I'm guessing that line 52 needs to be similar to the above ternary

pred = model.predict(Libsvm::Node.features(dictionary.map{|x| test_document.include?(x) ? 1 : 0 }))

otherwise, test_document.include?(x) returns [false, false] and the entire thing errors ( no method to_f for false:Falseclass, etc)

the result with the ternary, which means little right now is....

=> 543175208.0

I've installed ruby, devkit and rubygems and sucessfully installed the rb-libsvm gem. When I try to run the examples or go require 'libsvm' in irb i get this:

C:\dev>ruby
require 'libsvm'
^Z
c:/dev/ruby193/lib/ruby/site_ruby/1.9.1/rubygems/custom_require.rb:36:in `require': 193: %1 is not a valid Win32 application.   - c:/dev/ruby193/lib/ruby/gems/1.9.1/gems/rb-libsvm-1.0.8/lib/libsvm/libsvm_ext.so (LoadError)
        from c:/dev/ruby193/lib/ruby/site_ruby/1.9.1/rubygems/custom_require.rb:36:in `require'
        from c:/dev/ruby193/lib/ruby/gems/1.9.1/gems/rb-libsvm-1.0.8/lib/libsvm.rb:2:in `'
        from c:/dev/ruby193/lib/ruby/site_ruby/1.9.1/rubygems/custom_require.rb:60:in `require'
        from c:/dev/ruby193/lib/ruby/site_ruby/1.9.1/rubygems/custom_require.rb:60:in `rescue in require'
        from c:/dev/ruby193/lib/ruby/site_ruby/1.9.1/rubygems/custom_require.rb:35:in `require'
        from -:1:in `'

It seems like it would be good to either throw an exception or return nil in the event that svm_load_model(path) (libsvm.c line 500) returns NULL.

/* throw exception if svm_load_model returns NULL */
if (!model) {
rb_raise(rb_eIOError, "Unable to read model file: %s", path);
}

The other approach would look like:

return model ? (Data_Wrap_Struct(cModel,0, model_free, model)) : Qnil;

Without any feedback from svm_load_model, Ruby segfaults when I call predict on the returned model.

Thanks,

Tom

I tried to get prediction probability, but even with the simple example provided in readme file, the return value of method is always 0.0 for each predicted class.

Example

require 'libsvm'

# This library is namespaced.
problem = Libsvm::Problem.new
parameter = Libsvm::SvmParameter.new

parameter.cache_size = 1 # in megabytes

parameter.eps = 0.001
parameter.c = 10

examples = [ [1,0,1], [-1,0,-1] ].map {|ary| Libsvm::Node.features(ary) }
labels = [1, -1]

problem.set_examples(labels, examples)

model = Libsvm::Model.train(problem, parameter)

test = [0,0,0]

pred = model.predict_probability(Libsvm::Node.features(test))
puts "Example #{test} - Predicted #{pred}"

>>> Example [0, 0, 0] - Predicted [-1.0, [0.0, 0.0]]

I'm using OS X 10.8.4 but I tried also on latest ubuntu with the same result.

In Ruby community there is a strong directory structure pattern, which makes the modules correspond to directories.

I know that the library name might not be changed to libsvm, since this a different implementation of the libsvm wrapper, but I think rb-libsvm would be less surprising for developers, if the paths inside the lib directory were corresponding the the module Libsvm.

So instead of

  lib/rb-libsvm

we would have

  lib/libsvm

This also applies to the rb-libsvm.so which is generated during library installation.

How does one perform cross validation? Is it possible to provide some examples?

Does it possible use SVR through this gem?

Hi.

I would like to make a ROC-curve for the results.
Are there any way to get the decimal values for each prediction?
Or probably some other ways to do it?

Thank you

Hey,
I am trying to use your gem to make one-class svm work. I was just trying to make it work with the modified example code:

problem = Libsvm::Problem.new

parameter = Libsvm::SvmParameter.new
parameter.cache_size = 1 # in megabytes
parameter.eps = 0.001
parameter.nu = 0.5
parameter.svm_type = Libsvm::SvmType::ONE_CLASS

examples = [ [1,1,1], [0,0,0] ].map {|ary| Libsvm::Node.features(ary) }
labels = [1,1]
problem.set_examples(labels, examples)

model = Libsvm::Model.train(problem, parameter)

puts model.predict(Libsvm::Node.features(1, 1, 1))
puts model.predict(Libsvm::Node.features(0, 0, 0))
puts model.predict(Libsvm::Node.features(100, 100, 100))

However the output is -1 -1 -1 insted of 1 1 -1 as I would expect. What am I doing wrong?

It does not have NU_CVC and C_CVC enabled?
Returns an error.

How to enable and use it?

Pleas convert tabs in source files into two spaces. This is a very strong convention in the Ruby community.

Does this lib support the JRuby? I can not install it with jRuby interpreter

Dear Florian,

we've recently added rb-libsvm to our RubyML list: https://github.com/arbox/machine-learning-with-ruby

In order to connect all Ruby & ML related projects we introduced the rubyml topic on GitHub.

Please cosider adding the rubyml topic to this repository as well so the Ruby community can discover rb-libsvm faster 😃

Thank you in advance!

I haven't found any way to save a trained model, so I am making this a feature request.

I was connected to this project through neovintages blog articles I found in my search to understand machine learning and what I can do with it in ruby.

I'm interested doing text processing, and the one article specifically pointed this out for text processing -- but it seems to take only libsvm formats (which I'm still trying to learn) and to do text processing requires a bit more along the lines of sally(http://www.mlsec.org/sally) which I've also stumbled across and am integrating into what I'm trying to accomplish.

So, 2 question points:

1 - Is there some sort of text processing feature I'm unaware of? From what I saw there is not, but if there would be more examples would be appreciated.
2 - What is the exact data format(s)? I know it is libsvm data, but the example is sparse at best.

here is sample https://github.com/aero2a/kape-
is possible to recognise this code?
I found library to generate this

Hi,

Using rb-libsvm (1.0.4), ruby 1.9.2p136 (x86_64-darwin10.5.0) the following code causes a segfault:

require 'libsvm'

# This library is namespaced.
problem = Libsvm::Problem.new
parameter = Libsvm::SvmParameter.new

parameter.cache_size = 1 # in megabytes
parameter.eps = 0.001
parameter.c = 10
parameter.kernel_type = 2
parameter.gamma = 0.0025

examples = [ [1,0,1], [-1,0,-1] ].map {|ary| Libsvm::Node.features(ary) }
labels = [1, -1]

problem.set_examples(labels, examples)

model = Libsvm::Model.train(problem, parameter)

pred = model.predict(Libsvm::Node.features(1, 1, 1))
puts "Example [1, 1, 1] - Predicted #{pred}"

The segfault message reads

Example [1, 1, 1] - Predicted 1.0
lib/test_params.rb: [BUG] Segmentation fault
ruby 1.9.2p136 (2010-12-25 revision 30365) [x86_64-darwin10.5.0]

-- control frame ----------
c:0001 p:0000 s:0002 b:0002 l:0014f8 d:0014f8 TOP   
---------------------------

-- C level backtrace information -------------------------------------------

[NOTE]
You may have encountered a bug in the Ruby interpreter or extension libraries.
Bug reports are welcome.
For details: http://www.ruby-lang.org/bugreport.html

Abort trap

The code runs fine if I comment out the parameter.kernel_type = 2 line.

Any ideas?

Running ruby 1.9.2p136, rubygems 1.8.11 and rb-libsvm-1.0.1 I get the following error when attempting to install the gem:

Building native extensions. This could take a while...
Invalid gemspec in [$gemspec_location]: Illformed requirement ["#Syck::DefaultKey:0x000001020dac70 2.7.0"]
Successfully installed rb-libsvm-1.0.1

I don't think rb-libsvm has been installed however as 'require 'libsvm', as per the documentation, gives 'LoadError: no such file to load -- libsvm'