--- Log opened Thu Nov 01 00:00:17 2012 | ||
-!- wiking_ [~wiking@info2k1.hu] has joined #shogun | 00:17 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 264 seconds] | 00:19 | |
-!- ptizoom [~christian@85.210.80.20] has quit [Quit: Ex-Chat] | 03:24 | |
-!- ptizoom_ [~christian@85.210.80.20] has joined #shogun | 03:24 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 07:09 | |
-!- ptizoom_ is now known as ptizoom | 08:04 | |
blackburn | n4nd0: hey you were writing to my webirc client at job - but I wasn't at job :) | 08:19 |
---|---|---|
n4nd0 | blackburn: oops :P | 08:19 |
@sonney2k | blackburn, do you even have the slightest idea what ptizoom is talking about? | 08:19 |
blackburn | n4nd0: well my mistake to not close it | 08:19 |
blackburn | sonney2k: NO~ | 08:20 |
blackburn | !!! | 08:20 |
blackburn | 0% | 08:20 |
blackburn | nothing | 08:20 |
blackburn | ?????? | 08:20 |
@sonney2k | net | 08:20 |
@sonney2k | exactly | 08:20 |
n4nd0 | nada | 08:20 |
n4nd0 | ? | 08:20 |
blackburn | n4nd0: mas que nada! | 08:20 |
blackburn | :D | 08:20 |
@sonney2k | alright, let me continue the gsoc billing | 08:21 |
@sonney2k | *urgs* | 08:21 |
blackburn | sonney2k: big boss counting money? | 08:21 |
blackburn | sonney2k: no news from carol? | 08:22 |
@sonney2k | blackburn, is it one week yet? | 08:25 |
@sonney2k | blackburn, yeah counting money but heiko will receive it :D | 08:25 |
blackburn | sonney2k: one day more and yes I guess :) | 08:25 |
@sonney2k | well sorting bills lets say | 08:25 |
@sonney2k | even worse | 08:25 |
ptizoom | oye, ptizoom, was doing metaphors and poetry, as sure as cosacks can ride...killing time on ics. but anyway forget it... | 09:20 |
sonne|work | ptizoom: we appreciate some daily poem from ptizoom on shogun planet :D | 09:22 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 240 seconds] | 09:24 | |
ptizoom | I was also praising in the dojo of ML. | 09:24 |
ptizoom | anyway enough unflattering flaf...I still do not understand why ... | 09:26 |
ptizoom | *** Testing in ../data/classifier/*.m | 09:26 |
ptizoom | ../data/classifier/LibLinear_0017_1en05_1_True.m ERROR | 09:26 |
ptizoom | LibLinear - yesclassified: 5.644953e-02, bias: 9.286755e-04 <--- accuracy: 1.000000e-04 | 09:26 |
ptizoom | sonne*: shall this test be spot on, below accuracy .0001 ? | 09:27 |
blackburn | exactly | 09:28 |
ptizoom | so my perl stuff is very dodgy indeed... | 09:28 |
blackburn | ptizoom: but it is .m? how perl? | 09:30 |
ptizoom | I checked memory with debugger of vectorisation and matrification in/out and it shall be ok, but ... there must be a miss match in the engine... | 09:30 |
sonne|work | blackburn: man thats the test | 09:30 |
blackburn | sonne|work: matlab/octave test right? | 09:30 |
sonne|work | blackburn: no | 09:30 |
ptizoom | blackburn... yes | 09:31 |
ptizoom | oops | 09:31 |
blackburn | sonne|work: it tests perl? | 09:31 |
sonne|work | blackburn: this is just some file with the output | 09:31 |
blackburn | hmm okay | 09:31 |
ptizoom | there | 09:31 |
sonne|work | ptizoom: it suggests that sth is wrong with vector/matrix typemaps | 09:32 |
sonne|work | ptizoom: check if sth basic like the gaussian kernel works | 09:32 |
ptizoom | I was slowly but surely perlifying the tests in /tests/unit... | 09:37 |
ptizoom | that would tell me where are the deficiencies in the swig perlmodular... | 09:39 |
ptizoom | but hey, thanks for the tip. | 09:39 |
ptizoom | for info, not a single .m test works within accuracy ... | 09:41 |
ptizoom | ho gobble... | 09:42 |
sonne|work | ptizoom: well do sth very simple | 09:43 |
sonne|work | create some matrix | 09:43 |
sonne|work | from that x=RealFeatures(matrix) | 09:43 |
sonne|work | and then x.get_feature_matrix() (or what it is called) | 09:43 |
sonne|work | and compare if results match | 09:43 |
sonne|work | if that works do k=GaussianKernel(x,x) | 09:44 |
sonne|work | k.get_kernel_matrix() | 09:44 |
ptizoom | noted, I will come to that. thanks | 09:46 |
sonne|work | you can compare the result w/ what you get on python_modular | 09:51 |
ptizoom | I know... but that would be to easy, that would be cheating! | 09:51 |
-!- n4nd0 [~nando@dyna218-083.csc.kth.se] has joined #shogun | 10:12 | |
sonne|work | ptizoom: I prefer cheating^H^H^H^H^H^H^H^Hwinning | 11:34 |
-!- n4nd0 [~nando@dyna218-083.csc.kth.se] has quit [Ping timeout: 256 seconds] | 12:24 | |
-!- n4nd0 [~nando@dyna218-083.csc.kth.se] has joined #shogun | 12:50 | |
-!- romi_ [~mizobe@187.57.5.251] has joined #shogun | 13:30 | |
-!- n4nd0 [~nando@dyna218-083.csc.kth.se] has quit [Ping timeout: 260 seconds] | 14:52 | |
-!- Arya [bc9fd512@gateway/web/freenode/ip.188.159.213.18] has joined #shogun | 15:19 | |
Arya | Hi,I'm getting assertion failed when trying to read binary machines for multiclass classification. the line is : "CSVM *the_svm = (CSVM *)svm->get_machine(i);" which the_svm is Null | 15:22 |
blackburn | Arya: hi | 15:35 |
blackburn | Arya: how did you learn that "svm"? | 15:35 |
blackburn | what is the type? | 15:36 |
blackburn | Arya: my first guess is that binary machine could be of CLinearMachine class | 15:46 |
Arya | yes libsvmm , I'm running the classifier_multiclasslibsvm.cpp example on my dataset | 15:47 |
Arya | when the code reaches svm->apply() it happens | 15:48 |
blackburn | should be a bug when | 15:49 |
blackburn | Arya: my other guess is that could happen if you have more classes in test data than in train data | 15:50 |
Arya | no, I test it on train data, but the same happens | 15:51 |
blackburn | Arya: can you please check number of machines after training? method 'get_num_machines' | 15:52 |
Arya | it's correct, n(n-1)/2 | 15:53 |
Arya | int mc_libsvm(const char* ds, float64_t c, float64_t width, int32_t split) { init_shogun_with_defaults(); std::stringstream path; path<< "/home/arya/datasets/"<<ds << ".libsvm.train."<<split; uint32_t n=0,d=0; std::string train_file=path.str(); const char* train_file_name=train_file.c_str(); get_size(train_file, &n, &d); SGMatrix<float64_t> X(d,n); SGVector<float64_t> y(n); read_data(train_file_name, X, y); CLabels* train_ | 15:53 |
Arya | ooops | 15:53 |
Arya | can I send you my function? | 15:53 |
blackburn | Arya: pastebin maybe | 15:54 |
blackburn | or mail me blackburn91 at gmail.com | 15:55 |
blackburn | Arya: gtg now - please email me I'll take a look | 15:59 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 15:59 | |
-!- Arya [bc9fd512@gateway/web/freenode/ip.188.159.213.18] has quit [Quit: Page closed] | 16:05 | |
-!- romi_ [~mizobe@187.57.5.251] has quit [Quit: Leaving] | 16:25 | |
-!- arya [05c7817b@gateway/web/freenode/ip.5.199.129.123] has joined #shogun | 16:30 | |
-!- arya [05c7817b@gateway/web/freenode/ip.5.199.129.123] has quit [Quit: Page closed] | 16:54 | |
-!- n4nd0 [~nando@n183-p114.kthopen.kth.se] has joined #shogun | 16:57 | |
-!- blackburn [~blackburn@46.20.65.220] has joined #shogun | 17:13 | |
-!- wiking_ is now known as wiking | 17:58 | |
-!- wiking [~wiking@info2k1.hu] has quit [Changing host] | 17:58 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 17:58 | |
-!- audy [~audy@unaffiliated/audy] has quit [Remote host closed the connection] | 18:28 | |
-!- n4nd0 [~nando@n183-p114.kthopen.kth.se] has quit [Ping timeout: 252 seconds] | 19:55 | |
-shogungit:#shogun- [shogun] lisitsyn pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/a55b25e0b87ae2bbc5383a232f26d63ce2fad2ff | 20:04 | |
-shogungit:#shogun- shogun/master a55b25e Sergey Lisitsyn: Fix for 3-class training of multiclass libsvm | 20:04 | |
@sonney2k | blackburn, any clues on the bug? | 20:33 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:55 | |
blackburn | sonney2k: which bug? | 21:07 |
blackburn | ahhh | 21:07 |
blackburn | solved | 21:08 |
n4nd0 | blackburn: haha I see in the news Oksana is getting into shogun | 21:10 |
n4nd0 | well done! :) | 21:10 |
blackburn | :D | 21:10 |
blackburn | well I've spent a while trying to understand the error and she helped | 21:11 |
blackburn | so credit goes to | 21:11 |
@sonney2k | blackburn, what is the SG_PRINT("Num classes = %d\n",num_classes); in your patch? | 21:17 |
blackburn | sonney2k: oops juk | 21:17 |
blackburn | junk | 21:17 |
blackburn | here you go | 21:18 |
-shogungit:#shogun- [shogun] lisitsyn pushed 1 new commit to master: https://github.com/shogun-toolbox/shogun/commit/5c03620519cbc21b6c9009ebbb0c1ea5beb524d2 | 21:19 | |
-shogungit:#shogun- shogun/master 5c03620 Sergey Lisitsyn: Removed junk debug print | 21:19 | |
blackburn | n4nd0: # vectors10000 | 21:21 |
blackburn | Covertree construction took 7.880000 seconds | 21:21 |
blackburn | Covertree-based neighbors search took 21.230000 seconds | 21:21 |
n4nd0 | blackburn: what about the na?ve strategy? | 21:22 |
n4nd0 | what is the speedup with covertree wrt it? | 21:22 |
blackburn | n4nd0: partial_sort? | 21:22 |
n4nd0 | blackburn: yeah, the other one that was at the beginning | 21:22 |
blackburn | much slower I bet | 21:22 |
blackburn | n4nd0: before it was 14.94s for 10000 | 21:24 |
blackburn | 23s now | 21:24 |
blackburn | so a room for improvement still :D | 21:24 |
n4nd0 | so, that covertree is slower? | 21:25 |
blackburn | n4nd0: probably | 21:25 |
blackburn | JL is faster but it you are the only guy that managed to get it working | 21:25 |
n4nd0 | I still don't understand how is it possible that it can be slower than sorting! | 21:25 |
blackburn | n4nd0: it is impossible :D | 21:27 |
n4nd0 | but it is like that, isn't it? | 21:30 |
blackburn | n4nd0: ?? | 21:31 |
n4nd0 | that covertree is slower than na?ve sorting | 21:31 |
blackburn | which? | 21:31 |
n4nd0 | the one you are testing | 21:32 |
n4nd0 | you pasted ^ | 21:32 |
blackburn | no | 21:32 |
blackburn | you misunderstood me | 21:32 |
n4nd0 | oh I see, sorry | 21:32 |
blackburn | it was a covertree as well | 21:32 |
-!- travis-ci [~travis-ci@ec2-50-19-168-119.compute-1.amazonaws.com] has joined #shogun | 21:32 | |
travis-ci | [travis-ci] it's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/3022784 | 21:32 |
-!- travis-ci [~travis-ci@ec2-50-19-168-119.compute-1.amazonaws.com] has left #shogun [] | 21:32 | |
blackburn | travis-ci you are drunk go away | 21:33 |
n4nd0 | I love that guy, he has style | 21:34 |
n4nd0 | you know logs in | 21:34 |
n4nd0 | blame some people and go away | 21:34 |
n4nd0 | hehe | 21:34 |
blackburn | :D | 21:34 |
@sonney2k | blackburn, hurray - I finished the billing stuff :) | 22:38 |
@sonney2k | Just submitted to google | 22:38 |
@sonney2k | lets hope it works... | 22:38 |
blackburn | sonney2k: what is it? what you had to do? | 22:39 |
@sonney2k | blackburn, collect bills write some itemized billing and submit | 22:41 |
blackburn | no idea | 22:42 |
blackburn | :D | 22:42 |
blackburn | we just steal money here and stash it under a pillow | 22:43 |
@sonney2k | yeah I guess that is what you do in rossia | 22:43 |
@sonney2k | wiking, any ideas about https://github.com/shogun-toolbox/shogun/pull/785 | 22:44 |
@sonney2k | blackburn, do you know whether wiking is busy with some paper or sth? | 22:44 |
blackburn | sonney2k: he should be probably yes | 22:44 |
@sonney2k | argh | 22:45 |
@sonney2k | hungry | 22:45 |
@sonney2k | I need to have a second dinner | 22:45 |
blackburn | are you growing up? | 22:45 |
@sonney2k | not volume wise - but yes I guess doing the bills did cost a lot of energy ;) | 22:46 |
@sonney2k | yay pizza is waiting for me :) | 22:49 |
@sonney2k | yummy yummy yummy | 22:50 |
blackburn | :D | 22:50 |
n4nd0 | damn it, that made me hungry | 22:51 |
@sonney2k | n4nd0, just join me | 22:52 |
* sonney2k hands over a slice of pizza to n4nd0 | 22:52 | |
n4nd0 | sonney2k: danke | 22:52 |
@sonney2k | n4nd0, how is it going? | 22:53 |
blackburn | haha superfriendly sonney2k now | 22:53 |
* sonney2k hands another slice to blackburn | 22:53 | |
n4nd0 | sonney2k: good, what about you? | 22:53 |
@sonney2k | n4nd0, long time no code contributions? I guess you are pretty busy currently... | 22:54 |
@sonney2k | n4nd0, I am just glad that I managed to write this blogpost somehow | 22:54 |
@sonney2k | I suck at blogging | 22:55 |
blackburn | sonney2k: I employed him | 22:55 |
blackburn | :D | 22:55 |
blackburn | my scout reached him by email | 22:55 |
n4nd0 | sonney2k: yeah, I am trying to put KNN into the EDRT | 22:55 |
blackburn | you know I have scout team | 22:55 |
n4nd0 | hopefully I will have something ready by tom night | 22:55 |
@sonney2k | I see, you hijacked him | 22:55 |
blackburn | sonney2k: you have JMLR pubs we don't | 22:56 |
blackburn | :D | 22:56 |
@sonney2k | blackburn, does it make me feel better in any way? | 22:56 |
blackburn | sonney2k: you are the only who knows | 22:56 |
@sonney2k | then guess what the answer is | 22:57 |
blackburn | we are young and want to try various dangerous things haha | 22:57 |
n4nd0 | hehe good answer for a good question I think | 22:57 |
n4nd0 | sonney2k: I want to put the word recognizer application inside though | 22:57 |
@sonney2k | n4nd0, that would be cool | 22:57 |
n4nd0 | I think I will go a little bit slow with it but I'll do it | 22:57 |
@sonney2k | n4nd0, could we make this some live thing? | 22:58 |
blackburn | sonney2k: come on, all the EDRT stuff will be included to shogun anyway | 22:58 |
@sonney2k | blackburn, I always wanted to have KNN :P | 22:58 |
blackburn | sonney2k: it is already in shogun | 22:58 |
@sonney2k | blackburn, I know :) I wrote it | 22:58 |
n4nd0 | sonney2k: I have done a live one that worked fine but the problem is the data | 22:59 |
blackburn | sonney2k: I mean with covertree not sort | 22:59 |
n4nd0 | the one I did worked fine with my and bernard's voices | 22:59 |
n4nd0 | but it will probably explode with other's | 22:59 |
blackburn | sonney2k: NNs is a bottleneck for me so I hired a best specialist in JL's covertree around :) | 22:59 |
@sonney2k | blackburn, s/hired/hijacked :) | 23:00 |
@sonney2k | n4nd0, well then we need more people :) | 23:00 |
n4nd0 | blackburn, sonney2k : http://bernard.260mb.org/ | 23:00 |
n4nd0 | open speech recognition | 23:00 |
@sonney2k | n4nd0, speaking of bernard - is he doing fine? | 23:00 |
@sonney2k | since he wrote the website he basically disappeared | 23:00 |
blackburn | n4nd0: impressive CV I'd hire him too | 23:01 |
blackburn | :D | 23:01 |
n4nd0 | sonney2k: yeah, he is | 23:01 |
n4nd0 | sonney2k: we are doing together a CV course now and he is starting to look for a company where to do his master thesis | 23:01 |
blackburn | OMFG I have to do people detector | 23:02 |
blackburn | good you did remind me; | 23:02 |
blackburn | :D | 23:02 |
blackburn | ehm sleep reduction technique anyone? | 23:02 |
@sonney2k | n4nd0, we are trying to do some live demos on the website with shogun as backend and django & d3js etc as frontend. maybe he would be interested in playing with this? | 23:02 |
blackburn | yeah we will consult him with hyperplanes | 23:03 |
n4nd0 | sonney2k: I'll ask him right now | 23:03 |
blackburn | n4nd0: is he your flatmate? :D | 23:03 |
n4nd0 | not really | 23:03 |
@sonney2k | n4nd0, basically stuff like we have in the graphical examples + e.g. the ocr demo | 23:03 |
@sonney2k | but face recognition would be even cooler :D | 23:04 |
@sonney2k | + gender detection | 23:04 |
@sonney2k | but quite some effort | 23:04 |
n4nd0 | sonney2k: he said he is sorry but no :( | 23:04 |
blackburn | sonney2k: haha I came up with very offensive app | 23:04 |
@sonney2k | n4nd0, too bad. | 23:04 |
n4nd0 | sonney2k: face detection would be awesome but we need adaboost for that! | 23:04 |
@sonney2k | n4nd0, no | 23:04 |
blackburn | so you enter the website it takes image from your camera | 23:04 |
n4nd0 | that reminds me when I started to come around here, adaboost | 23:04 |
blackburn | and says you are a man | 23:05 |
@sonney2k | n4nd0, I would rather just use opencv's FD | 23:05 |
@sonney2k | it works OKish | 23:05 |
@sonney2k | and on top of it we put the gender classifier | 23:05 |
blackburn | opencv is a heap of api crap | 23:05 |
blackburn | :D | 23:05 |
@sonney2k | who cares | 23:05 |
@sonney2k | it does the job :D | 23:06 |
n4nd0 | hehe | 23:06 |
n4nd0 | I know they use Haar features | 23:06 |
n4nd0 | is it with adaboost too? | 23:06 |
blackburn | it has god damn WORST HOG thing ever | 23:06 |
blackburn | they should rather call it humandetectorarraycomparator :D | 23:06 |
@sonney2k | n4nd0, some cascades - decision trees maybe? | 23:06 |
@sonney2k | (wildly guessing) | 23:06 |
n4nd0 | maybe a cascade of strong classifiers with adaboost | 23:07 |
n4nd0 | that's what we did in our course actually | 23:07 |
@sonney2k | no boosting I think | 23:07 |
blackburn | n4nd0: why do you think adaboost is so required here? | 23:07 |
@sonney2k | blackburn, it really is very well suited for that task | 23:07 |
blackburn | for haar yes | 23:08 |
n4nd0 | sonney2k: http://books.google.se/books?id=seAgiOfu2EIC&pg=PA508&lpg=PA508&dq=opencv+face+detector+adaboost&source=bl&ots=hSI1dfiGNi&sig=YblAMxB_D1hv5cYSbZoTLS7dtD8&hl=sv&sa=X&ei=l_KSUKakB5DNswb9w4GABg&redir_esc=y#v=onepage&q=opencv%20face%20detector%20adaboost&f=false | 23:08 |
@sonney2k | and other object recognition | 23:08 |
n4nd0 | sorry for that ugly link... | 23:08 |
n4nd0 | but it looks like it is adaboost | 23:08 |
n4nd0 | Viola-Jones | 23:08 |
n4nd0 | blackburn: I think it is a pretty awesome algorithm | 23:09 |
blackburn | n4nd0: dependsssssss | 23:09 |
n4nd0 | blackburn: and also, it is the only thing I have done for face detection :D idk if there are other strategies | 23:09 |
blackburn | even LBP + linear svm works | 23:09 |
blackburn | with HKM even better I think | 23:10 |
@sonney2k | indeed http://www.cognotics.com/opencv/servo_2007_series/part_2/sidebar.html | 23:10 |
@sonney2k | blackburn, but your traffic sign stuff would also be very cool | 23:10 |
n4nd0 | that link was definitely nicer :S | 23:10 |
@sonney2k | blackburn, it would actually be very easy to put it on the website once you put the thing into /applications | 23:10 |
blackburn | sonney2k: draw me how do you see it | 23:11 |
@sonney2k | blackburn, just some image upload button -> name or what do you mean? | 23:12 |
blackburn | sonney2k: n4nd0: [BENCHMARK] Randomized eigendecomposition took 0.57 seconds | 23:12 |
blackburn | sonney2k: n4nd0: 15K x 15K eigenproblem | 23:12 |
blackburn | result is a little approximate though but come on it is linear | 23:13 |
blackburn | sonney2k: I am a good composer of stolen methods :D | 23:13 |
@sonney2k | blackburn, its an art! | 23:14 |
@sonney2k | stealing is believing :) | 23:14 |
@sonney2k | blackburn, http://docs.opencv.org/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.html | 23:16 |
@sonney2k | looks like they have one based on LBP | 23:16 |
blackburn | sonney2k: yes they do | 23:16 |
@sonney2k | probably better these days | 23:16 |
blackburn | LBP is as well good for faces | 23:16 |
blackburn | HOG+LBP is unbeatable :D | 23:17 |
@sonney2k | blackburn, don't tell that yann lecun :D | 23:17 |
blackburn | sonney2k: well I respect his vision that we should learn features not models | 23:18 |
@sonney2k | alright I guess when I find the time I will do some shogun demoing | 23:19 |
@sonney2k | but for now | 23:19 |
@sonney2k | bed time | 23:19 |
@sonney2k | cya | 23:19 |
n4nd0 | night | 23:19 |
n4nd0 | time to sleep here too, good night | 23:24 |
blackburn | n4nd0: night | 23:24 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 23:24 | |
--- Log closed Fri Nov 02 00:00:17 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!