--- Log opened Sat Apr 16 00:00:36 2011 | ||
alesis-novik | 4 pints later, I return | 02:33 |
---|---|---|
serialhex | niiice, how are things alesis-novik? | 02:38 |
* serialhex thinks of pippin from LotR every time he hears 'pint' | 02:39 | |
alesis-novik | Good so far, going to fix some other things in the pull request | 02:39 |
serialhex | cool cool | 02:39 |
* serialhex has been lazy these past few days | 02:40 | |
alesis-novik | There... Now I can think about starting working on 1 dimensional EM | 02:43 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 03:08 | |
alesis-novik | Right, new plan: go to sleep relatively early today and get up relatively early tomorrow to do work | 04:07 |
serialhex | :D that'll work alesis-novik! | 04:08 |
@bettyboo | ;> | 04:08 |
serialhex | 'nite! | 04:09 |
alesis-novik | goodnight :) | 04:12 |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 05:48 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 05:49 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Remote host closed the connection] | 05:50 | |
-!- serialhex [~quassel@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 05:53 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Read error: Connection reset by peer] | 06:19 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 06:22 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 250 seconds] | 06:37 | |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 06:52 | |
siddharth | anyone here? | 07:32 |
@sonney2k | dammed, took the whole night | 07:57 |
-!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 260 seconds] | 08:10 | |
-!- blackburn [~qdrgsm@188.122.237.148] has joined #shogun | 09:17 | |
@mlsec | sonney2k: No sleep? | 09:48 |
@mlsec | Hmm | 09:49 |
@mlsec | bettyboo: did you talk to sonney2k? | 09:49 |
@bettyboo | mlsec: when you told me to integrate shogun with the parser yesterday, what things about shogun did you have in mind? | 09:49 |
blackburn | bettyboo: make doc | 09:49 |
@bettyboo | blackburn: qdrgsm@blackburn-R519:~/Documents/GSoC-SHOGUN/shogun_myfork/shogun/examples/undocumented/python_modular$ python kernel_exponential_modular.py | 09:50 |
blackburn | :D | 09:50 |
@mlsec | That's your code. Nice catch | 09:50 |
blackburn | I see :) | 09:51 |
blackburn | bettyboo: make doc | 09:51 |
@bettyboo | blackburn: (only the .h file needs to be documented) | 09:51 |
@mlsec | Surprisingly good. | 09:52 |
@mlsec | Breakfast time!! | 09:53 |
blackburn | hm, that's a good idea | 09:55 |
-!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 10:14 | |
@sonney2k | blackburn, ok... | 11:27 |
blackburn | sonney2k: ok abouuut? | 11:27 |
@sonney2k | about perf measures :) | 11:30 |
CIA-110 | shogun: Alesis Novik master * rc702400 / src/libshogun/distributions/Gaussian.cpp : Fixed - http://bit.ly/eWZZjO | 11:32 |
CIA-110 | shogun: Soeren Sonnenburg master * r6e6999c / (5 files in 3 dirs): Merge branch 'master' of https://github.com/alesis/shogun - http://bit.ly/gMtBn0 | 11:32 |
@sonney2k | alesis-novik, I applied you patches - next time please use more meaningful commit msg | 12:05 |
@sonney2k | I mean 'fixed' doesn't really tell anyone (besides me *today*) what you did :) | 12:05 |
blackburn | Commit: 31434 files, description "Fixed some performance issues" :D | 12:08 |
-!- blackburn [~qdrgsm@188.122.237.148] has quit [Ping timeout: 276 seconds] | 12:14 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 12:16 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 12:18 | |
-!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun | 12:18 | |
blackburn | rghhh | 12:19 |
Tanmoy | @sonney2k do we have an implementation of Bessel function ? | 12:25 |
siddharth | nope | 12:33 |
Tanmoy | maybe need to implement it from GSL ? | 12:35 |
Tanmoy | even LAPACK does not have it | 12:35 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 13:16 | |
-!- yayo3 [~jakub@ip-78-45-113-245.net.upcbroadband.cz] has joined #shogun | 13:23 | |
yayo3 | hey | 13:23 |
yayo3 | how does disabling of features that require LAPACK when there's none work? | 13:33 |
yayo3 | I can't seem to find it | 13:36 |
yayo3 | but it needs to be done for the Gaussian distribution, because current master won't compile on Mac OS at all (there's no lapack and stuff) | 13:38 |
yayo3 | but there's aparently BLAS and LAPACK integrated in Mac OS, so I'm diving into configure once more :) | 13:47 |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 14:16 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 14:42 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 14:42 | |
@sonney2k | yayo3, where does it fail - in Gaussian.cpp ? | 14:43 |
@sonney2k | if so ... git pull | 14:44 |
CIA-110 | shogun: Soeren Sonnenburg master * rebf4cf6 / (2 files): Add HAVE_LAPACK checks to Gaussian distribution - http://bit.ly/ii0qrA | 14:44 |
yayo3 | yep. thanks. | 14:46 |
* sonney2k takes a nap | 14:48 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 252 seconds] | 14:49 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 15:17 | |
dvevre | sonney2k: i've made a StreamingFeatures class implementing the get_next_feature_vector method.. haven't modified the I/O back-end much, | 15:21 |
dvevre | can i mail it to you for opinions on the implementation? | 15:22 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 15:28 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 15:30 | |
dvevre | sonney2k: mail sent | 15:32 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 15:41 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 15:47 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Client Quit] | 15:49 | |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 16:00 | |
yayo3 | sonney2k: there's still a problem with tha guassian stuff. it uses function that's only in ATLAS | 16:00 |
yayo3 | clapack_dpotri to be exact. there's clapack_dpotrf reimplemented in shogun but not this one | 16:01 |
yayo3 | so it won't compile on computers without ATLAS. While Mac OS have LAPACK integrated and I'm working on using it, there's no ATLAS here | 16:03 |
-!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Client Quit] | 16:04 | |
alesis-novik | Morning people and bots | 16:08 |
yayo3 | hi | 16:08 |
alesis-novik | yayo3, I'll look into that | 16:08 |
yayo3 | alesis that guassian stuff is your, right? | 16:08 |
yayo3 | oh, you saw. good :) | 16:08 |
yayo3 | anyway good morning :) | 16:08 |
siddharth | hi all | 16:21 |
siddharth | Is there any way to access the elements of an 'i'th index feature vector? | 16:22 |
alesis-novik | what class for features are you using siddharth ? | 16:23 |
siddharth | Cdotfeatures | 16:23 |
siddharth | i was trying get_feature_vector()...but getting segmentation fault :| | 16:24 |
alesis-novik | then you might be trying to get a feature that isn't there | 16:24 |
siddharth | nope features are there | 16:25 |
siddharth | w8 let me see once again if I am making some mistake | 16:27 |
siddharth | alesis-novik, I am doing this | 16:34 |
siddharth | w_dim=features->get_dim_feature_space(); | 16:34 |
siddharth | int32_t num_vec=features->get_num_vectors(); | 16:34 |
siddharth | for (int32_t i=0; i<num_vec; i++) | 16:34 |
siddharth | { | 16:34 |
siddharth | features->get_feature_vector(dst,alen,i); | 16:34 |
siddharth | output of num_vec == 3 | 16:35 |
siddharth | so the features exist | 16:35 |
alesis-novik | what are dst and alen? | 16:35 |
siddharth | float64_t** dst; to store the feature vector | 16:36 |
siddharth | copy | 16:36 |
siddharth | alen==length of the feature vector | 16:37 |
alesis-novik | alen is int32_t* ? | 16:37 |
siddharth | yes | 16:37 |
siddharth | http://www.shogun-toolbox.org/doc/classshogun_1_1CDotFeatures.html#a6c74837680cf15810c9b989a6908de09 | 16:38 |
alesis-novik | and were does it break? | 16:38 |
alesis-novik | does the code actually break at get_feature_vector or somewhere further? | 16:38 |
siddharth | further I am doing some more calculations | 16:39 |
siddharth | but i can break it | 16:39 |
alesis-novik | can you post the code in pastebin or somewhere? | 16:39 |
siddharth | ok w8 | 16:39 |
siddharth | alesis-novik, maybe u can try urself...just add 'features->get_feature_vector()' in SVMSGD.cpp in classifier/svm in the train function | 16:46 |
siddharth | if u have time | 16:47 |
alesis-novik | wait, did you post the code anywhere? | 16:49 |
siddharth | its a lengthy code...I am trying myself in this code SVMSGD.cpp | 16:49 |
alesis-novik | because I'm using get_feature_vector() in my code, but it doesn't break, so just randomly adding it in other places won't really achieve anything | 16:50 |
siddharth | yes but you will get segmentation fault while running this classifier after adding these lines | 16:51 |
siddharth | you can run the code from here... examples/documented/libshogun/classifier_minimal_svm.cpp .In this just change the classifier name from LibSVM to SVMSGD | 16:53 |
siddharth | or you can run from python interface | 16:58 |
alesis-novik | siddharth, can you show me what you do with dst and alen after you get them? | 17:00 |
siddharth | alesis-novik, I am not using them as of now... | 17:00 |
alesis-novik | so just adding get_feature_vector produces seg-faults? | 17:01 |
siddharth | yeah | 17:01 |
siddharth | alesis-novik,you can run the code from static/modular interface | 17:04 |
siddharth | alesis-novik, I will be back in 10 min...have to go | 17:05 |
* siddharth brb | 17:05 | |
yayo3 | alesis-novik: what does clapack_dpotri() do? | 17:07 |
alesis-novik | computes the inverse of a matrix using stuff from dpotrf | 17:08 |
alesis-novik | if done a small patch that might help | 17:09 |
siddharth | alesis-novik, any findings? | 17:27 |
alesis-novik | siddharth, as far as I can tell without actually testing it, it shouldn't cause any problems... | 17:27 |
siddharth | not able to find the error | 17:28 |
siddharth | alesis-novik, I have posted the code..u can have a look at that | 17:47 |
siddharth | http://codepad.org/BkqLEVvk | 17:47 |
siddharth | i have also marked where I have done the changes | 17:47 |
siddharth | /CHANGES------------------------ | 17:47 |
alesis-novik | Will take a look now siddharth | 17:52 |
alesis-novik | ah | 18:08 |
alesis-novik | right | 18:08 |
alesis-novik | You should probably change | 18:08 |
alesis-novik | float64_t** dst; | 18:08 |
alesis-novik | int32_t* alen; | 18:08 |
alesis-novik | features->get_feature_vector(dst,alen,i); | 18:08 |
alesis-novik | to | 18:09 |
alesis-novik | float64_t* dst; | 18:09 |
alesis-novik | int32_t alen; | 18:09 |
alesis-novik | features->get_feature_vector(&dst,&alen,i); | 18:09 |
siddharth | alesis-novik, hmm let me try | 18:09 |
alesis-novik | because get_feature_vector does *len = dim, which if no memory is allocated, gets you your faults | 18:10 |
siddharth | ohh did not see the function | 18:11 |
siddharth | ya solved | 18:11 |
siddharth | thanks :) | 18:11 |
alesis-novik | sonney2k, I've made a new pull request. I think I might have done some stuff you already added, not too sure. | 18:15 |
-!- dvevre [b49531e5@gateway/web/freenode/ip.180.149.49.229] has joined #shogun | 18:24 | |
alesis-novik | There, fixed some of the merging issues, should be fine now. Main points of interest are changes in lapack.h/.cpp and PCACut.cpp | 18:28 |
yayo3 | aaaaaaaaaaaaand we have BLAS and LAPACK Mac OS support :) | 18:30 |
* yayo3 's gotta run | 18:31 | |
-!- yayo3 [~jakub@ip-78-45-113-245.net.upcbroadband.cz] has quit [Quit: leaving] | 18:31 | |
-!- dvevre [b49531e5@gateway/web/freenode/ip.180.149.49.229] has quit [Ping timeout: 252 seconds] | 18:35 | |
-!- siddharth [~siddharth@117.211.88.150] has quit [Remote host closed the connection] | 18:56 | |
-!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 19:42 | |
blackburn | sonney2k: here? | 20:50 |
CIA-110 | shogun: Jakub Jirku master * r0911622 / (src/configure src/libshogun/lib/lapack.h): Mac OS BLAS and LAPACK support - http://bit.ly/eH6SsL | 20:50 |
CIA-110 | shogun: Jakub Jirku master * rc8026b1 / src/configure : Configure fix - http://bit.ly/egj0f1 | 20:51 |
@sonney2k | blackburn, yes | 20:52 |
blackburn | sonney2k: have time for discuss some issues with 'evaluation'? | 20:52 |
@sonney2k | lets try | 20:59 |
blackburn | so, we could have some 'Evaluation.h' with virtual evaluate(), right? | 21:00 |
CIA-110 | shogun: Soeren Sonnenburg master * r9988d90 / (2 files): rename member variables to start with m_ to fix warnings - http://bit.ly/eFEH5p | 21:00 |
blackburn | CEvaluation, sorry | 21:00 |
blackburn | and there could be things like in mldata-utils: CRegressionEvaluate, CClassificationEvaluate, inherited from CEvaluate | 21:01 |
blackburn | evaluate() there would be something like init: it computes some things and then one can compute ROC or mean absolute error | 21:02 |
blackburn | sonney2k: am I going in right direction? | 21:04 |
@sonney2k | sonney2k, or just CAccuracy derived from CEvaluation | 21:05 |
@sonney2k | blackburn, there a two ways: Either we have one class CEvaluation which gets two label objects as input | 21:06 |
@sonney2k | and then all the different evaluation functions | 21:06 |
blackburn | sonney2k: why CAccuracy? both for regression and classification? | 21:06 |
@sonney2k | or various evaluation classes | 21:06 |
@sonney2k | blackburn, accuracy is only defined for Classification. | 21:07 |
@sonney2k | MSE would be for regression... | 21:07 |
blackburn | sonney2k: ah, I mean it, sorry for misunderstanding terms :) | 21:07 |
blackburn | sonney2k: so, classes for evaluating these types will be specified | 21:10 |
blackburn | sonney2k: and I could implement for clustering too | 21:10 |
serialhex | bettyboos next project: http://xkcd.com/329/ | 21:16 |
@bettyboo | serialhex: I guess you need to assume that the number of features is known and that you have to read that many bytes or so to get the next vector | 21:16 |
@sonney2k | yes | 21:16 |
@sonney2k | blackburn, I don't know what do you prefer a evaluation class and derived classes or jsut one class | 21:17 |
blackburn | sonney2k: derived classes seems to be better | 21:17 |
blackburn | sonney2k: what way do you prefer? | 21:18 |
@sonney2k | blackburn, then one has to assume that each method returns a label object as outputs | 21:18 |
@sonney2k | and then one could compare the true label object with the predicted one | 21:18 |
@sonney2k | sounds good to me | 21:18 |
blackburn | sonney2k: each method of ..? | 21:19 |
@sonney2k | in case labels become to complex we have to have more than the current label object - but an interface | 21:19 |
@sonney2k | method == SVM etc | 21:19 |
blackburn | eh, but thought evaluation works only with labels | 21:20 |
blackburn | for example, when classifying, we just do some e.evalute(labels,right_labels) and after that can do e.ROC() or etc | 21:21 |
@sonney2k | serialhex, heh :) | 21:21 |
@sonney2k | blackburn, yes | 21:22 |
@sonney2k | but the predicted labels come from a method right | 21:22 |
blackburn | labels = SVM.classify() and then e.evaluate(labels, right_labels)? | 21:23 |
@sonney2k | exactly | 21:23 |
blackburn | what you mean saying labels become to complex? | 21:23 |
@sonney2k | blackburn, e.g. a graph, or some other thing | 21:25 |
blackburn | don't sure I understand what is complex labels.. | 21:25 |
blackburn | ah | 21:25 |
blackburn | sonney2k: how could it be? | 21:25 |
@sonney2k | blackburn, structured output prediction... predict a graph | 21:25 |
blackburn | so in that case there will not any 'evaluate()' in CEvaluation | 21:25 |
blackburn | all things will be overloaded | 21:26 |
blackburn | in derived classes | 21:26 |
@sonney2k | blackburn, well then we simply need a StructuredLabel object derived from CLables | 21:26 |
blackburn | sonney2k: well, but what should we do for clustering? | 21:27 |
@sonney2k | ehh no idea yet | 21:28 |
blackburn | we should use objects as labels | 21:28 |
blackburn | and may be we will even need CDistance | 21:29 |
@sonney2k | actually a clustering is just some integer class assignment | 21:29 |
@sonney2k | so some integer vector is sufficient | 21:29 |
@sonney2k | no need to do anything then | 21:29 |
blackburn | hm. okay, I'm just about some internal cluster distances and etc | 21:30 |
@sonney2k | blackburn, but doesn't matter right? I mean that is in the perfmeasure then | 21:30 |
blackburn | ah, okay, will implement it in this way ;) | 21:31 |
blackburn | sonney2k: I will make a pull request with draft design (CEvaluation - CClassifyEvaluation, CRegressionEvaluation, ...) - for discuss methods to be implemented, okay? | 21:32 |
@sonney2k | blackburn, I would only do CEvaluation and derive directly from there | 21:33 |
@sonney2k | e.g. CEvalution, CAccuracy, CBalancedError, CMeanSquaredError etc... | 21:33 |
blackburn | is CEvaluation base class? | 21:34 |
@sonney2k | yes - the interface | 21:34 |
blackburn | what is BalancedError? | 21:34 |
@sonney2k | BAL | 21:35 |
blackburn | anyway now I will just port things available in mldata-utils | 21:36 |
blackburn | sorry, but I don't know what it is :( | 21:36 |
@sonney2k | blackburn, ahh then look in mldata-utils - I documented all functions in there | 21:36 |
@sonney2k | and BAL is there too | 21:36 |
blackburn | sonney2k: ok, thank you | 21:36 |
@sonney2k | ok ... time for dinner | 21:37 |
blackburn | enjoy your meal | 21:37 |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 21:40 | |
Tanmoy | ok when i am trying to push it says Oops! The shogun-toolbox:master branch is already up-to-date with Tankiit:master — maybe you want to try something else? | 21:52 |
blackburn | did your set up your git properly? | 21:55 |
blackburn | btw what are you doing exactly, pushing your commits to your fork? | 21:57 |
Tanmoy | i think so the setup is fine | 21:57 |
Tanmoy | yup | 21:57 |
alesis-novik | Evening folks | 21:57 |
Tanmoy | evening | 21:57 |
blackburn | but for some reason it says sth about 'shogun-toolbox', looks like you try to commit in main repo.. | 21:58 |
blackburn | alesis-novik: evening | 21:58 |
alesis-novik | sonney2k, around? | 22:01 |
blackburn | alesis-novik: (11:37:12 PM) sonney2k: ok ... time for dinner | 22:01 |
alesis-novik | ok then | 22:02 |
Tanmoy | hmm well i forked it made changes | 22:02 |
Tanmoy | committed | 22:02 |
Tanmoy | and then tried pushing | 22:02 |
Tanmoy | i think i did what is reqd | 22:02 |
blackburn | did you clone your fork? | 22:02 |
Tanmoy | yup | 22:03 |
blackburn | then don't know :) | 22:03 |
Tanmoy | this is what push is doing | 22:03 |
Tanmoy | ssh: connect to host hithub.com port 22: Connection timed out | 22:03 |
blackburn | hithub? | 22:03 |
Tanmoy | yeah | 22:04 |
Tanmoy | : | 22:04 |
Tanmoy | ok got it now | 22:04 |
blackburn | what is hithub :D | 22:04 |
Tanmoy | typo | 22:05 |
Tanmoy | ok done | 22:07 |
Tanmoy | thanks blackburn | 22:07 |
blackburn | np, Tanmoy | 22:07 |
Tanmoy | @blackburn hows the graph library coming up | 22:10 |
blackburn | Tanmoy: now working on CEvaluation class and it's derivates | 22:10 |
blackburn | will you need it soon? | 22:10 |
Tanmoy | nope had some ideas | 22:11 |
Tanmoy | r u working on it | 22:11 |
Tanmoy | or u planing to merge some library | 22:11 |
blackburn | nope, think will implement some 'compact' graph class | 22:11 |
blackburn | anyway we don't need serious graph library | 22:12 |
Tanmoy | hmm modules of finding_neigbors,BFS, DFS shld do | 22:13 |
blackburn | it seems to be a minimal graph 'library' | 22:13 |
Tanmoy | ok gr8 | 22:15 |
blackburn | think I will work on it during next week | 22:16 |
alesis-novik | Right, while I'm waiting for comments on my pull request, I might as well start working on the next thing | 22:18 |
blackburn | alesis-novik: just not commit next thing before pull will be closed :) | 22:18 |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 22:19 | |
blackburn | seems many candidates are flunked by themselves :) | 22:20 |
@bettyboo | ;D | 22:20 |
blackburn | only 5-6-7 ones are active now | 22:21 |
alesis-novik | That's why I decided to only apply to 1 project, so I can focus on it instead of spreading out and really not doing anything at all | 22:22 |
serialhex | well if there's only 5 the i guess they 'win' by default :D | 22:22 |
blackburn | alesis-novik: me too | 22:23 |
blackburn | serialhex: that will be fine :) | 22:23 |
serialhex | i agree blackburn!! | 22:24 |
serialhex | i take it tanmoy got his github troubles figured out? | 22:24 |
serialhex | FFS!!! i hate stinging bugs!!! | 22:24 |
blackburn | serialhex: yeap, he fixed his 'hithub' | 22:24 |
serialhex | :P cool, i was having a bunch of troubles with it before, but now i've got the hang of it... i'm glad it's fixed | 22:25 |
serialhex | ~43,608,608.5494047km until spring classes end :D | 22:26 |
blackburn | was ist das? | 22:27 |
serialhex | ~5,460,804.968458357km until shogun candidates are announced :D | 22:28 |
serialhex | how far we have to travel before these events happen :P | 22:28 |
blackburn | with 'c' speed? :) | 22:28 |
alesis-novik | I wonder how many applicants have submitted anything so far... | 22:28 |
serialhex | though i dont know if in russia you use a "," or "." for thousands, millions & such... | 22:28 |
blackburn | alesis-novik: submitted what? | 22:29 |
alesis-novik | a patch or any code | 22:29 |
blackburn | alesis-novik: not so many, only 1/3 of all candidates | 22:29 |
serialhex | we can count: https://github.com/shogun-toolbox/shogun/commits/master | 22:29 |
blackburn | serialhex: we don't use for sth like 1,000 but use ',' for 10,0 | 22:30 |
blackburn | https://github.com/shogun-toolbox/shogun/contributors | 22:30 |
serialhex | i think less... sonney2k said there were 60 applicants and i'm not sure there were 20 ppl | 22:30 |
serialhex | ahh, even better :D | 22:30 |
blackburn | I really laugh always when see 'invalid-email-address (You didn't set your git config correctly)' | 22:30 |
blackburn | the most funny contributor | 22:31 |
serialhex | who is that??? :P | 22:31 |
blackburn | some spanish guy, who fucked up his git | 22:31 |
serialhex | niiice | 22:32 |
alesis-novik | :) | 22:32 |
@bettyboo | ;D | 22:32 |
alesis-novik | Even if there were 60 applicants, I assume they won't really consider people who haven't submitted anything... | 22:32 |
serialhex | anyway blackburn, i made a calculator that figure out how far you travel in a given amount of time. | 22:32 |
serialhex | probably not al | 22:32 |
serialhex | it's here: https://github.com/serialhex/fun-stuff/blob/master/spaceship_earth.rb | 22:33 |
blackburn | alesis-novik: at least some applicant contributors don't do anything now | 22:33 |
blackburn | alesis-novik: more than half of them implement some kernel at first :D | 22:34 |
blackburn | implemented* | 22:34 |
serialhex | it takes the time in the future and optionally your latitue and the time you want "now" to be based of of, and calculates how far you travel (through space) | 22:34 |
serialhex | yeah, and theres only like 4 of us who communicate on the channel... | 22:35 |
alesis-novik | serialhex, I really like google calculator, it can compute complex things like (number of horns on a unicorn/the answer to life, the universe and everything) | 22:35 |
serialhex | lol i know, it rocks! | 22:35 |
blackburn | serialhex: may be a little more, 5-6 | 22:35 |
blackburn | let's count :D | 22:35 |
serialhex | something around there :P | 22:35 |
@bettyboo | hihi? | 22:35 |
serialhex | hihi bettyboo!! | 22:36 |
@bettyboo | serialhex: >>> x=Labels() | 22:36 |
blackburn | serialhex, alesis-novik, I, heiko, yayo3, siddharth, dvevre | 22:36 |
blackburn | seems to be 7 | 22:36 |
@sonney2k | alesis-novik, I still don't get your follow up pull requests btw | 22:42 |
@sonney2k | alesis-novik, did you rebase your commit against master? | 22:42 |
serialhex | yeah, alesis-novik probably made the same mistake i did... | 22:44 |
@sonney2k | alesis-novik, would you mind putting the gaussian pdf formula in gaussian.h? | 22:45 |
alesis-novik | sonney2k, I did, under compute PDF function | 22:45 |
serialhex | i was using the fork queue on github to keep my fork up to date, which added all those other commits. | 22:45 |
@sonney2k | alesis-novik, I would prefer if you put it in the class description | 22:46 |
alesis-novik | sonney2k, should I remove it from computePDF or keep it in both places? | 22:47 |
@sonney2k | remove from computepdf (it is enough to state that it computes the gaussian pdf as described in class descr) | 22:47 |
serialhex | alesis-novik: if that is what you did you can rebase your fork on your computer & when it asks you to change stuff simply --skip a bunch of times (assuming that it's up-to-date on your computer) and that should fix the problem | 22:47 |
@sonney2k | alesis-novik, don't have time to look at your other patch in detail now - looks good but I have to have a closer look. | 22:48 |
@sonney2k | thanks! | 22:48 |
blackburn | sonney2k: get_accuracy() or accuracy()? | 22:49 |
@sonney2k | blackburn, err multiple classes or one? | 22:49 |
blackburn | sonney2k: binary classification | 22:49 |
@sonney2k | c++ classes that is | 22:50 |
blackburn | a. | 22:50 |
blackburn | CAccuracy :) | 22:50 |
blackburn | ah! | 22:50 |
blackburn | you mean the way I implement it? | 22:51 |
blackburn | CEvaluation and derivates: CAccuracy, ... | 22:51 |
alesis-novik | sonney2k, I added the formula to the Class description as well | 22:51 |
blackburn | sonney2k: so how better, get_accuracy() or accuracy(), get_BAL() or BAL()? | 22:52 |
CIA-110 | shogun: Soeren Sonnenburg master * r1ee5185 / (4 files in 2 dirs): Merge branch 'master' of https://github.com/kisa12012/shogun - http://bit.ly/eHg5Qw | 22:52 |
@sonney2k | blackburn, I thought you are doing several classes. And then only modify the evaluate function? | 22:54 |
blackburn | sonney2k: we don't decided how 'several' it could be.. | 22:54 |
blackburn | sonney2k: should it be a class for BAL, a class for error, etc? | 22:55 |
blackburn | sonney2k: just say how often do you use multiple evaluations same time | 22:56 |
blackburn | i.e. if multiple of them often get used same time - it is better to do one 'big' class for classify | 22:57 |
blackburn | oh, I see now how I raveled: I tried to implement BAL in CAccuracy :D | 22:58 |
@sonney2k | blackburn, I think one only ever uses one. | 22:58 |
@sonney2k | heh | 22:59 |
@sonney2k | I am too tired now | 22:59 |
@sonney2k | sorry | 22:59 |
@sonney2k | cu tomorrow | 22:59 |
blackburn | sonney2k: thank you, you helped me ;) | 22:59 |
blackburn | see you | 22:59 |
alesis-novik | see you | 22:59 |
--- Log closed Sun Apr 17 00:00:36 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!