |
yoo | CMKL instead of COneClass ? | 16:29 |
-!- zxtx [~zv@n80-237-234-136.cnet.hosteurope.de] has quit [Ping timeout: 264 seconds] | 16:51 | |
-!- zxtx [~zv@82.207.176.18] has joined #shogun | 17:15 | |
-!- heiko [~heiko@dhcp-180-207.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:25 | |
-!- blackburn [~blackburn@83.234.169.110] has joined #shogun | 17:53 | |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has quit [Quit: Page closed] | 18:32 | |
-shogungit:#shogun- [shogun] karlnapf opened pull request #845: some clean ups towars kernel selection framework for MMD (master...master) https://github.com/shogun-toolbox/shogun/pull/845 | 18:54 | |
-shogungit:#shogun- [shogun] karlnapf pushed 5 new commits to master: https://github.com/shogun-toolbox/shogun/compare/bbb2be6fad1b...fa3566b5db4b | 18:54 | |
-shogungit:#shogun- shogun/master 7303295 Heiko Strathmann: removed old kernel selection examples due to soon to be introduced new framework | 18:54 | |
-shogungit:#shogun- shogun/master 33f4eb1 Heiko Strathmann: documentation | 18:54 | |
-shogungit:#shogun- shogun/master 836d10a Heiko Strathmann: initialise optimisation parameters | 18:54 | |
-!- heiko [~heiko@dhcp-180-207.internal.eduroam.ucl.ac.uk] has left #shogun [] | 19:07 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 19:19 | |
shogun-buildbot | build #720 of deb3 - modular_interfaces is complete: Failure [failed compile octave_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/720 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:39 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 265 seconds] | 20:57 | |
-!- blackburn [~blackburn@83.234.169.110] has left #shogun [] | 21:33 | |
-!- Netsplit *.net <-> *.split quits: naywhayare | 21:47 | |
-!- Netsplit over, joins: naywhayare | 21:48 | |
-!- Hausas [~admin@unaffiliated/hausas] has quit [Ping timeout: 245 seconds] | 23:07 | |
-!- Hausas [~admin@185.6.154.4] has joined #shogun | 23:25 | |
-!- Hausas [~admin@185.6.154.4] has quit [Changing host] | 23:25 | |
-!- Hausas [~admin@unaffiliated/hausas] has joined #shogun | 23:25 | |
--- Log closed Fri Dec 14 00:00:38 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!