--- Log opened Sat Oct 29 00:00:10 2011 | ||
CIA-3 | shogun: Soeren Sonnenburg master * re9c8ecd / src/interfaces/lua_modular/swig_typemaps.i : fix an error in the lua typemaps (checking for the typecode without using strcmp) - http://git.io/PIgx2g | 00:30 |
---|---|---|
-!- BrianC_ [43dc92a2@gateway/web/freenode/ip.67.220.146.162] has quit [Ping timeout: 265 seconds] | 00:39 | |
-!- blackburn [~blackburn@188.168.2.187] has joined #shogun | 12:32 | |
blackburn | sonney2k: wow you did a lot of fixes | 12:34 |
-!- blackburn [~blackburn@188.168.2.187] has quit [Read error: Operation timed out] | 18:32 | |
-!- blackburn [~blackburn@188.168.4.216] has joined #shogun | 21:46 | |
-!- blackburn1 [~blackburn@188.168.5.175] has joined #shogun | 22:03 | |
-!- blackburn [~blackburn@188.168.4.216] has quit [Ping timeout: 258 seconds] | 22:04 | |
supernaught | sonney2k: When you say "use the signum of these numbers, e.g. if the sign is negative then it is a letter from group A, if positve then from group B" | 22:05 |
supernaught | how do I feed in a number for which training has not been done? | 22:06 |
supernaught | e.g. In the program, one creates points {0,1,2} with labels {-1,1,-1} | 22:06 |
supernaught | but when I attempt to extrapolate from my inputs and determine which group {3} belongs to, I pull an exception "terminate called after throwing an instance of 'shogun::ShogunException'" | 22:11 |
blackburn1 | supernaught: are you doing multi-class classification? | 22:11 |
supernaught | I believe not | 22:12 |
blackburn1 | how many letter groups you have? | 22:12 |
supernaught | I'm working with third one from top at http://www.shogun-toolbox.org/doc/en/current/developer_tutorial.html | 22:12 |
supernaught | I'm not sure what a letter group is | 22:12 |
blackburn1 | well you said group A, group B, ..? | 22:13 |
supernaught | oh, sure | 22:13 |
blackburn1 | and there are only A and B, right? | 22:13 |
supernaught | 2 | 22:13 |
supernaught | +1, -1 | 22:13 |
supernaught | yes | 22:13 |
blackburn1 | and you set up labels with +1 and -1 | 22:13 |
blackburn1 | and got exception? | 22:14 |
supernaught | yes | 22:14 |
supernaught | the example is nearly verbatim what I have | 22:14 |
supernaught | except when I go to "classify over training variables" | 22:14 |
supernaught | I loop from 0 to 4 instead of 0 to 3 | 22:15 |
blackburn1 | can you send the example reproducing the error to me? | 22:15 |
supernaught | sure | 22:15 |
blackburn1 | I would check it and fix then | 22:15 |
blackburn1 | or say what is wrong | 22:15 |
supernaught | let me paste it somewhere | 22:15 |
blackburn1 | just send it by email | 22:16 |
blackburn1 | to blackburn91@gmail.com | 22:16 |
-!- blackburn1 is now known as blackburn | 22:16 | |
blackburn | supernaught: btw what are the features you do use? | 22:17 |
supernaught | https://gist.github.com/1325029 | 22:17 |
blackburn | ah sure gist works too | 22:17 |
supernaught | I'm a complete newb to machine learning | 22:18 |
supernaught | I'm not sure what you mean by "features" ? | 22:18 |
blackburn | hmm for example simple features is just dense vectors | 22:18 |
blackburn | [1,2,3,4] | 22:18 |
supernaught | whoops, I forgot the error | 22:19 |
blackburn | and feature is 1 and 2 and 3 and 4 | 22:19 |
blackburn | then [1,2,3,4] is a feature vector | 22:19 |
blackburn | and matrix containing feature vector is a feature matrix | 22:19 |
supernaught | ok | 22:19 |
blackburn | but sometimes feature vector is just a string ("some string") | 22:20 |
blackburn | and there is no feature matrix | 22:20 |
supernaught | Ok: so a feature vector is an input which is used for training? | 22:20 |
blackburn | yes, it is vector representing some object | 22:21 |
supernaught | ok. Lingo :) | 22:21 |
blackburn | something like that :) | 22:21 |
blackburn | okay | 22:21 |
supernaught | so in this case, I am using a feature vector of [0,1,2,3,4,5] | 22:21 |
blackburn | so do you understand what are train set and test set? | 22:21 |
blackburn | no | 22:21 |
blackburn | [[0,2,4], | 22:22 |
blackburn | [1,3,5]] | 22:22 |
blackburn | it is a feature matrix | 22:22 |
supernaught | oh, right | 22:22 |
blackburn | [0,1] [2,3] [4,5] is a feature vectors | 22:22 |
supernaught | and the fortran-style indexing | 22:22 |
supernaught | gotcha | 22:22 |
blackburn | yeah old-school hah | 22:22 |
blackburn | okay I got it | 22:22 |
blackburn | about train features and test features | 22:23 |
blackburn | you are trying to classify 4th object/vector | 22:23 |
blackburn | that are not here anyhow | 22:23 |
supernaught | perhaps I am misunderstanding a training and a test | 22:23 |
supernaught | you say | 22:23 |
supernaught | this makes sense | 22:24 |
supernaught | how do I input some test data then ? | 22:24 |
blackburn | supernaught: you could use set_features | 22:25 |
blackburn | or apply to whole features | 22:25 |
supernaught | I will check out the documentation now | 22:25 |
blackburn | well we have pretty bad doc | 22:25 |
supernaught | just so I am sure, | 22:25 |
blackburn | svm->apply(features) | 22:26 |
blackburn | will return CLabels instance | 22:26 |
blackburn | when you can get classified labels | 22:26 |
blackburn | supernaught: btw why are you using C++ for doing that? | 22:27 |
supernaught | so what is happening is I will get: <interpretted/classified data> = svm->apply( <test data> ); | 22:27 |
supernaught | ? | 22:27 |
supernaught | I am tinkering with embedding shogun in a set of medical physics libraries which are written in c++ | 22:27 |
blackburn | ah | 22:28 |
blackburn | I see | 22:28 |
blackburn | apply(features) returns CLabels | 22:28 |
blackburn | which are containing labels (-1,+1,..) | 22:28 |
supernaught | Awesome. Thanks for the clarification. I'll check out the documentation and come back when I get stuck again :) | 22:28 |
blackburn | i.e. the results of classification | 22:28 |
supernaught | thanks! | 22:28 |
blackburn | you could just ask here :) | 22:29 |
supernaught | unfortunately I never seem to learn anything unless I struggle through it a little | 22:29 |
supernaught | maybe it is 'starving artist' syndrome | 22:29 |
supernaught | : / | 22:29 |
blackburn | everyone do the same I guess | 22:29 |
blackburn | https://gist.github.com/1325045 | 22:29 |
blackburn | good thing is you understand my broken english :D | 22:30 |
supernaught | not a problem | 22:30 |
supernaught | thanks again - this is some unbelieveably cool software | 22:31 |
supernaught | I'm grateful, thanks | 22:31 |
blackburn | supernaught: in the example I sent to you | 22:31 |
blackburn | there are features and features_test | 22:31 |
blackburn | ah I did it wrong | 22:31 |
blackburn | okay when there are features and features_test | 22:31 |
blackburn | and if you do svm->train with features | 22:32 |
blackburn | and then svm->apply with features_test | 22:32 |
blackburn | you will get classification results | 22:32 |
supernaught | :D | 22:32 |
blackburn | supernaught: it is nice to hear our software is any cool :) | 22:33 |
supernaught | Ok | 22:33 |
supernaught | I understand | 22:33 |
supernaught | one more question, though | 22:34 |
supernaught | how stable is the c++ api ? | 22:34 |
blackburn | supernaught: in what means? | 22:34 |
supernaught | in a few years, will I be able to compile my code? | 22:35 |
supernaught | e.g. shogun is 1.0.0 right now. How much will things change at 8.0.0 ? | 22:35 |
blackburn | oh, I hope so | 22:35 |
blackburn | are you doing eternal software? :D | 22:35 |
supernaught | in the medical world, everything is supposed to be 'eternal' and 'error-free' | 22:36 |
supernaught | ;) | 22:36 |
blackburn | yeah sure | 22:36 |
blackburn | I'm not sure about every method but I think apply/train/CLabels/etc things won't be changed | 22:36 |
supernaught | cool | 22:37 |
supernaught | Well, I'm off for now. I'll catch you later blackburn | 22:37 |
blackburn | supernaught: okay see you | 22:37 |
blackburn | supernaught: I'm really interested with attaching shogun to any other software (as Soeren does too) - so don't hesitate to ask | 22:38 |
CIA-3 | shogun: Sergey Lisitsyn master * r5ced67a / (src/shogun/base/SGObject.cpp src/shogun/base/SGObject.h): Fixed spacings at SGObject - http://git.io/I0J7rQ | 22:54 |
supernaught | blackburn: the application will be a compatibility layer between raw medical data and a uniform-surface application layer | 23:47 |
blackburn | supernaught: I see, I hope that will work | 23:47 |
supernaught | it is a for a master's thesis, but the (public) organization I work for covers a large part of Canada | 23:48 |
supernaught | and they are eager to adopt it | 23:48 |
supernaught | so I would _love_ to put shogun to work | 23:48 |
blackburn | supernaught: what are the tasks that software specialize with? | 23:54 |
blackburn | supernaught: anyway as I said before don't hesitate to ask me or Soeren and if there is no one of us here - ask on the mailing list or by email | 23:59 |
--- Log closed Sun Oct 30 00:00:17 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!