--- Log opened Sat Jul 09 00:00:57 2011 | ||
-!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 00:18 | |
@sonney2k | f-x, yes? | 00:21 |
---|---|---|
@sonney2k | I am too sleepy now g'night. | 00:23 |
f-x | okay, see ya! | 00:29 |
-!- f-x [~user@117.192.202.110] has quit [Remote host closed the connection] | 02:12 | |
-!- in3xes_ [~in3xes@210.212.58.111] has joined #shogun | 05:22 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Read error: Operation timed out] | 05:24 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 10:44 | |
-!- blackburn [~blackburn@188.122.238.99] has quit [Read error: Connection reset by peer] | 11:58 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 12:03 | |
-!- sploving1 [~sploving@2001:cc0:2020:2022:5eff:35ff:fe04:f091] has joined #shogun | 13:15 | |
-!- sploving1 [~sploving@2001:cc0:2020:2022:5eff:35ff:fe04:f091] has quit [Ping timeout: 260 seconds] | 13:40 | |
-!- f-x [~user@117.192.197.89] has joined #shogun | 13:49 | |
-!- f-x [~user@117.192.197.89] has quit [Remote host closed the connection] | 13:59 | |
-!- sploving1 [~sploving@210.77.14.109] has joined #shogun | 14:07 | |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 14:10 | |
-!- f-x [~user@117.192.197.89] has joined #shogun | 14:40 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Quit: Ooops..!!] | 15:01 | |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 15:01 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Client Quit] | 15:02 | |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 15:05 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Remote host closed the connection] | 15:11 | |
-!- in3xes [~in3xes@180.149.49.230] has joined #shogun | 15:13 | |
sploving1 | blackburn, | 15:27 |
blackburn | yes? | 15:27 |
sploving1 | parameter_list.add(Arrays.asList(new Double(0.9), new Double(1e-3), new Integer(1))); | 15:27 |
sploving1 | why int num_threads = ((Integer)para.get(2)).intValue(); said: Exception in thread "main" java.lang.ArrayIndexOutOfBoundsException: 2 | 15:27 |
sploving1 | blackburn, do you know what is the matter? | 15:28 |
blackburn | sploving1: no, looks strange | 15:30 |
sploving1 | blackburn, I got know. my mistake | 15:31 |
sploving1 | blackburn, svm.parallel.set_num_threads(num_threads); | 15:32 |
sploving1 | cannot find variable parallel | 15:32 |
blackburn | no idea | 15:33 |
sploving1 | oh | 15:34 |
-!- in3xes [~in3xes@180.149.49.230] has quit [Quit: Ooops..!!] | 15:45 | |
sploving1 | what does this error mean? [ERROR] Not a two class labeling - no positively labeled examples found | 15:54 |
sploving1 | Exception in thread "main" java.lang.RuntimeException: [ERROR] Not a two class labeling - no positively labeled examples found | 15:54 |
sploving1 | blackburn, sonney2k, | 15:54 |
blackburn | exactly what it says - you have olny one class labelling | 15:55 |
sploving1 | blackburn, svm.set_linear_term(new DoubleMatrix(new double[] {-1,-2,-3,-4,-5,-6,-7,-8,-7,-6})); | 15:55 |
sploving1 | how to fix the above statement? | 15:55 |
blackburn | check labels passed to svm | 15:56 |
sploving1 | blackburn, double label_train_dna[] = {-1,-1,-1,-1,-1,1,1,1,1,1}; | 15:56 |
sploving1 | Labels labels = new Labels(new DoubleMatrix(label_train_dna)); | 15:56 |
sploving1 | maybe this statement is wrong | 15:56 |
blackburn | it should be ok | 15:56 |
sploving1 | so strange | 15:57 |
sploving1 | blackturn, i know. DoubleMatrix(new double[] {value1, value2, ...})Constructs a column vector. | 15:58 |
blackburn | ah | 15:59 |
blackburn | make a row :) | 15:59 |
sploving1 | yeap | 16:00 |
-!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 16:18 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 16:21 | |
-!- f-x [~user@117.192.197.89] has quit [Remote host closed the connection] | 16:25 | |
-!- Guest587 [~mobile@117.192.197.89] has joined #shogun | 16:33 | |
-!- in3xes_ is now known as in3xes | 16:53 | |
sploving1 | hello Math_init_random(17) in python, then in java, how to use it? | 17:13 |
-!- sploving1 [~sploving@210.77.14.109] has left #shogun [] | 17:38 | |
-!- Guest587 [~mobile@117.192.197.89] has quit [Ping timeout: 240 seconds] | 19:16 | |
CIA-32 | shogun: Baozeng Ding master * re34a3b8 / (3 files): add remaining classifier java examples - http://bit.ly/rnrZh3 | 20:35 |
CIA-32 | shogun: Baozeng Ding master * r00b9a3c / (2 files): add clustering java examples - http://bit.ly/ngHumD | 20:35 |
CIA-32 | shogun: Baozeng Ding master * r53b6f91 / examples/undocumented/java_modular/distribution_histogram_modular.java : add distribution_histogram_modular java example - http://bit.ly/nSVFBg | 20:35 |
CIA-32 | shogun: Soeren Sonnenburg master * r08b175b / (6 files): | 20:35 |
CIA-32 | shogun: Merge pull request #184 from sploving/master | 20:35 |
CIA-32 | shogun: add remaining classifier java examples - http://bit.ly/n85Bq4 | 20:35 |
blackburn | sonney2k: I came with new task for me: static memory relaxed fibonacci heap :( | 20:37 |
@sonney2k | blackburn, you dont' like the easy things do you ? | 20:37 |
blackburn | sonney2k: it could be the fastest isomap in shogun ;) | 20:38 |
blackburn | even for really large datasets it could act well in theory | 20:39 |
blackburn | sth like O(n^2 log 1/n) | 20:40 |
@sonney2k | that's the spirit | 20:40 |
blackburn | spirit? | 20:40 |
blackburn | from next week work on cover tree porting should be started | 20:42 |
@sonney2k | all very welcome | 20:42 |
blackburn | chris will be grandmentor (just like father becomes grandfather) | 20:42 |
blackburn | :D | 20:42 |
blackburn | cause I will be some kind of mentor for guy that will try to port cover tree | 20:42 |
@sonney2k | will that turn me into grand-grand-mentor or grand-admin? | 20:43 |
blackburn | yes | 20:44 |
blackburn | Soeren the godfather I guess | 20:45 |
blackburn | why not to play some mafia here :D | 20:45 |
@sonney2k | blackburn, I prefer benevolent dictator - but hey I am no longer alone :) | 22:22 |
blackburn | sonney2k: who are? :) | 22:22 |
@sonney2k | rumor has it - that some mad vodka drinking Russian is launching nuclear missiles too | 22:28 |
blackburn | sonney2k: I'm sure you are ;) | 22:29 |
* sonney2k wonders if blackburn got replaced by some bot | 22:30 | |
blackburn | hehe | 22:30 |
-!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 22:43 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 22:43 | |
-!- f-x [~user@117.192.197.89] has joined #shogun | 22:47 | |
@sonney2k | f-x, hi | 22:48 |
f-x | sonney2k: hey! | 22:48 |
@sonney2k | I am more alive today | 22:48 |
f-x | I too just got back now, read your mail | 22:48 |
f-x | have you observed this kind of behaviour with that dataset before? i mean performance decreases with the last 100k or so examples | 22:49 |
f-x | i'm pretty sure the online SGD and the original are essentially identical, except for lambda and epochs | 22:50 |
@sonney2k | f-x, yeah but I might have made a mistake translating it | 22:50 |
@sonney2k | f-x, so I would suggest to use liblinear | 22:50 |
@sonney2k | for liblinear we have guarantees | 22:50 |
f-x | sonney2k: you mean a version of liblinear using the new framework? | 22:51 |
@sonney2k | f-x, I meant just the batch version | 22:51 |
f-x | ok | 22:52 |
@sonney2k | f-x, there is a script on the website to convert the data files into svmlight format | 22:52 |
@sonney2k | which sparse features can directly read | 22:52 |
f-x | sonney2k: ok thanks, i'll use that. but my computer can't take like more than 1 GB data at once | 22:53 |
@sonney2k | f-x, yeah then just use a subset | 22:53 |
f-x | so I think the examples will be in the 100k-200k range | 22:53 |
f-x | and testing on say 10k examples which were not used for training? | 22:53 |
@sonney2k | f-x, you should get similar results with liblinear | 22:54 |
@sonney2k | f-x, I would test on 100k examples too | 22:54 |
f-x | sonney2k: ah. okay.. btw how was testing done for the alpha dataset? | 22:54 |
f-x | did you have some extra labelled data with you? | 22:55 |
@sonney2k | f-x, labels are non-public | 22:55 |
@sonney2k | so use say first 100-200k examples for traiing | 22:55 |
@sonney2k | and the next 100k examples for testing | 22:55 |
f-x | was just asking :) sure.. | 22:56 |
@sonney2k | f-x, but shouldn't make a difference | 22:56 |
f-x | sonney2k: yeah.. i'll get on with that.. anything else you think i should be working on in the meantime? | 22:57 |
@sonney2k | f-x, I mean the next steps would be to integrate liblinear/sgd-qn/vw | 22:58 |
@sonney2k | of course getting this benchmark to work and speed optimizations | 22:58 |
@sonney2k | f-x, sgd-qn should be trivial to do for you (very similar to sgd) | 22:59 |
f-x | yes - i browsed through the code, and they looked pretty similar | 22:59 |
@sonney2k | f-x, regarding liblinear I would only translate the CLibLinear::solve_l2r_l1l2_svc function | 23:00 |
-!- phrhbk [~asdfasdf@host-57-20.warwick.ac.uk] has quit [Quit: Leaving] | 23:00 | |
@sonney2k | and this time provide an OnlineLibLinear in addition to the batch variant | 23:01 |
f-x | sonney2k: yes.. just like we did for sgd | 23:01 |
@sonney2k | f-x, for sgd we will only keep the online variant at some point | 23:02 |
@sonney2k | but liblinear has many solvers and the svm one is sufficient to us | 23:02 |
f-x | okay.. for now i'll do the testing with liblinear and begin with OnlineLibLinear | 23:03 |
@sonney2k | yeah these online learners won't be a lot of work - but it will be more interesting and the benchmarking will actually tell us whether things are correct and what we can improve | 23:05 |
blackburn | another day another wrapper | 23:10 |
blackburn | sonney2k: grant me rights to execute CustomDistance ;) | 23:26 |
bettyboo | HA | 23:26 |
@sonney2k | blackburn, ? | 23:26 |
blackburn | sonney2k: I use it in landmark mds - I pass landmark distance matrix using it - but it creates a copy of matrix | 23:26 |
blackburn | because it is only float32_t and I use float64_t | 23:26 |
@sonney2k | blackburn, of course it is float32_t | 23:27 |
@sonney2k | it tries to be efficient | 23:27 |
blackburn | and produces necessarity of copy matrix.. | 23:27 |
@sonney2k | I mean when you have to store the matrix you should try to be | 23:27 |
@sonney2k | well then don't get the copy but the float32_t version | 23:27 |
blackburn | distance always returns float64_t? | 23:28 |
blackburn | btw I guess I should use triangular matrices | 23:29 |
-!- f-x [~user@117.192.197.89] has left #shogun ["ERC Version 5.3 (IRC client for Emacs)"] | 23:30 | |
-!- f-x [~user@117.192.197.89] has joined #shogun | 23:30 | |
blackburn | shit I've got the most beautiful embedding I ever seen | 23:31 |
blackburn | http://imageshack.us/photo/my-images/718/imagepw.png/ | 23:31 |
@sonney2k | :) | 23:39 |
@sonney2k | blackburn, yes distcance returns float64_t /scalars/ | 23:40 |
@sonney2k | but the matrix itself is squeezed as float32... | 23:40 |
blackburn | phew | 23:40 |
blackburn | don't like it a bit | 23:40 |
@sonney2k | I guess I will never understand | 23:41 |
blackburn | what? | 23:41 |
@sonney2k | what your problem with this is | 23:42 |
blackburn | I get the whole matrix | 23:42 |
@sonney2k | and? | 23:42 |
@sonney2k | just get the float32 one | 23:42 |
blackburn | how to get float32 one? | 23:42 |
@sonney2k | blackburn, I see - it is not templated like in Kernel.h | 23:43 |
@sonney2k | basically, the distance class evolved from the kernel class | 23:44 |
@sonney2k | the kernel class has that function | 23:44 |
@sonney2k | the distance class has not | 23:44 |
@sonney2k | uh | 23:45 |
blackburn | aha | 23:45 |
@sonney2k | and I see a new problem | 23:45 |
@sonney2k | we do return some SGMatrix<xx> | 23:45 |
blackburn | I have found a way to slow down LLE 20x | 23:46 |
@sonney2k | but is it possbile to have two functions with same name and no input args but different output args? | 23:46 |
blackburn | no | 23:46 |
blackburn | I guess :) | 23:46 |
bettyboo | weird | 23:46 |
blackburn | how can compiler know what do you want? :) | 23:47 |
@sonney2k | blackburn, it could - I mean it knows the output type | 23:47 |
@sonney2k | and that should match | 23:47 |
blackburn | ehh. if you call foo() how can compiler know you want int foo()? | 23:49 |
@sonney2k | blackburn, well if you do int x = foo() it is clear | 23:54 |
blackburn | sonney2k: I'm not sure it is clear for gcc ;) | 23:55 |
@sonney2k | blackburn, the problem is that one can drop return arguments | 23:55 |
@sonney2k | i.e. call a function ignoring the result | 23:55 |
blackburn | get distance matrix and lost result | 23:56 |
blackburn | hmm pretty way to get garbage | 23:56 |
blackburn | :) | 23:56 |
--- Log closed Sun Jul 10 00:00:59 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!