rumale-linear_model 0.24.0 → 0.25.0

data/lib/rumale/linear_model/ridge.rb

@@ -4,19 +4,18 @@ require 'lbfgsb'
 
 require 'rumale/base/regressor'
 require 'rumale/validation'
-require 'rumale/linear_model/base_sgd'
+
+require_relative 'base_estimator'
 
 module Rumale
   module LinearModel
     # Ridge is a class that implements Ridge Regression
-    # with stochastic gradient descent (SGD) optimization,
-    # singular value decomposition (SVD), or L-BFGS optimization.
+    # with singular value decomposition (SVD) or L-BFGS optimization.
     #
     # @example
     #   require 'rumale/linear_model/ridge'
     #
-    #   estimator =
-    #     Rumale::LinearModel::Ridge.new(reg_param: 0.1, max_iter: 1000, batch_size: 20, random_seed: 1)
+    #   estimator = Rumale::LinearModel::Ridge.new(reg_param: 0.1)
     #   estimator.fit(training_samples, traininig_values)
     #   results = estimator.predict(testing_samples)
     #
@@ -27,73 +26,40 @@ module Rumale
     #   estimator = Rumale::LinearModel::Ridge.new(reg_param: 0.1, solver: 'svd')
     #   estimator.fit(training_samples, traininig_values)
     #   results = estimator.predict(testing_samples)
-    #
-    # *Reference*
-    # - Bottou, L., "Large-Scale Machine Learning with Stochastic Gradient Descent," Proc. COMPSTAT'10, pp. 177--186, 2010.
-    class Ridge < BaseSGD
-      include ::Rumale::Base::Regressor
-
-      # Return the weight vector.
-      # @return [Numo::DFloat] (shape: [n_outputs, n_features])
-      attr_reader :weight_vec
-
-      # Return the bias term (a.k.a. intercept).
-      # @return [Numo::DFloat] (shape: [n_outputs])
-      attr_reader :bias_term
-
-      # Return the random generator for random sampling.
-      # @return [Random]
-      attr_reader :rng
+    class Ridge < Rumale::LinearModel::BaseEstimator
+      include Rumale::Base::Regressor
 
       # Create a new Ridge regressor.
       #
-      # @param learning_rate [Float] The initial value of learning rate.
-      #   The learning rate decreases as the iteration proceeds according to the equation: learning_rate / (1 + decay * t).
-      #   If solver is not 'sgd', this parameter is ignored.
-      # @param decay [Float] The smoothing parameter for decreasing learning rate as the iteration proceeds.
-      #   If nil is given, the decay sets to 'reg_param * learning_rate'.
-      #   If solver is not 'sgd', this parameter is ignored.
-      # @param momentum [Float] The momentum factor.
-      #   If solver is not 'sgd', this parameter is ignored.
       # @param reg_param [Float] The regularization parameter.
       # @param fit_bias [Boolean] The flag indicating whether to fit the bias term.
       # @param bias_scale [Float] The scale of the bias term.
       # @param max_iter [Integer] The maximum number of epochs that indicates
       #   how many times the whole data is given to the training process.
       #   If solver is 'svd', this parameter is ignored.
-      # @param batch_size [Integer] The size of the mini batches.
-      #   If solver is not 'sgd', this parameter is ignored.
       # @param tol [Float] The tolerance of loss for terminating optimization.
       #   If solver is 'svd', this parameter is ignored.
-      # @param solver [String] The algorithm to calculate weights. ('auto', 'sgd', 'svd', or 'lbfgs').
+      # @param solver [String] The algorithm to calculate weights. ('auto', 'svd', or 'lbfgs').
       #   'auto' chooses the 'svd' solver if Numo::Linalg is loaded. Otherwise, it chooses the 'lbfgs' solver.
-      #   'sgd' uses the stochastic gradient descent optimization.
       #   'svd' performs singular value decomposition of samples.
       #   'lbfgs' uses the L-BFGS method for optimization.
-      # @param n_jobs [Integer] The number of jobs for running the fit method in parallel.
-      #   If nil is given, the method does not execute in parallel.
-      #   If zero or less is given, it becomes equal to the number of processors.
-      #   This parameter is ignored if the Parallel gem is not loaded or solver is not 'sgd'.
       # @param verbose [Boolean] The flag indicating whether to output loss during iteration.
       #   If solver is 'svd', this parameter is ignored.
-      # @param random_seed [Integer] The seed value using to initialize the random generator.
-      def initialize(learning_rate: 0.01, decay: nil, momentum: 0.9,
-                     reg_param: 1.0, fit_bias: true, bias_scale: 1.0,
-                     max_iter: 1000, batch_size: 50, tol: 1e-4,
-                     solver: 'auto',
-                     n_jobs: nil, verbose: false, random_seed: nil)
+      def initialize(reg_param: 1.0, fit_bias: true, bias_scale: 1.0, max_iter: 1000, tol: 1e-4, solver: 'auto', verbose: false)
         super()
-        @params.merge!(method(:initialize).parameters.to_h { |_t, arg| [arg, binding.local_variable_get(arg)] })
+        @params = {
+          reg_param: reg_param,
+          fit_bias: fit_bias,
+          bias_scale: bias_scale,
+          max_iter: max_iter,
+          tol: tol,
+          verbose: verbose
+        }
         @params[:solver] = if solver == 'auto'
                              enable_linalg?(warning: false) ? 'svd' : 'lbfgs'
                            else
-                             solver.match?(/^svd$|^sgd$|^lbfgs$/) ? solver : 'lbfgs'
+                             solver.match?(/^svd$|^lbfgs$/) ? solver : 'lbfgs'
                            end
-        @params[:decay] ||= @params[:reg_param] * @params[:learning_rate]
-        @params[:random_seed] ||= srand
-        @rng = Random.new(@params[:random_seed])
-        @penalty_type = L2_PENALTY
-        @loss_func = ::Rumale::LinearModel::Loss::MeanSquaredError.new
       end
 
       # Fit the model with given training data.
@@ -102,17 +68,15 @@ module Rumale
       # @param y [Numo::DFloat] (shape: [n_samples, n_outputs]) The target values to be used for fitting the model.
       # @return [Ridge] The learned regressor itself.
       def fit(x, y)
-        x = ::Rumale::Validation.check_convert_sample_array(x)
-        y = ::Rumale::Validation.check_convert_target_value_array(y)
-        ::Rumale::Validation.check_sample_size(x, y)
-
-        if @params[:solver] == 'svd' && enable_linalg?(warning: false)
-          fit_svd(x, y)
-        elsif @params[:solver] == 'lbfgs'
-          fit_lbfgs(x, y)
-        else
-          fit_sgd(x, y)
-        end
+        x = Rumale::Validation.check_convert_sample_array(x)
+        y = Rumale::Validation.check_convert_target_value_array(y)
+        Rumale::Validation.check_sample_size(x, y)
+
+        @weight_vec, @bias_term = if @params[:solver] == 'svd' && enable_linalg?(warning: false)
+                                    partial_fit_svd(x, y)
+                                  else
+                                    partial_fit_lbfgs(x, y)
+                                  end
 
         self
       end
@@ -122,25 +86,24 @@ module Rumale
       # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to predict the values.
       # @return [Numo::DFloat] (shape: [n_samples, n_outputs]) Predicted values per sample.
       def predict(x)
-        x = ::Rumale::Validation.check_convert_sample_array(x)
+        x = Rumale::Validation.check_convert_sample_array(x)
 
         x.dot(@weight_vec.transpose) + @bias_term
       end
 
       private
 
-      def fit_svd(x, y)
+      def partial_fit_svd(x, y)
         x = expand_feature(x) if fit_bias?
-
         s, u, vt = Numo::Linalg.svd(x, driver: 'sdd', job: 'S')
         d = (s / (s**2 + @params[:reg_param])).diag
         w = vt.transpose.dot(d).dot(u.transpose).dot(y)
-
-        @weight_vec, @bias_term = single_target?(y) ? split_weight(w) : split_weight_mult(w)
+        w = w.transpose.dup unless single_target?(y)
+        split_weight(w)
       end
 
-      def fit_lbfgs(x, y)
-        fnc = proc do |w, x, y, a| # rubocop:disable Lint/ShadowingOuterLocalVariable
+      def partial_fit_lbfgs(base_x, base_y)
+        fnc = proc do |w, x, y, a|
           n_samples, n_features = x.shape
           w = w.reshape(y.shape[1], n_features) unless y.shape[1].nil?
           z = x.dot(w.transpose)
@@ -150,57 +113,25 @@ module Rumale
           [loss, gradient.flatten.dup]
         end
 
-        x = expand_feature(x) if fit_bias?
+        base_x = expand_feature(base_x) if fit_bias?
 
-        n_features = x.shape[1]
-        n_outputs = single_target?(y) ? 1 : y.shape[1]
+        n_features = base_x.shape[1]
+        n_outputs = single_target?(base_y) ? 1 : base_y.shape[1]
+        w_init = Numo::DFloat.zeros(n_outputs * n_features)
 
         res = Lbfgsb.minimize(
-          fnc: fnc, jcb: true, x_init: init_weight(n_features, n_outputs), args: [x, y, @params[:reg_param]],
+          fnc: fnc, jcb: true, x_init: w_init, args: [base_x, base_y, @params[:reg_param]],
           maxiter: @params[:max_iter], factr: @params[:tol] / Lbfgsb::DBL_EPSILON,
           verbose: @params[:verbose] ? 1 : -1
         )
 
-        @weight_vec, @bias_term =
-          if single_target?(y)
-            split_weight(res[:x])
-          else
-            split_weight_mult(res[:x].reshape(n_outputs, n_features).transpose)
-          end
-      end
-
-      def fit_sgd(x, y)
-        if single_target?(y)
-          @weight_vec, @bias_term = partial_fit(x, y)
-        else
-          n_outputs = y.shape[1]
-          n_features = x.shape[1]
-          @weight_vec = Numo::DFloat.zeros(n_outputs, n_features)
-          @bias_term = Numo::DFloat.zeros(n_outputs)
-          if enable_parallel?
-            models = parallel_map(n_outputs) { |n| partial_fit(x, y[true, n]) }
-            n_outputs.times { |n| @weight_vec[n, true], @bias_term[n] = models[n] }
-          else
-            n_outputs.times { |n| @weight_vec[n, true], @bias_term[n] = partial_fit(x, y[true, n]) }
-          end
-        end
+        w = single_target?(base_y) ? res[:x] : res[:x].reshape(n_outputs, n_features)
+        split_weight(w)
       end
 
       def single_target?(y)
         y.ndim == 1
       end
-
-      def init_weight(n_features, n_outputs)
-        ::Rumale::Utils.rand_normal([n_outputs, n_features], @rng.dup).flatten.dup
-      end
-
-      def split_weight_mult(w)
-        if fit_bias?
-          [w[0...-1, true].dup, w[-1, true].dup]
-        else
-          [w.dup, Numo::DFloat.zeros(w.shape[1])]
-        end
-      end
     end
   end
 end

data/lib/rumale/linear_model/sgd_classifier.rb

@@ -0,0 +1,262 @@
+# frozen_string_literal: true
+
+require 'rumale/base/classifier'
+require 'rumale/probabilistic_output'
+require 'rumale/validation'
+
+require_relative 'sgd_estimator'
+
+module Rumale
+  module LinearModel
+    # SGDClassifier is a class that implements linear classifier with stochastic gradient descent optimization.
+    #
+    # @example
+    #   require 'rumale/linear_model/sgd_classifier'
+    #
+    #   estimator =
+    #     Rumale::LinearModel::SGDClassifier.new(loss: 'hinge', reg_param: 1.0, max_iter: 1000, batch_size: 50, random_seed: 1)
+    #   estimator.fit(training_samples, traininig_labels)
+    #   results = estimator.predict(testing_samples)
+    #
+    # *Reference*
+    # - Shalev-Shwartz, S., and Singer, Y., "Pegasos: Primal Estimated sub-GrAdient SOlver for SVM," Proc. ICML'07, pp. 807--814, 2007.
+    # - Tsuruoka, Y., Tsujii, J., and Ananiadou, S., "Stochastic Gradient Descent Training for L1-regularized Log-linear Models with Cumulative Penalty," Proc. ACL'09, pp. 477--485, 2009.
+    # - Bottou, L., "Large-Scale Machine Learning with Stochastic Gradient Descent," Proc. COMPSTAT'10, pp. 177--186, 2010.
+    class SGDClassifier < Rumale::LinearModel::SGDEstimator # rubocop:disable Metrics/ClassLength
+      include Rumale::Base::Classifier
+
+      # Return the class labels.
+      # @return [Numo::Int32] (shape: [n_classes])
+      attr_reader :classes
+
+      # Return the random generator for performing random sampling.
+      # @return [Random]
+      attr_reader :rng
+
+      # Create a new linear classifier with stochastic gradient descent optimization.
+      #
+      # @param loss [String] The loss function to be used ('hinge' and 'log_loss').
+      # @param learning_rate [Float] The initial value of learning rate.
+      #   The learning rate decreases as the iteration proceeds according to the equation: learning_rate / (1 + decay * t).
+      # @param decay [Float] The smoothing parameter for decreasing learning rate as the iteration proceeds.
+      #   If nil is given, the decay sets to 'reg_param * learning_rate'.
+      # @param momentum [Float] The momentum factor.
+      # @param penalty [String] The regularization type to be used ('l1', 'l2', and 'elasticnet').
+      # @param l1_ratio [Float] The elastic-net type regularization mixing parameter.
+      #   If penalty set to 'l2' or 'l1', this parameter is ignored.
+      #   If l1_ratio = 1, the regularization is similar to Lasso.
+      #   If l1_ratio = 0, the regularization is similar to Ridge.
+      #   If 0 < l1_ratio < 1, the regularization is a combination of L1 and L2.
+      # @param reg_param [Float] The regularization parameter.
+      # @param fit_bias [Boolean] The flag indicating whether to fit the bias term.
+      # @param bias_scale [Float] The scale of the bias term.
+      # @param max_iter [Integer] The maximum number of epochs that indicates
+      #   how many times the whole data is given to the training process.
+      # @param batch_size [Integer] The size of the mini batches.
+      # @param tol [Float] The tolerance of loss for terminating optimization.
+      # @param n_jobs [Integer] The number of jobs for running the fit and predict methods in parallel.
+      #   If nil is given, the methods do not execute in parallel.
+      #   If zero or less is given, it becomes equal to the number of processors.
+      #   This parameter is ignored if the Parallel gem is not loaded.
+      # @param verbose [Boolean] The flag indicating whether to output loss during iteration.
+      # @param random_seed [Integer] The seed value using to initialize the random generator.
+      def initialize(loss: 'hinge', learning_rate: 0.01, decay: nil, momentum: 0.9,
+                     penalty: 'l2', reg_param: 1.0, l1_ratio: 0.5,
+                     fit_bias: true, bias_scale: 1.0,
+                     max_iter: 1000, batch_size: 50, tol: 1e-4,
+                     n_jobs: nil, verbose: false, random_seed: nil)
+        super()
+        @params.merge!(
+          loss: loss,
+          learning_rate: learning_rate,
+          decay: decay,
+          momentum: momentum,
+          penalty: penalty,
+          reg_param: reg_param,
+          l1_ratio: l1_ratio,
+          fit_bias: fit_bias,
+          bias_scale: bias_scale,
+          max_iter: max_iter,
+          batch_size: batch_size,
+          tol: tol,
+          n_jobs: n_jobs,
+          verbose: verbose,
+          random_seed: random_seed
+        )
+        @params[:decay] ||= @params[:reg_param] * @params[:learning_rate]
+        @params[:random_seed] ||= srand
+        @rng = Random.new(@params[:random_seed])
+        @penalty_type = @params[:penalty]
+        @loss_func = case @params[:loss]
+                     when Rumale::LinearModel::Loss::HingeLoss::NAME
+                       Rumale::LinearModel::Loss::HingeLoss.new
+                     when Rumale::LinearModel::Loss::LogLoss::NAME
+                       Rumale::LinearModel::Loss::LogLoss.new
+                     else
+                       raise ArgumentError, "given loss '#{loss}' is not supported."
+                     end
+      end
+
+      # 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.
+      # @param y [Numo::Int32] (shape: [n_samples]) The labels to be used for fitting the model.
+      # @return [SGDClassifier] The learned classifier itself.
+      def fit(x, y)
+        x = Rumale::Validation.check_convert_sample_array(x)
+        y = Rumale::Validation.check_convert_label_array(y)
+        Rumale::Validation.check_sample_size(x, y)
+
+        @classes = Numo::Int32[*y.to_a.uniq.sort]
+
+        send("fit_#{@loss_func.name}", x, y)
+
+        self
+      end
+
+      # Calculate confidence scores for samples.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to compute the scores.
+      # @return [Numo::DFloat] (shape: [n_samples, n_classes]) Confidence score per sample.
+      def decision_function(x)
+        x = ::Rumale::Validation.check_convert_sample_array(x)
+
+        x.dot(@weight_vec.transpose) + @bias_term
+      end
+
+      # Predict class labels for samples.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to predict the labels.
+      # @return [Numo::Int32] (shape: [n_samples]) Predicted class label per sample.
+      def predict(x)
+        x = ::Rumale::Validation.check_convert_sample_array(x)
+
+        send("predict_#{@loss_func.name}", x)
+      end
+
+      # Predict probability for samples.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to predict the probailities.
+      # @return [Numo::DFloat] (shape: [n_samples, n_classes]) Predicted probability of each class per sample.
+      def predict_proba(x)
+        x = ::Rumale::Validation.check_convert_sample_array(x)
+
+        send("predict_proba_#{@loss_func.name}", x)
+      end
+
+      private
+
+      def fit_hinge(x, y)
+        if multiclass_problem?
+          n_classes = @classes.size
+          n_features = x.shape[1]
+          @weight_vec = Numo::DFloat.zeros(n_classes, n_features)
+          @bias_term = Numo::DFloat.zeros(n_classes)
+          @prob_param = Numo::DFloat.zeros(n_classes, 2)
+          models = if enable_parallel?
+                     parallel_map(n_classes) do |n|
+                       bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
+                       w, b = partial_fit(x, bin_y)
+                       prb = Rumale::ProbabilisticOutput.fit_sigmoid(x.dot(w.transpose) + b, bin_y)
+                       [w, b, prb]
+                     end
+                   else
+                     Array.new(n_classes) do |n|
+                       bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
+                       w, b = partial_fit(x, bin_y)
+                       prb = Rumale::ProbabilisticOutput.fit_sigmoid(x.dot(w.transpose) + b, bin_y)
+                       [w, b, prb]
+                     end
+                   end
+          # store model.
+          models.each_with_index { |model, n| @weight_vec[n, true], @bias_term[n], @prob_param[n, true] = model }
+        else
+          negative_label = @classes[0]
+          bin_y = Numo::Int32.cast(y.ne(negative_label)) * 2 - 1
+          @weight_vec, @bias_term = partial_fit(x, bin_y)
+          @prob_param = Rumale::ProbabilisticOutput.fit_sigmoid(x.dot(@weight_vec.transpose) + @bias_term, bin_y)
+        end
+      end
+
+      def fit_log_loss(x, y)
+        if multiclass_problem?
+          n_classes = @classes.size
+          n_features = x.shape[1]
+          @weight_vec = Numo::DFloat.zeros(n_classes, n_features)
+          @bias_term = Numo::DFloat.zeros(n_classes)
+          if enable_parallel?
+            models = parallel_map(n_classes) do |n|
+              bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
+              partial_fit(x, bin_y)
+            end
+            n_classes.times { |n| @weight_vec[n, true], @bias_term[n] = models[n] }
+          else
+            n_classes.times do |n|
+              bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
+              @weight_vec[n, true], @bias_term[n] = partial_fit(x, bin_y)
+            end
+          end
+        else
+          negative_label = @classes[0]
+          bin_y = Numo::Int32.cast(y.ne(negative_label)) * 2 - 1
+          @weight_vec, @bias_term = partial_fit(x, bin_y)
+        end
+      end
+
+      def predict_proba_hinge(x)
+        if multiclass_problem?
+          probs = 1.0 / (Numo::NMath.exp(@prob_param[true, 0] * decision_function(x) + @prob_param[true, 1]) + 1.0)
+          (probs.transpose / probs.sum(axis: 1)).transpose.dup
+        else
+          n_samples = x.shape[0]
+          probs = Numo::DFloat.zeros(n_samples, 2)
+          probs[true, 1] = 1.0 / (Numo::NMath.exp(@prob_param[0] * decision_function(x) + @prob_param[1]) + 1.0)
+          probs[true, 0] = 1.0 - probs[true, 1]
+          probs
+        end
+      end
+
+      def predict_proba_log_loss(x)
+        proba = 1.0 / (Numo::NMath.exp(-decision_function(x)) + 1.0)
+        return (proba.transpose / proba.sum(axis: 1)).transpose.dup if multiclass_problem?
+
+        n_samples = x.shape[0]
+        probs = Numo::DFloat.zeros(n_samples, 2)
+        probs[true, 1] = proba
+        probs[true, 0] = 1.0 - proba
+        probs
+      end
+
+      def predict_hinge(x)
+        n_samples = x.shape[0]
+        predicted = if multiclass_problem?
+                      decision_values = decision_function(x)
+                      if enable_parallel?
+                        parallel_map(n_samples) { |n| @classes[decision_values[n, true].max_index] }
+                      else
+                        Array.new(n_samples) { |n| @classes[decision_values[n, true].max_index] }
+                      end
+                    else
+                      decision_values = decision_function(x).ge(0.0).to_a
+                      Array.new(n_samples) { |n| @classes[decision_values[n]] }
+                    end
+        Numo::Int32.asarray(predicted)
+      end
+
+      def predict_log_loss(x)
+        n_samples = x.shape[0]
+        decision_values = predict_proba_log_loss(x)
+        predicted = if enable_parallel?
+                      parallel_map(n_samples) { |n| @classes[decision_values[n, true].max_index] }
+                    else
+                      Array.new(n_samples) { |n| @classes[decision_values[n, true].max_index] }
+                    end
+        Numo::Int32.asarray(predicted)
+      end
+
+      def multiclass_problem?
+        @classes.size > 2
+      end
+    end
+  end
+end

data/lib/rumale/linear_model/{base_sgd.rb → sgd_estimator.rb}

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-require 'rumale/base/estimator'
+require_relative 'base_estimator'
 
 module Rumale
   module LinearModel
@@ -96,6 +96,14 @@ module Rumale
         def dloss(out, y)
           2.fdiv(y.shape[0]) * (out - y)
         end
+
+        # @!visibility private
+        def name
+          NAME
+        end
+
+        # @!visibility private
+        NAME = 'squared_error'
       end
 
       # @!visibility private
@@ -110,6 +118,14 @@ module Rumale
         def dloss(out, y)
           y / (1 + Numo::NMath.exp(-y * out)) - y
         end
+
+        # @!visibility private
+        def name
+          NAME
+        end
+
+        # @!visibility private
+        NAME = 'log_loss'
       end
 
       # @!visibility private
@@ -127,6 +143,14 @@ module Rumale
           d[tids] = -y[tids] if tids.count.positive?
           d
         end
+
+        # @!visibility private
+        def name
+          NAME
+        end
+
+        # @!visibility private
+        NAME = 'hinge'
       end
 
       # @!visibility private
@@ -151,13 +175,21 @@ module Rumale
           d[tids] = -1 if tids.count.positive?
           d
         end
+
+        # @!visibility private
+        def name
+          NAME
+        end
+
+        # @!visibility private
+        NAME = 'epsilon_insensitive'
       end
     end
 
-    # BaseSGD is an abstract class for implementation of linear model with mini-batch stochastic gradient descent (SGD) optimization.
+    # SGDEstimator is an abstract class for implementation of linear model with mini-batch stochastic gradient descent (SGD) optimization.
     # This class is used internally.
-    class BaseSGD < ::Rumale::Base::Estimator
-      # Create an initial linear model.
+    class SGDEstimator < Rumale::LinearModel::BaseEstimator
+      # Create an initial linear model with SGD.
       def initialize
         super()
         @params = {
@@ -224,23 +256,6 @@ module Rumale
         split_weight(weight)
       end
 
-      def expand_feature(x)
-        n_samples = x.shape[0]
-        Numo::NArray.hstack([x, Numo::DFloat.ones([n_samples, 1]) * @params[:bias_scale]])
-      end
-
-      def split_weight(weight)
-        if fit_bias?
-          [weight[0...-1].dup, weight[-1]]
-        else
-          [weight, 0.0]
-        end
-      end
-
-      def fit_bias?
-        @params[:fit_bias] == true
-      end
-
       def apply_l2_penalty?
         @penalty_type == L2_PENALTY || @penalty_type == ELASTICNET_PENALTY
       end