--- Log opened Tue Mar 27 00:00:19 2012 | ||
@sonney2k | blackburn, yeah I know (you switch topics every 5 minutes so it must be you :) | 00:00 |
---|---|---|
blackburn | sonney2k: well, sorry :) | 00:01 |
blackburn | http://latex.codecogs.com/gif.latex?L(w,\xi)%20=%20\frac{1}{2}%20\sum_m%20\|w_m\|^2%20+%20C%20\sum_i%20\xi_i%20+%20\sum_m%20\sum_i%20\alpha_i^m%20(\langle%20w_{y_i},x_i\rangle%20-%20\langle%20w_m,x_i%20\rangle%20-%201%20-%20\delta_{y_i,m}%20-%20\xi_i\rangle)%20+%20\sum_i%20\beta_i%20\xi_i | 00:01 |
@sonney2k | gsomix, submit what you have then I will try to have a look at the remaining examples tomorrow | 00:02 |
@sonney2k | gsomix, always note that you are not alone - when you get stuck ask! | 00:02 |
blackburn | sonney2k: is d/d(w_m) of \sum_m ||w_m||^2 actually equal to 2 \sum_m ||w_m||? | 00:03 |
blackburn | a bit tricky | 00:03 |
@sonney2k | blackburn, no b in that formula? | 00:04 |
blackburn | sonney2k: no, C-S makes no use of B | 00:04 |
blackburn | sonney2k: btw did you know that there are really strong relations between crammer-singer and SO svm? | 00:05 |
@sonney2k | blackburn, yes | 00:05 |
blackburn | I was happy to know it | 00:05 |
@sonney2k | same idea | 00:05 |
@sonney2k | or better SO extends this idea... | 00:05 |
gsomix | sonney2k, thx. of course, I'm not alone. blackburn lives just on one floor above. :) | 00:05 |
n4nd0 | blackburn: is it what we discussed yesterday a bit? | 00:05 |
blackburn | n4nd0: yes you made me think about it | 00:06 |
@sonney2k | gsomix, he doesn't need much sleep so feel free to wake him up at any time ;-) | 00:06 |
shogun-buildbot | build #390 of ruby_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/390 | 00:06 |
blackburn | LOL | 00:06 |
@sonney2k | blackburn, 2 || w_m|| you mean | 00:06 |
blackburn | actually it is rather difficult to catch me there | 00:06 |
blackburn | :D | 00:06 |
blackburn | sonney2k: where is sum? | 00:07 |
@sonney2k | derivative? | 00:07 |
blackburn | I am confused with derivative of sum | 00:07 |
blackburn | :D | 00:07 |
@sonney2k | shouldn't it be zero for all w_i i!=m | 00:07 |
blackburn | ah | 00:07 |
blackburn | hahaha | 00:07 |
blackburn | haha | 00:07 |
@sonney2k | maybe you choose some other sub index | 00:08 |
blackburn | I got it | 00:08 |
blackburn | sonney2k: remember I told you about ECOC SVM? | 00:08 |
@sonney2k | ? | 00:09 |
blackburn | sonney2k: I think it could be a great idea | 00:09 |
blackburn | I asked vojtech and he agreed | 00:09 |
blackburn | idea is to extend any MC formulation | 00:09 |
blackburn | in means of ECOC | 00:09 |
@sonney2k | now I don't know what you are talking about | 00:09 |
blackburn | yes, yes, I'll describe | 00:10 |
blackburn | sonney2k: recall s.t. in structural svm | 00:10 |
@sonney2k | we just received the first student application! | 00:10 |
@sonney2k | (officially submitted via gsoc melange) | 00:11 |
blackburn | F is the same as in C-S, but \delta(y_i,y) is code distance (say hamming) | 00:11 |
blackburn | <w_yi,xi> - <w_m,x_i> >= hamming distance - \xi_i | 00:11 |
blackburn | sonney2k: crazy, is it someone you know already? | 00:12 |
n4nd0 | harshit said he was planning to start quite early with it ... | 00:12 |
@sonney2k | blackburn, no | 00:13 |
@sonney2k | you all should - this way you can get feedback | 00:13 |
n4nd0 | but not directly to submit via melange right? | 00:14 |
n4nd0 | sonney2k: maybe mail it to you or something? | 00:14 |
blackburn | n4nd0: why not? | 00:14 |
@sonney2k | n4nd0, you can alwasy update until the deadline | 00:14 |
blackburn | yeah | 00:14 |
n4nd0 | ah ok, forgot about that! | 00:15 |
n4nd0 | then I think it is the most comfortable thing for all | 00:15 |
blackburn | sonney2k: last thing before I leave - is my idea clear for you? | 00:15 |
@sonney2k | and we (mentors) are being spammed like hell too :) | 00:15 |
@sonney2k | blackburn, yes - I have no feeling though how it will change the learned result and if it will be useful | 00:17 |
blackburn | sonney2k: no serious change | 00:17 |
blackburn | but something similar to GMNPSVM | 00:17 |
blackburn | scaling mostly | 00:17 |
blackburn | while ECOC is more disbalanced than OvR it may impact | 00:18 |
blackburn | sonney2k: I gave this idea to my gf to derive dual task and stucked myself :D | 00:18 |
blackburn | that's why I asked | 00:19 |
@sonney2k | poor gf :) | 00:19 |
@sonney2k | weird way of showing appreciation | 00:19 |
blackburn | hahahah | 00:19 |
n4nd0 | :D | 00:19 |
blackburn | well she has to come up with some great bachelor work, right? ;) | 00:20 |
blackburn | sonney2k: my paper on sign recognition is almost ready btw http://dl.dropbox.com/u/10139213/rsr_paper.pdf | 00:21 |
blackburn | take a look on that funny characters lol | 00:22 |
@sonney2k | blackburn, ok so what you would hope for is that e.g. hamming code is more robust than ovr code above? | 00:22 |
blackburn | sonney2k: yes | 00:22 |
blackburn | sonney2k: first shogun mentioning paper in russian ever | 00:22 |
@sonney2k | blackburn, it is so funny that this is pure russian :) | 00:23 |
@sonney2k | ahh english abstract :) | 00:23 |
@sonney2k | let me try to decipher how much I understand from the russian one :) | 00:23 |
blackburn | sonney2k: I bet not much ;) | 00:24 |
@sonney2k | haha maschina opornich vectorow :) | 00:25 |
blackburn | yes that is svm | 00:25 |
@sonney2k | never heard of anyone translating 'SVM' to german | 00:25 |
@sonney2k | it would be Stuetzvektormaschine | 00:25 |
blackburn | hahahah | 00:25 |
@sonney2k | which sounds incredibly funny | 00:25 |
@sonney2k | blackburn, why do you publish in russian? | 00:26 |
blackburn | sonney2k: cause it is a useless crap | 00:26 |
blackburn | :D | 00:26 |
n4nd0 | when I came last year to Spain after studying some machine learning at KTH I herad a professor talking about MVS | 00:26 |
@sonney2k | n4nd0, MVS? | 00:27 |
n4nd0 | I think he said it like three times until I realized it was the short for SVM in Spanish :) | 00:27 |
@sonney2k | SVMs? | 00:27 |
n4nd0 | M?quinas de Vectores Soporte | 00:27 |
blackburn | ohohhoh | 00:27 |
blackburn | oloooooool | 00:27 |
@sonney2k | bahh | 00:27 |
n4nd0 | I really didn't know what they were talking about when the said it | 00:27 |
@sonney2k | I guess wikipedia has it all | 00:27 |
blackburn | sonney2k: that's the difficult question why I publish in russian | 00:28 |
@sonney2k | https://es.wikipedia.org/wiki/M%C3%A1quinas_de_vectores_de_soporte | 00:28 |
@sonney2k | https://ru.wikipedia.org/wiki/%D0%9C%D0%B5%D1%82%D0%BE%D0%B4_%D0%BE%D0%BF%D0%BE%D1%80%D0%BD%D1%8B%D1%85_%D0%B2%D0%B5%D0%BA%D1%82%D0%BE%D1%80%D0%BE%D0%B2 | 00:28 |
@sonney2k | and japanese : https://ja.wikipedia.org/wiki/%E3%82%B5%E3%83%9D%E3%83%BC%E3%83%88%E3%83%99%E3%82%AF%E3%82%BF%E3%83%BC%E3%83%9E%E3%82%B7%E3%83%B3 | 00:29 |
-!- av3ngr [av3ngr@nat/redhat/x-dhyqmsdozibizipk] has joined #shogun | 00:29 | |
@sonney2k | ??????????? | 00:29 |
blackburn | sonney2k: chervonenkis used to call SVM as it is in wiki | 00:29 |
blackburn | V-C one you know | 00:30 |
@sonney2k | well I guess russians have a good reason | 00:30 |
n4nd0 | hahaha it is even in Swedish | 00:30 |
@sonney2k | :) | 00:30 |
n4nd0 | I bet the people in the department here have not heard it either | 00:30 |
@sonney2k | anyway seriously bed time for me now | 00:30 |
@sonney2k | cu all | 00:30 |
blackburn | see you | 00:30 |
n4nd0 | good night | 00:31 |
blackburn | for me too probably | 00:31 |
blackburn | sonney2k: I guess main answer is this paper just supports my bachelor's work and contains nothing really new | 00:31 |
blackburn | so it would fit only to some paid journals or local russian journal like this one | 00:32 |
-!- PhilTillet [~Philippe@38.124.160.27] has joined #shogun | 00:32 | |
blackburn | n4nd0: see you tomorrow | 00:32 |
n4nd0 | blackburn: bye! | 00:33 |
n4nd0 | good night | 00:33 |
blackburn | gsomix: hey you | 00:33 |
blackburn | :D | 00:33 |
blackburn | good night | 00:33 |
gsomix | blackburn, "Hey you" is the Pink Floyd's song. | 00:33 |
gsomix | blackburn, good, ok | 00:33 |
blackburn | really? I am not a big fan of pink floyd | 00:33 |
blackburn | I used to listen to stoopid music :) | 00:34 |
PhilTillet | Hey :) | 00:34 |
-!- blackburn [~qdrgsm@188.168.128.91] has left #shogun [] | 00:35 | |
n4nd0 | hi PhilTillet | 00:38 |
PhilTillet | hi n4nd0 :) | 00:39 |
gsomix | sonney2k, i sent pull request. | 00:41 |
gsomix | sorry for commit "Merge branch 'master'..." | 00:42 |
gsomix | good night, guys | 00:48 |
n4nd0 | good night! | 00:49 |
gsomix | tomorrow I will have a day full of theoretical mechanics. =____= | 00:50 |
n4nd0 | oh | 00:50 |
n4nd0 | good luck with that ;) | 00:50 |
n4nd0 | gsomix: what do you study btw? | 00:50 |
gsomix | n4nd0, physics, math, cs. bit of everything. | 00:53 |
n4nd0 | gsomix: cool stuff | 00:53 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 244 seconds] | 01:04 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 01:20 | |
-!- makokal [~makokal@212.201.44.246] has quit [Quit: makokal] | 01:23 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 01:39 | |
-!- makokal [~makokal@brmn-4d0acdfe.pool.mediaWays.net] has joined #shogun | 01:39 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 02:11 | |
-!- makokal [~makokal@brmn-4d0acdfe.pool.mediaWays.net] has quit [Ping timeout: 246 seconds] | 02:39 | |
-!- makokal [~makokal@brmn-4db71722.pool.mediaWays.net] has joined #shogun | 02:40 | |
-!- av3ngr [av3ngr@nat/redhat/x-dhyqmsdozibizipk] has quit [Quit: That's all folks!] | 03:36 | |
-!- pluskid [~chatzilla@60.191.2.230] has joined #shogun | 03:40 | |
pluskid | where is LinearByteKernel? is it removed? or renamed to something else? | 03:57 |
-!- derekhh [8f59be22@gateway/web/freenode/ip.143.89.190.34] has joined #shogun | 04:06 | |
-!- pluskid_ [~chatzilla@60.191.2.230] has joined #shogun | 04:16 | |
-!- pluskid [~chatzilla@60.191.2.230] has quit [Ping timeout: 260 seconds] | 04:18 | |
-!- pluskid_ is now known as pluskid | 04:18 | |
-!- pluskid [~chatzilla@60.191.2.230] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120312181643]] | 04:24 | |
-!- vikram360 [~vikram360@117.192.168.229] has joined #shogun | 05:14 | |
-!- Varun_Agrawal [~chatzilla@triband-mum-120.62.15.172.mtnl.net.in] has joined #shogun | 05:41 | |
Varun_Agrawal | Hey guys! | 05:41 |
Varun_Agrawal | Sorry for the long hiatus | 05:41 |
Varun_Agrawal | Had my GRE yesterday. | 05:41 |
-!- Varun_Agrawal [~chatzilla@triband-mum-120.62.15.172.mtnl.net.in] has quit [Client Quit] | 05:44 | |
derekhh | wish you gl :) | 05:52 |
-!- derekhh [8f59be22@gateway/web/freenode/ip.143.89.190.34] has quit [Quit: Page closed] | 05:52 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 07:30 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 08:12 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 08:27 | |
-!- PhilTillet [~Philippe@38.124.160.27] has quit [Ping timeout: 248 seconds] | 09:02 | |
-!- Miggy [~Piggy@14.139.82.6] has joined #shogun | 09:03 | |
-!- Miggy [~Piggy@14.139.82.6] has quit [Quit: Leaving] | 09:16 | |
gsomix | hi | 09:18 |
n4nd0 | gsomix: hey! | 09:59 |
n4nd0 | gsomix: do you think you can help me a moment with git? | 09:59 |
gsomix | n4nd0, let's try. however, i sometimes experience problems with it. | 10:07 |
n4nd0 | gsomix: so I was on the point of doing a pull request | 10:07 |
n4nd0 | but it didn't allow me to push before pulling first | 10:08 |
n4nd0 | so I fetched shogun's repo and rebased it | 10:08 |
n4nd0 | and later I pushed | 10:08 |
n4nd0 | and it has turned out to be like this | 10:08 |
n4nd0 | https://github.com/iglesias/shogun/commits/multiclass_1vs1?page=1 | 10:08 |
-!- harshit_ [~harshit@59.177.203.139] has joined #shogun | 10:08 | |
n4nd0 | you can see that lot of the commits are done twice :( | 10:08 |
n4nd0 | one authored by me and the other the original | 10:09 |
n4nd0 | and now if I try to pull request | 10:09 |
n4nd0 | it understands that I am pushing all that stuff as new | 10:09 |
n4nd0 | not just the changes I have done but also the changes I got when I rebased | 10:09 |
n4nd0 | do you see what I mean? | 10:10 |
gsomix | hmm | 10:10 |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 10:16 | |
n4nd0 | harshit_: hey! yesterday we spoke a bit with sonney2k about deep learning and stuff, I suggest you to check the irclogs | 10:20 |
n4nd0 | if you have not already done so of course :) | 10:21 |
harshit_ | okay i'll have a look at it | 10:21 |
harshit_ | actually i am in my class right now, so i'll look at it in an hour or so.. | 10:22 |
n4nd0 | sure, no problem, they will remain there ;) | 10:22 |
harshit_ | and thanks for talking dude | 10:23 |
harshit_ | :) | 10:23 |
-!- in3xes [~in3xes@180.149.49.230] has quit [Ping timeout: 264 seconds] | 10:25 | |
gsomix | n4nd0, unfortunately, I have no ideas about it. | 10:36 |
gsomix | i will try to appeal to blackburn. | 10:36 |
n4nd0 | gsomix: no worries man! I just fixed it :) | 10:36 |
n4nd0 | https://github.com/iglesias/shogun/commits/multiclass_1vs1 so nice again :) | 10:37 |
n4nd0 | gsomix: thank you very much anyway for taking a look to it ;) | 10:37 |
gsomix | n4nd0, oh, ok. | 10:38 |
-!- blackburn [~qdrgsm@188.168.128.91] has joined #shogun | 10:38 | |
gsomix | n4nd0, tell me what you did, please. | 10:39 |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 10:40 | |
blackburn | what's up | 10:41 |
gsomix | blackburn, hi | 10:41 |
blackburn | hi | 10:41 |
wiking | blackburn: yo | 10:41 |
blackburn | wiking: yo | 10:41 |
blackburn | :D | 10:41 |
wiking | :>> | 10:42 |
blackburn | gsomix: please rebase your pull request up to latest master | 10:43 |
blackburn | and push it again | 10:43 |
blackburn | git push --force origin if needed | 10:44 |
n4nd0 | blackburn: that was the trick I had to use git push --force | 10:44 |
n4nd0 | blackburn: why does that happen? | 10:44 |
blackburn | n4nd0: what exactly? | 10:45 |
n4nd0 | blackburn: that we need to use --force after rebasing | 10:45 |
blackburn | n4nd0: origin at github becomes unsynchronized with your local git | 10:47 |
n4nd0 | n4nd0: aham, I see | 10:48 |
blackburn | actually it shouldn't require force anytime | 10:48 |
n4nd0 | blackburn: fuck, talking to me again | 10:48 |
n4nd0 | I had to do it before, at the beginning I avoided it but I almost screwed my branch | 10:48 |
blackburn | I don't get why it can't fast-forward it though | 10:48 |
wiking | blackburn: don't force it | 10:49 |
wiking | n4nd0: don't force it | 10:49 |
wiking | you'll fuck up your commit history... | 10:49 |
wiking | and then people following you will have problems | 10:49 |
wiking | so rather try a rebase | 10:50 |
wiking | and if that does not work | 10:50 |
n4nd0 | wiking: how should we do it then? | 10:50 |
wiking | do a merge | 10:50 |
n4nd0 | wiking: ok I see | 10:50 |
n4nd0 | wiking: I did the rebase but later it didn't allow me to push to my branch because of this non-fast-forward not allowed | 10:50 |
n4nd0 | wiking: so when that happens, better to merge? | 10:51 |
wiking | well | 10:51 |
wiking | first try a git pull | 10:51 |
wiking | so if you did rebase and it did not allow you to have a git push after that | 10:51 |
wiking | try a git pull | 10:51 |
wiking | it'll try to do a simple merge | 10:51 |
wiking | and then after that try a git push | 10:51 |
blackburn | hah, that makes more sense, agree | 10:52 |
n4nd0 | ok, thank you guys :) | 10:52 |
wiking | no worries | 10:52 |
blackburn | actually I think force is required only in case of wrong rebasing | 10:52 |
blackburn | it should work as fast-forward thing.. | 10:52 |
n4nd0 | blackburn: I have new news about one-vs-one | 10:53 |
blackburn | yes? | 10:53 |
n4nd0 | blackburn: I had to change a couple of things more because the seg faults persisted https://github.com/iglesias/shogun/commit/b0d1a28229aba6604bc325a4fec3aab6bf5900bc | 10:54 |
n4nd0 | the features in m_machines were NULL at training | 10:54 |
n4nd0 | I fixed that with the changes in MulticlassMachine:89 | 10:55 |
blackburn | hmm | 10:57 |
n4nd0 | and the other thing I had to do, I don't like that much how it's though is MulticlassMachine:68 | 10:57 |
blackburn | n4nd0: I am afraid line 90 is wrong | 10:59 |
blackburn | consider case data=NULL and m_features was set before | 11:00 |
blackburn | and train() is called | 11:00 |
blackburn | it will init train machine with NULL | 11:00 |
n4nd0 | but init train machine doesn't set anything if data is NULL | 11:02 |
n4nd0 | it just refreshes the las set features for m_machines | 11:02 |
-!- harshit_ [~harshit@59.177.203.139] has quit [Ping timeout: 248 seconds] | 11:03 | |
blackburn | hmm sorry I have to go | 11:03 |
blackburn | I'll reach you in a hour or little more | 11:03 |
n4nd0 | no problem | 11:03 |
-!- blackburn [~qdrgsm@188.168.128.91] has quit [Quit: Leaving.] | 11:03 | |
-!- harshit_ [~harshit@59.177.203.139] has joined #shogun | 11:08 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 11:18 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 11:19 | |
-!- makokal [~makokal@brmn-4db71722.pool.mediaWays.net] has quit [Quit: makokal] | 11:36 | |
-!- gsomix [~gsomix@188.168.128.179] has quit [Ping timeout: 244 seconds] | 11:39 | |
-!- gsomix [~gsomix@188.168.13.36] has joined #shogun | 11:49 | |
-!- harshit_ [~harshit@59.177.203.139] has quit [Ping timeout: 276 seconds] | 12:06 | |
-!- romovpa [bc2c2ad0@gateway/web/freenode/ip.188.44.42.208] has joined #shogun | 12:22 | |
-!- vikram360 [~vikram360@117.192.168.229] has quit [Ping timeout: 264 seconds] | 12:30 | |
romovpa | hi there, I want to participate in gsoc. I should demonstrate my familiarity with git and ability to build the sources | 12:31 |
romovpa | is there somebody who tell me how to do that? | 12:31 |
romovpa | hmm..? | 12:32 |
n4nd0 | hey romovpa | 12:40 |
n4nd0 | first you should fork the repo in github, checkout the source, get a bit used to it ... | 12:40 |
n4nd0 | so you can work on something and submit a small patch to the project! :) | 12:40 |
n4nd0 | if you and idea you'd like to do for this patch, it is ok then | 12:40 |
romovpa | I have already forked and touched it | 12:40 |
n4nd0 | romovpa: good! | 12:41 |
n4nd0 | so would you like to contribute with something in special? | 12:41 |
romovpa | here is my small patch: https://github.com/shogun-toolbox/shogun/pull/331 | 12:41 |
n4nd0 | that one is already merged | 12:42 |
n4nd0 | good then | 12:42 |
romovpa | hmm... should I do anything more? | 12:42 |
n4nd0 | should?? | 12:43 |
n4nd0 | I mean it's up to you :) | 12:43 |
romovpa | ok, =) thanks | 12:43 |
romovpa | Let it be my success report (I'm filling up my application now) | 12:44 |
n4nd0 | romovpa: I hope you didn't take that wrong | 12:44 |
n4nd0 | good luck with it | 12:45 |
n4nd0 | may I ask what project are you applying for? | 12:45 |
gsomix | romovpa, hello from Russia. | 12:45 |
romovpa | I am going to implement a bundle method solver for SO-SVM | 12:46 |
romovpa | gsomix: hello, could we know each other? | 12:47 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 12:47 | |
blackburn | n4nd0: ok lets check | 12:48 |
gsomix | romovpa, i do not think so. | 12:48 |
n4nd0 | blackburn: so what did you say before that could go wrong? something that was getting null | 12:49 |
blackburn | ok let me describe logic | 12:49 |
blackburn | m_features/m_kernel is always a state | 12:49 |
blackburn | so then you set features/kernel and call train() it should train machine and prepare for applying to exactly these ones | 12:50 |
blackburn | when apply(CFeatures) is called it sets features for machines and applies | 12:50 |
n4nd0 | ok, I understand | 12:50 |
n4nd0 | the problem here is that somehow if we just do init_machines_for_applly before starting to train; it seg faults | 12:52 |
blackburn | oh | 12:52 |
blackburn | sh | 12:52 |
blackburn | yes | 12:52 |
blackburn | init_machines_for_apply should be called AFTER train | 12:52 |
n4nd0 | aham I see | 12:52 |
n4nd0 | just before the return? | 12:52 |
blackburn | yes probably | 12:52 |
blackburn | did you get this logic? | 12:53 |
n4nd0 | ok | 12:53 |
n4nd0 | yeah, I think so | 12:53 |
n4nd0 | I think that solution is better | 12:53 |
blackburn | 3 cases | 12:53 |
blackburn | machine.set_features(train_features) | 12:53 |
blackburn | machine.train() | 12:53 |
blackburn | machine.apply() | 12:53 |
blackburn | next | 12:53 |
n4nd0 | but still I don't know what may be wrond doing init_machines_for_train(NULL) in apply() | 12:53 |
blackburn | machine.train(features) | 12:53 |
blackburn | machine.apply() | 12:53 |
blackburn | in apply? | 12:54 |
blackburn | wait | 12:54 |
n4nd0 | yes | 12:54 |
blackburn | damn this naming is pretty hard to understand | 12:54 |
n4nd0 | at it's right now in MulticlassMachine.cpp:68 | 12:54 |
blackburn | n4nd0: could you please rename it? | 12:54 |
blackburn | init_training_machine or so | 12:54 |
blackburn | init_apply_machines | 12:54 |
blackburn | may be tihs way? | 12:54 |
n4nd0 | ok | 12:55 |
blackburn | machine and machines look confusing | 12:55 |
blackburn | lemme check the source | 12:55 |
blackburn | n4nd0: ok let me describe a wrong case | 12:56 |
blackburn | machine.apply(features) | 12:56 |
n4nd0 | blackburn: that was one issue, there is another to look at | 12:56 |
n4nd0 | blackburn: sorry, continue | 12:57 |
blackburn | when you call it | 12:57 |
blackburn | it sets features | 12:57 |
blackburn | and calls apply() | 12:57 |
blackburn | this time apply() sets features with NULL | 12:57 |
blackburn | and BADDABOOM | 12:57 |
n4nd0 | I don't think so | 12:58 |
n4nd0 | because init_machines_for_apply | 12:58 |
n4nd0 | (init_apply_machines) | 12:58 |
n4nd0 | doesn't set the features if they're NULL | 12:59 |
blackburn | really? | 12:59 |
n4nd0 | if will skip the set features and reload the ones for m_machines | 12:59 |
n4nd0 | the latter is what we are interested in | 12:59 |
blackburn | it was beyond my intention | 12:59 |
blackburn | oh really | 12:59 |
blackburn | I got wrong with my own logic | 12:59 |
n4nd0 | idk, maybe init_machines_for_apply doesn't do the right thing in LinearMulticlassClassifier | 13:00 |
n4nd0 | but it does so this far I think | 13:00 |
blackburn | damn I've got some doubts | 13:02 |
blackburn | anyway | 13:03 |
blackburn | use cases are | 13:03 |
blackburn | set features / set kernel | 13:03 |
blackburn | train() | 13:03 |
blackburn | apply() | 13:03 |
n4nd0 | in any case I think that to do init_machines_for_apply at the end of train is better | 13:03 |
blackburn | set features / set kernel | 13:03 |
blackburn | train() | 13:03 |
blackburn | apply(features) | 13:03 |
blackburn | and | 13:03 |
blackburn | train(features) | 13:03 |
blackburn | apply(other_features) | 13:03 |
blackburn | no more cases I guess | 13:03 |
n4nd0 | I will check if they're ok | 13:04 |
n4nd0 | apart from that | 13:04 |
n4nd0 | I also changed MulticlassMachine::train_machine | 13:05 |
n4nd0 | the very beginning | 13:05 |
n4nd0 | lines 87 - 90 | 13:06 |
n4nd0 | it's a similar idea, but the thing is that if it is not done this way, it seg faults in training | 13:06 |
blackburn | it should fail if it !is_ready() | 13:07 |
n4nd0 | I am thinking now that maybe it is appropriate to do init_train_machines at the end of the constructor that takes features as argument | 13:07 |
blackburn | and there is no data | 13:07 |
blackburn | if there is no data* | 13:07 |
n4nd0 | yeah I think that it's true, if features was not initialized in the constructor | 13:09 |
n4nd0 | then the constructor with features must call init_machines_for_train | 13:10 |
n4nd0 | do you see my point? | 13:10 |
blackburn | yes! | 13:10 |
blackburn | that was forgotten probably | 13:10 |
n4nd0 | ok | 13:11 |
blackburn | machine_for_train | 13:11 |
blackburn | not machines :D | 13:11 |
blackburn | yeah embarrasing naming | 13:11 |
n4nd0 | haha yeah I noted that | 13:11 |
blackburn | feel free to change it | 13:11 |
n4nd0 | ok | 13:12 |
blackburn | it is in guts of these classes so ok to change | 13:12 |
blackburn | no API | 13:12 |
n4nd0 | can we impose somehow that class that inherit from MulticlassMachine | 13:12 |
n4nd0 | are compulsed to do these init_training_machine | 13:12 |
n4nd0 | in the constructor that accepts Features? | 13:13 |
blackburn | it calls constructor of base class | 13:13 |
n4nd0 | ok | 13:13 |
n4nd0 | I think I need to implement that constructor then | 13:13 |
blackburn | so while init_training_machine is virtual | 13:13 |
blackburn | oh wait | 13:14 |
n4nd0 | what happened? | 13:15 |
blackburn | hmm no way to do that probably | 13:16 |
n4nd0 | yeah | 13:16 |
blackburn | then it shouldn't be in the constructor | 13:17 |
n4nd0 | wait I don't see it | 13:17 |
n4nd0 | why no way? | 13:17 |
blackburn | well multiclass machine constructor can't get any features | 13:17 |
n4nd0 | why not? | 13:17 |
blackburn | currently I mean | 13:18 |
n4nd0 | even if features is not a member | 13:18 |
blackburn | wait | 13:18 |
blackburn | why not to init_training_machine in train_machine? | 13:18 |
blackburn | in case of NULL data and is_ready()? | 13:18 |
blackburn | isn't it exactly what we need? | 13:18 |
n4nd0 | yes ... I think so at least | 13:19 |
n4nd0 | then we will call init_training_machine with NULL | 13:19 |
blackburn | yes | 13:19 |
n4nd0 | ok | 13:20 |
blackburn | I hope this all makes sense | 13:20 |
n4nd0 | we'll see | 13:21 |
n4nd0 | one thing about the doc, where should I put it? this might what S?ren told me yesterday but I might have not understood | 13:26 |
n4nd0 | I mean for example, for LDA | 13:26 |
n4nd0 | I would like to do sth similar for QDA | 13:27 |
n4nd0 | as it appears in doxygen doc | 13:27 |
n4nd0 | but I have not found that documentation in the code | 13:27 |
n4nd0 | wait, I said nothing | 13:27 |
n4nd0 | it is in the code ... | 13:28 |
blackburn | n4nd0: yeah @brief part in .h | 13:29 |
n4nd0 | blackburn: I am working in the code, in the train_machine part | 13:32 |
n4nd0 | I really think that if ( !is_read() && !data ) then SG_ERROR | 13:32 |
n4nd0 | in any other case init_training_machine | 13:33 |
n4nd0 | init_training_machine(data) simply | 13:33 |
n4nd0 | i.e. as it is in my branch right now | 13:33 |
blackburn | yes | 13:34 |
n4nd0 | :) | 13:34 |
n4nd0 | is it normal that LibLinear doesn't always give the same results? | 13:35 |
blackburn | n4nd0: multiclass? | 13:36 |
n4nd0 | the example I am using for MulticlassMachine uses LinearMulticlassMachine with LibLinear | 13:36 |
n4nd0 | blackburn: yes | 13:36 |
blackburn | yes it is a little random | 13:36 |
blackburn | results shouldn't be *very* different though | 13:36 |
n4nd0 | :O | 13:37 |
n4nd0 | every once in a while | 13:37 |
blackburn | what kind of random you've got? | 13:37 |
n4nd0 | I get a bunch of class 3 detected for what it is class 0 | 13:37 |
blackburn | random labels? | 13:37 |
n4nd0 | yeah | 13:37 |
n4nd0 | normally it classifies good | 13:37 |
n4nd0 | but every 1 out 4 or 5 let's say, that happens | 13:37 |
blackburn | that's strange | 13:37 |
n4nd0 | possible bug you think? | 13:38 |
blackburn | yes it should be almost equal.. | 13:39 |
n4nd0 | it happens just with ONE_VS_ONE actually :S | 13:39 |
blackburn | ah | 13:41 |
blackburn | wait | 13:41 |
blackburn | what? | 13:41 |
blackburn | oh | 13:41 |
blackburn | :D | 13:41 |
blackburn | ah | 13:41 |
n4nd0 | haha | 13:41 |
n4nd0 | what's going on? | 13:41 |
n4nd0 | haha | 13:41 |
blackburn | I thought you mean MulticlassLibLinear | 13:41 |
blackburn | and was very afraid of it | 13:41 |
blackburn | :D | 13:41 |
n4nd0 | no no | 13:41 |
n4nd0 | so what do you think then? | 13:42 |
n4nd0 | bad result? | 13:42 |
blackburn | n4nd0: well depends on data actually but strange anyway | 13:42 |
-!- romovpa [bc2c2ad0@gateway/web/freenode/ip.188.44.42.208] has quit [Quit: Page closed] | 13:46 | |
blackburn | n4nd0: valgrind is the tool ;) | 13:46 |
blackburn | n4nd0: what is the email addr you reached nico with? | 13:50 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 14:10 | |
-!- Peter_Gr [~Peter_Gr@115.248.130.148] has joined #shogun | 14:43 | |
-!- vikram360 [~vikram360@117.192.168.229] has joined #shogun | 14:45 | |
wiking | blackburn: here | 15:16 |
wiking | ? | 15:17 |
blackburn | yeah | 15:17 |
wiking | ok so now it's starting to be ready | 15:20 |
wiking | just one more commit | 15:20 |
blackburn | wiking: what exactly? ;) | 15:21 |
wiking | latent svm | 15:21 |
wiking | api | 15:21 |
blackburn | ah | 15:22 |
blackburn | lets see | 15:22 |
wiking | ok | 15:25 |
wiking | check this | 15:25 |
wiking | https://github.com/vigsterkr/shogun/commits/latent | 15:25 |
wiking | last two commits | 15:25 |
wiking | or rather | 15:25 |
wiking | https://github.com/vigsterkr/shogun/blob/6fef385b84ea003f398c32338337ae73ca06293f/src/shogun/classifier/svm/LatentLinearMachine.h | 15:26 |
blackburn | I'd suggest to move these things to shogun'/latent | 15:27 |
wiking | ? | 15:28 |
blackburn | i think it is ok to move all latent domain specific things to shogun/latent | 15:28 |
blackburn | new dir | 15:28 |
blackburn | :) | 15:28 |
wiking | but why? | 15:28 |
wiking | anyhow here you go a new stupid pull request for shogun :)) | 15:29 |
wiking | sent | 15:29 |
blackburn | wiking: I think domain specific separation is ok | 15:30 |
blackburn | I actually plan to move ready multitask things to shogun/multitask | 15:30 |
blackburn | and multiclass things to shogun/multiclass | 15:30 |
wiking | aha | 15:30 |
wiking | btw is this your email adds: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 15:31 |
wiking | ? | 15:31 |
blackburn | wiking: yes both works | 15:31 |
blackburn | blackburn91 at gmail | 15:31 |
blackburn | and lisitsyn.s.o | 15:32 |
wiking | ok i'll send there now the email to you as well | 15:32 |
wiking | that goes for alex as well | 15:32 |
blackburn | ah ok sure | 15:32 |
wiking | ok email sent... apply my pull request :D | 15:41 |
blackburn | wiking: can't right now | 15:47 |
wiking | ok | 15:47 |
blackburn | hmm didn't received your mail | 15:47 |
blackburn | would be funny if it was considered as spam | 15:47 |
wiking | ahhahahahah | 15:49 |
wiking | maybe it is | 15:49 |
blackburn | ah it is ok | 15:51 |
-!- harshit_ [~harshit@182.68.160.94] has joined #shogun | 15:52 | |
-!- harshit_ [~harshit@182.68.160.94] has quit [Ping timeout: 252 seconds] | 16:08 | |
-!- harshit_ [~harshit@182.68.160.94] has joined #shogun | 16:22 | |
-!- harshit_ [~harshit@182.68.160.94] has quit [Ping timeout: 248 seconds] | 16:47 | |
-!- harshit_ [~harshit@182.68.160.94] has joined #shogun | 16:47 | |
-!- jckrz [~jacek@89-69-164-5.dynamic.chello.pl] has joined #shogun | 16:51 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 16:51 | |
-!- harshit_ [~harshit@182.68.160.94] has quit [Ping timeout: 248 seconds] | 16:52 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 17:08 | |
-!- blackburn [~qdrgsm@188.168.128.91] has joined #shogun | 18:01 | |
-!- harshit_ [~harshit@182.68.160.94] has joined #shogun | 18:02 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 18:14 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 18:20 | |
-!- blackburn [~qdrgsm@188.168.128.91] has quit [Quit: Leaving.] | 18:30 | |
-!- flxb [~cronor@fb.ml.tu-berlin.de] has joined #shogun | 18:34 | |
flxb | hey all | 18:37 |
flxb | I saved a classifier with pickle.dump() and then loaded it. When I do apply it throws "terminate called after throwing an instance of 'shogun::ShogunException'". Is there a way to get a more meaningful error? | 18:39 |
-!- flxb [~cronor@fb.ml.tu-berlin.de] has left #shogun [] | 18:42 | |
-!- flxb [~cronor@fb.ml.tu-berlin.de] has joined #shogun | 18:44 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 18:45 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 18:55 | |
-!- flxb [~cronor@fb.ml.tu-berlin.de] has quit [Quit: flxb] | 18:55 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 19:00 | |
-!- blink_123 [~blink@117.227.136.222] has joined #shogun | 19:08 | |
-!- harshit_ [~harshit@182.68.160.94] has quit [Ping timeout: 240 seconds] | 19:15 | |
-!- blink_123 [~blink@117.227.136.222] has quit [Ping timeout: 246 seconds] | 19:19 | |
-!- makokal [~makokal@212.201.44.246] has joined #shogun | 19:20 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has quit [Ping timeout: 276 seconds] | 19:26 | |
-!- muddo [~muddo@gateway/tor-sasl/muddo] has joined #shogun | 19:29 | |
-!- makokal [~makokal@212.201.44.246] has quit [Quit: makokal] | 19:45 | |
-!- Marty [9eb54c39@gateway/web/freenode/ip.158.181.76.57] has joined #shogun | 20:04 | |
Marty | test | 20:04 |
Marty | Hi, anybody on? | 20:07 |
Marty | Running svm.train() Traceback (most recent call last): File "test_wd_Celegans.py", line 64, in <module> svm.train() File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 20344, in train return _modshogun.Machine_train(self, data) SystemError: [ERROR] assertion num<get_num_vectors() failed in file features/StringFeatures.cpp line 264 | 20:09 |
-!- Marty [9eb54c39@gateway/web/freenode/ip.158.181.76.57] has quit [Quit: Page closed] | 20:13 | |
-!- flxb [~cronor@g231228254.adsl.alicedsl.de] has joined #shogun | 20:19 | |
-!- MartysBrother [~chatzilla@158.181.76.57] has joined #shogun | 20:22 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 20:23 | |
-!- MartysBrother is now known as Marty28 | 20:24 | |
Marty28 | testing again | 20:25 |
Marty28 | Running svm.train() | 20:25 |
Marty28 | Traceback (most recent call last): | 20:25 |
Marty28 | File "test_wd_Celegans.py", line 64, in <module> | 20:25 |
Marty28 | svm.train() | 20:25 |
Marty28 | File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 20344, in train | 20:25 |
Marty28 | return _modshogun.Machine_train(self, data) | 20:26 |
Marty28 | SystemError: [ERROR] assertion num<get_num_vectors() failed in file features/StringFeatures.cpp line 264 | 20:26 |
Marty28 | ----------------------- | 20:26 |
Marty28 | This I get with WeightedDegreeStringKernel | 20:26 |
Marty28 | My small toy example works. Are there size limitations for feature data? I have 2200 x 141 DNA base pairs as training data. | 20:28 |
-!- makokal [~makokal@brmn-4db71722.pool.mediaWays.net] has joined #shogun | 20:28 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 20:38 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 20:40 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 20:42 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun | 20:46 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has quit [Client Quit] | 20:47 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 20:47 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 20:47 | |
-!- Marty28 [~chatzilla@158.181.76.57] has quit [Quit: ChatZilla 0.9.88.1 [Firefox 11.0/20120310010446]] | 20:47 | |
CIA-64 | shogun: Viktor Gal master * r306bc99 / src/shogun/distance/ChiSquareDistance.h : A little typo fix of define in ChiSquareDistance - http://git.io/vcwQnQ | 20:49 |
CIA-64 | shogun: Soeren Sonnenburg master * r1f71363 / src/shogun/distance/ChiSquareDistance.h : | 20:49 |
CIA-64 | shogun: Merge pull request #396 from vigsterkr/master | 20:49 |
CIA-64 | shogun: There was a little typo in the ChiSquareDistance header - http://git.io/qdOBqw | 20:49 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit] | 20:50 | |
-!- flxb_ [~cronor@e178175102.adsl.alicedsl.de] has joined #shogun | 20:58 | |
-!- flxb [~cronor@g231228254.adsl.alicedsl.de] has quit [Ping timeout: 246 seconds] | 21:01 | |
-!- flxb_ is now known as flxb | 21:01 | |
-!- blackburn [~qdrgsm@188.168.128.91] has joined #shogun | 21:14 | |
@sonney2k | gsomix, why did you close https://github.com/shogun-toolbox/shogun/pull/395? | 21:24 |
@sonney2k | except for missing rebase it looks very valid to me and should be merged...! | 21:26 |
blackburn | sonney2k: blame him in russian ;) | 21:39 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 21:45 | |
flxb | I saved a classifier with pickle.dump() and then loaded it. When I do apply it throws "terminate called after throwing an instance of 'shogun::ShogunException'". Is there a way to get a more meaningful error or what could this be? | 21:45 |
blackburn | flxb: try to add init_shogun_with_defaults() once you import shogun | 21:46 |
@sonney2k | blackburn, he is using python already (pickle...) | 21:50 |
@sonney2k | flxb, well you could do bla.io.set_loglevel(0) | 21:50 |
flxb | sonney2k: ok i don't know what effect init_shogun_with_defaults has. i'll try bla io log level 0 | 21:51 |
flxb | sonney2k: ha, but what is bla? it's not in modshogun | 21:54 |
blackburn | flxb: any shogun object you use | 21:54 |
@sonney2k | flxb, some shogun object :-) | 21:54 |
flxb | amusing, i'll give it a try | 22:02 |
n4nd0 | sonney2k: hi! I have a doubt related to solving SVM optimization with the primal | 22:08 |
@sonney2k | n4nd0, ? | 22:08 |
n4nd0 | sonney2k: so can it be solved directly using a QP solver (let's say CPLEX or libqp)? | 22:09 |
n4nd0 | sonney2k: is there no need in using an specific SVM solver for that? | 22:09 |
@sonney2k | n4nd0, svm's can be solved (in primal or dual) with off-the-shelve QP solvers | 22:10 |
@sonney2k | but that's slow and memory demanding | 22:10 |
@sonney2k | so tuned solutions are much faster / require less memory | 22:10 |
n4nd0 | sonney2k: aha, I understand | 22:10 |
n4nd0 | sonney2k: is there something already in shogun to solve the SVM in primal? | 22:11 |
blackburn | hey any crazy guy wanting to implement sparse matrix class? | 22:11 |
blackburn | :D | 22:11 |
@sonney2k | n4nd0, liblinear / libocas | 22:11 |
@sonney2k | SGD | 22:12 |
n4nd0 | blackburn: aren't CSparseFeatures for that? | 22:12 |
@sonney2k | SGD-QN | 22:12 |
blackburn | n4nd0: I need pretty specific one | 22:12 |
@sonney2k | probably more :) | 22:12 |
@sonney2k | blackburn, what's wrong with CSparseFeatures? | 22:12 |
n4nd0 | sonney2k: oh all these can be used for primal? cool! | 22:12 |
@sonney2k | n4nd0, primal (with no kernel!) | 22:12 |
blackburn | sonney2k: nothing but I need dymamic CSC matrix | 22:13 |
blackburn | dynamic | 22:13 |
blackburn | :D | 22:13 |
blackburn | sonney2k: I thought ocas is dual :( | 22:14 |
blackburn | my bad | 22:14 |
n4nd0 | sonney2k: hmm I have to check if they could be used for the problem I am thinking of, for SO-SVM optimization | 22:14 |
@sonney2k | blackburn, it uses cutting planes... | 22:14 |
blackburn | hmm true | 22:14 |
blackburn | I do not understand it thoroughly but I know it stands for s.t. part in primal ;) | 22:15 |
@sonney2k | n4nd0, I suspect you would need sth like libqp but ask nico | 22:18 |
n4nd0 | sonney2k: no way to do it with OCAS because of the kernel issue? | 22:19 |
@sonney2k | n4nd0, actually vojtech has an extension to ocas idea for SO | 22:20 |
@sonney2k | that is the other SO task btw | 22:20 |
@sonney2k | n4nd0, one is more the framework and some baseline algorithms, the other SO one would be one application of the framework but bleeding edge and potentially even better than svm struct | 22:21 |
n4nd0 | sonney2k: mmm but this idea other talks about libqp and not OCAS | 22:22 |
gsomix | sonney2k, I had some problems with repo. | 22:22 |
@sonney2k | gsomix, so can you open a new pull request? | 22:22 |
@sonney2k | n4nd0, this idea http://www.shogun-toolbox.org/gsoc-ideas.html#bmrm | 22:23 |
gsomix | sonney2k, yep. | 22:23 |
@sonney2k | hmmhh we have 5 student submissions by now - but none is from any of the active students here... | 22:24 |
n4nd0 | sonney2k: oh sorry, I'm kind of retarded ... I was checking another one | 22:24 |
gsomix | sonney2k, but only tomorrow. I have some changes and I will complete it tomorrow. | 22:24 |
n4nd0 | we have to get hands on to work with the application guys | 22:25 |
blackburn | oh not now | 22:25 |
n4nd0 | I think I am going to start tonight with it :) | 22:25 |
blackburn | sonney2k: would you mind I attach a list of my commits? | 22:25 |
n4nd0 | blackburn: :D | 22:26 |
@sonney2k | blackburn, the post gsoc ones only please :) | 22:26 |
blackburn | sonney2k: ok, I've commited about 200 after gsoc :D | 22:27 |
* gsomix was trying to do home work today. But he did little. (tag: whining.) | 22:29 | |
n4nd0 | blackburn: so should we attach the commits as well? I mean, isn't enough to give a brief description about what have we done? | 22:30 |
blackburn | haha | 22:30 |
blackburn | that was joke! | 22:30 |
n4nd0 | haha ok | 22:31 |
n4nd0 | at first I took it like that | 22:31 |
n4nd0 | but after the post gsoc comment | 22:31 |
n4nd0 | sonney2k: do you know if Nico has access to the applications? I'd like to ask him for feedback on the timeline for my application | 22:32 |
blackburn | well sonney2k is a great joker ;) | 22:32 |
n4nd0 | I wonder if I should send it to him directly | 22:32 |
@sonney2k | n4nd0, yes - he is a mentor but it doesn't hurt to explicitly ask him | 22:34 |
n4nd0 | thank you :) | 22:35 |
n4nd0 | sonney2k: is it important to write some references (e.g. professors we have interactuated with)? or not really that much | 22:43 |
@sonney2k | n4nd0, everything that separates you from other students helps! | 22:46 |
blackburn | e.g. if you've got three legs | 22:47 |
n4nd0 | blackburn: :D | 22:48 |
blackburn | n4nd0: want to try dimensionality reduction? | 22:50 |
blackburn | :D | 22:50 |
blackburn | gsomix: want to try dimensionality reduction? | 22:51 |
n4nd0 | blackburn: yeah! sure | 22:51 |
n4nd0 | blackburn: what do you have in mind? | 22:52 |
blackburn | n4nd0: examples, I lack examples | 22:52 |
n4nd0 | blackburn: yes, I like the idea | 22:52 |
n4nd0 | blackburn: have you thought anything in particular? | 22:53 |
blackburn | ok simplest | 22:53 |
blackburn | download a few images somewhere | 22:53 |
blackburn | compute hue histograms | 22:53 |
blackburn | embed it with isomap/mds | 22:53 |
blackburn | will be a nice picture :) | 22:53 |
blackburn | check applications/edrt/faces.py to see how to place images on plot with matplotlib | 22:54 |
blackburn | oh actually I had a script somewhere | 22:54 |
blackburn | btw it is my gf ;) | 22:54 |
blackburn | (faces.py) | 22:54 |
n4nd0 | blackburn: http://www.csc.kth.se/utbildning/kth/kurser/DD2427/bik12/DownloadMaterial/Lectures/Lecture1_part1.pdf | 22:55 |
n4nd0 | check the previous to last page | 22:55 |
n4nd0 | sth like that would be nice? | 22:55 |
blackburn | yeah | 22:56 |
blackburn | I'll give you some code in a min | 22:56 |
n4nd0 | blackburn: so these slides are from one of the courses I told you the other day | 22:57 |
blackburn | aha I got it | 22:57 |
n4nd0 | blackburn: we could prepare an example like this one we are talking about | 22:57 |
n4nd0 | blackburn: and I can mail the professor and show her | 22:58 |
blackburn | http://dl.dropbox.com/u/10139213/shogun/pics.png | 22:58 |
n4nd0 | blackburn: maybe that way we can start some shogun-university courses contact | 22:58 |
blackburn | why not :) | 22:58 |
blackburn | n4nd0: ^ that's how it worked | 22:58 |
n4nd0 | blackburn: looks nice | 22:59 |
n4nd0 | blackburn: you have the code for that right? | 22:59 |
blackburn | yeah let me find it | 22:59 |
gsomix | blackburn, dimensional reduction? no, I do not want to live in two-dimensional world. | 22:59 |
gsomix | if seriously, i will work on python3 and covertee. | 22:59 |
blackburn | remember I always have tasks | 22:59 |
blackburn | :D | 22:59 |
blackburn | n4nd0: ok it is pretty crappy but here it goes | 23:00 |
blackburn | http://pastebin.com/QzBwcph5 | 23:00 |
blackburn | there is even a hog function I borrowed somewhere | 23:00 |
CIA-64 | shogun: Soeren Sonnenburg master * rc375b8f / (5 files in 3 dirs): make director based classes optional - http://git.io/sXQslA | 23:00 |
blackburn | n4nd0: all you need is to get some pictures | 23:01 |
blackburn | ideally retrieve from google.. but it seems they do not provide such api | 23:01 |
n4nd0 | mmm not good | 23:02 |
blackburn | n4nd0: any other options are welcome | 23:02 |
blackburn | haha I just realized all the pictures are upside down | 23:02 |
n4nd0 | I will make a search about how to do get images, if it doesn't work, I'll pick some manually | 23:03 |
n4nd0 | probably we don't need that many to make a nice-looking example | 23:03 |
n4nd0 | any suggestion/preference of the class of pictures? | 23:03 |
blackburn | no, it is totally up to you | 23:03 |
blackburn | ah yes | 23:04 |
blackburn | install libsuperlu3 | 23:04 |
blackburn | and libarpack | 23:04 |
blackburn | to make this feasible | 23:04 |
blackburn | in case of 3k-5k images it would be slow w/o superlu+arpack | 23:04 |
blackburn | however I can hardly imagine that | 23:04 |
n4nd0 | ok | 23:04 |
blackburn | :D | 23:04 |
blackburn | n4nd0: have you checked applications/edrt btw? | 23:05 |
blackburn | there are words embedding example as well | 23:05 |
n4nd0 | I have not executed it | 23:05 |
n4nd0 | which one is it? | 23:05 |
blackburn | words.py | 23:05 |
n4nd0 | ... I feel stupid sometimes I ask :O | 23:06 |
blackburn | ah | 23:06 |
blackburn | words_embedding.py | 23:06 |
n4nd0 | blackburn: libarpack2 is ok I guess | 23:06 |
blackburn | yes | 23:06 |
n4nd0 | haha nice example | 23:08 |
blackburn | try faces_embedding.py as well | 23:08 |
n4nd0 | fun one :) | 23:09 |
n4nd0 | isn't she sad a couple of times but very separated on the plane :P? | 23:09 |
blackburn | too complex for any dim reduction algorithm | 23:09 |
blackburn | actually I should prepare data somehow | 23:10 |
n4nd0 | I see | 23:10 |
blackburn | with centering at leat | 23:10 |
blackburn | least* | 23:10 |
blackburn | http://shogun-toolbox.org/edrt/ | 23:10 |
blackburn | there are two more examples | 23:10 |
blackburn | with ORL and swissroll | 23:10 |
n4nd0 | I like swissroll | 23:10 |
blackburn | yes it is easy but not real | 23:11 |
blackburn | local tangent space alignment is ideal with swissroll | 23:11 |
blackburn | but not as good with real data | 23:11 |
@sonney2k | n4nd0, I want pictures of giraffes, zebras, horses etc :) | 23:17 |
n4nd0 | sonney2k: I'll look for those then ;) | 23:21 |
n4nd0 | oh, didn't understand anything in that last mail :S | 23:24 |
blackburn | hahah | 23:27 |
blackburn | heiko wrote in german oh lol | 23:27 |
blackburn | I shall answer him in russian | 23:30 |
n4nd0 | can you read it? | 23:31 |
blackburn | no, how can I | 23:31 |
blackburn | :) | 23:31 |
n4nd0 | haha don't know ... maybe you know German man | 23:32 |
blackburn | I wish I knew | 23:32 |
blackburn | ok sent | 23:33 |
blackburn | n4nd0: one more mail you won't understand ;) | 23:34 |
n4nd0 | blackburn: hahaha you are fun man | 23:34 |
n4nd0 | blackburn: wanna do a guess what google translate tells me? | 23:34 |
blackburn | something reasonable I think | 23:34 |
n4nd0 | quite reasonable actually | 23:35 |
n4nd0 | Heiko, | 23:35 |
n4nd0 | I generally do not understand what you wrote here! But to keep the conversation going write some thing about LibSVM too :) | 23:35 |
n4nd0 | Sergey | 23:35 |
blackburn | hah | 23:35 |
blackburn | right | 23:35 |
n4nd0 | so the reverser R in Russian is I | 23:35 |
n4nd0 | reversed* | 23:35 |
n4nd0 | or flipped | 23:35 |
blackburn | n4nd0: ya | 23:36 |
blackburn | not i | 23:36 |
blackburn | like in young | 23:36 |
blackburn | ah you mean "I" | 23:36 |
blackburn | :D | 23:36 |
blackburn | I thought of sound | 23:36 |
n4nd0 | ah ok | 23:36 |
blackburn | yes it is "I" | 23:36 |
n4nd0 | I was like ... there is no i in young so I lost sth probably | 23:36 |
blackburn | hahh | 23:37 |
@sonney2k | blackburn, heikos mail was a private mail to me ... | 23:40 |
blackburn | sonney2k: why did he CC shogun list then? | 23:41 |
@sonney2k | blackburn, I think he accidentally sent it to shogun list | 23:41 |
@sonney2k | in the end it says 'migration business is complex' | 23:41 |
blackburn | sonney2k: well why not to joke ;) | 23:41 |
@sonney2k | he managed to fix things but it took quite long | 23:41 |
@sonney2k | and he is going to submit a patch ... | 23:42 |
blackburn | yes I did make use of google translate already | 23:42 |
gsomix | good night, guys | 23:47 |
n4nd0 | gsomix: good night | 23:49 |
@sonney2k | yeah good night | 23:49 |
blackburn | sonney2k: I don't think it was private ;) | 23:54 |
-!- vikram360 [~vikram360@117.192.168.229] has quit [Ping timeout: 252 seconds] | 23:54 | |
@sonney2k | blackburn, you mean my reply? no it wasn't intended to be :) | 23:55 |
@sonney2k | should I have replied in german? | 23:55 |
blackburn | sonney2k: yes or japanese according to way it was carried out ;) | 23:56 |
@sonney2k | blackburn, you can still join the thread and reply in russian :D | 23:56 |
blackburn | I already did once | 23:56 |
@sonney2k | then n4nd0 can continue in swedish or spanish :) | 23:57 |
blackburn | n4nd0: yeah continue thread with something swenish | 23:57 |
blackburn | or spadish | 23:57 |
blackburn | I actually usually write in runglish | 23:58 |
n4nd0 | I am a shy guy to reply | 23:58 |
-!- flxb [~cronor@e178175102.adsl.alicedsl.de] has quit [Read error: Connection reset by peer] | 23:58 | |
-!- flxb_ [~cronor@e178175102.adsl.alicedsl.de] has joined #shogun | 23:58 | |
blackburn | heiko from londongrad | 23:59 |
blackburn | :D | 23:59 |
@sonney2k | n4nd0, blackburn consumes a lot of vodka - you know he is *russian* | 23:59 |
blackburn | yes I do | 23:59 |
--- Log closed Wed Mar 28 00:00:03 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!