class Rumale::SVM::SVR

SVR is a class that provides Kernel Epsilon-Support Vector Regressor in LIBSVM with Rumale interface.

@example

estimator = Rumale::SVM::SVR.new(reg_param: 1.0, kernel: 'rbf', gamma: 10.0, random_seed: 1)
estimator.fit(training_samples, traininig_target_values)
results = estimator.predict(testing_samples)

Public Class Methods

new(reg_param: 1.0, epsilon: 0.1, kernel: 'rbf', degree: 3, gamma: 1.0, coef0: 0.0, shrinking: true, cache_size: 200.0, tol: 1e-3, verbose: false, random_seed: nil) click to toggle source

Create a new regressor with Kernel Epsilon-Support Vector Regressor.

@param reg_param [Float] The regularization parameter. @param epsilon [Float] The epsilon parameter in loss function of espsilon-svr. @param kernel [String] The type of kernel function ('rbf', 'linear', 'poly', 'sigmoid', and 'precomputed'). @param degree [Integer] The degree parameter in polynomial kernel function. @param gamma [Float] The gamma parameter in rbf/poly/sigmoid kernel function. @param coef0 [Float] The coefficient in poly/sigmoid kernel function. @param shrinking [Boolean] The flag indicating whether to use the shrinking heuristics. @param cache_size [Float] The cache memory size in MB. @param tol [Float] The tolerance of termination criterion. @param verbose [Boolean] The flag indicating whether to output learning process message @param random_seed [Integer/Nil] The seed value using to initialize the random generator.

# File lib/rumale/svm/svr.rb, line 32
def initialize(reg_param: 1.0, epsilon: 0.1, kernel: 'rbf', degree: 3, gamma: 1.0, coef0: 0.0,
               shrinking: true, cache_size: 200.0, tol: 1e-3, verbose: false, random_seed: nil)
  check_params_numeric(reg_param: reg_param, degree: degree, epsilon: epsilon, gamma: gamma, coef0: coef0,
                       cache_size: cache_size, tol: tol)
  check_params_string(kernel: kernel)
  check_params_boolean(shrinking: shrinking, verbose: verbose)
  check_params_numeric_or_nil(random_seed: random_seed)
  @params = {}
  @params[:reg_param] = reg_param.to_f
  @params[:epsilon] = epsilon.to_f
  @params[:kernel] = kernel
  @params[:degree] = degree.to_i
  @params[:gamma] = gamma.to_f
  @params[:coef0] = coef0.to_f
  @params[:shrinking] = shrinking
  @params[:cache_size] = cache_size.to_f
  @params[:tol] = tol.to_f
  @params[:verbose] = verbose
  @params[:random_seed] = random_seed.nil? ? nil : random_seed.to_i
end

Public Instance Methods

duel_coef() click to toggle source

Return the coefficients of the support vector in decision function. @return [Numo::DFloat] (shape: [1, n_support_vectors])

# File lib/rumale/svm/svr.rb, line 115
def duel_coef
  @model[:sv_coef]
end
fit(x, y) click to toggle source

Fit the model with given training data.

@param x [Numo::DFloat] (shape: [n_samples, n_features]) The training data to be used for fitting the model.

If the kernel is 'precomputed', x must be a square distance matrix (shape: [n_samples, n_samples]).

@param y [Numo::DFloat] (shape: [n_samples]) The target values to be used for fitting the model. @return [SVR] The learned regressor itself.

# File lib/rumale/svm/svr.rb, line 59
def fit(x, y)
  x = check_convert_sample_array(x)
  y = check_convert_tvalue_array(y)
  check_sample_tvalue_size(x, y)
  xx = precomputed_kernel? ? add_index_col(x) : x
  @model = Numo::Libsvm.train(xx, y, libsvm_params)
  self
end
intercept() click to toggle source

Return the intercepts in decision function. @return [Numo::DFloat] (shape: [1])

# File lib/rumale/svm/svr.rb, line 121
def intercept
  @model[:rho]
end
marshal_dump() click to toggle source

Dump marshal data. @return [Hash] The marshal data about SVR.

# File lib/rumale/svm/svr.rb, line 82
def marshal_dump
  { params: @params,
    model: @model }
end
marshal_load(obj) click to toggle source

Load marshal data. @return [nil]

# File lib/rumale/svm/svr.rb, line 89
def marshal_load(obj)
  @params = obj[:params]
  @model = obj[:model]
  nil
end
n_support() click to toggle source

Return the number of support vectors. @return [Integer]

# File lib/rumale/svm/svr.rb, line 109
def n_support
  support.size
end
predict(x) click to toggle source

Predict values for samples.

@param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to predict the labels.

If the kernel is 'precomputed', the shape of x must be [n_samples, n_training_samples].

@return [Numo::DFloat] (shape: [n_samples]) Predicted value per sample.

# File lib/rumale/svm/svr.rb, line 73
def predict(x)
  raise "#{self.class.name}\##{__method__} expects to be called after training the model with the fit method." unless trained?
  x = check_convert_sample_array(x)
  xx = precomputed_kernel? ? add_index_col(x) : x
  Numo::Libsvm.predict(xx, libsvm_params, @model)
end
support() click to toggle source

Return the indices of support vectors. @return [Numo::Int32] (shape: [n_support_vectors])

# File lib/rumale/svm/svr.rb, line 97
def support
  @model[:sv_indices]
end
support_vectors() click to toggle source

Return the support_vectors. @return [Numo::DFloat] (shape: [n_support_vectors, n_features])

# File lib/rumale/svm/svr.rb, line 103
def support_vectors
  precomputed_kernel? ? del_index_col(@model[:SV]) : @model[:SV]
end

Private Instance Methods

add_index_col(x) click to toggle source
# File lib/rumale/svm/svr.rb, line 127
def add_index_col(x)
  idx = Numo::Int32.new(x.shape[0]).seq + 1
  Numo::NArray.hstack([idx.expand_dims(1), x])
end
del_index_col(x) click to toggle source
# File lib/rumale/svm/svr.rb, line 132
def del_index_col(x)
  x[true, 1..-1].dup
end
libsvm_params() click to toggle source
# File lib/rumale/svm/svr.rb, line 140
def libsvm_params
  res = @params.merge(svm_type: Numo::Libsvm::SvmType::EPSILON_SVR)
  res[:kernel_type] = case res.delete(:kernel)
                      when 'linear'
                        Numo::Libsvm::KernelType::LINEAR
                      when 'poly'
                        Numo::Libsvm::KernelType::POLY
                      when 'sigmoid'
                        Numo::Libsvm::KernelType::SIGMOID
                      when 'precomputed'
                        Numo::Libsvm::KernelType::PRECOMPUTED
                      else
                        Numo::Libsvm::KernelType::RBF
                      end
  res[:C] = res.delete(:reg_param)
  res[:p] = res.delete(:epsilon)
  res[:eps] = res.delete(:tol)
  res
end
precomputed_kernel?() click to toggle source
# File lib/rumale/svm/svr.rb, line 136
def precomputed_kernel?
  @params[:kernel] == 'precomputed'
end
trained?() click to toggle source
# File lib/rumale/svm/svr.rb, line 160
def trained?
  !@model.nil?
end