rumale 0.8.0

data/lib/rumale/linear_model/svc.rb

@@ -0,0 +1,183 @@
+# frozen_string_literal: true
+
+require 'rumale/linear_model/base_linear_model'
+require 'rumale/base/classifier'
+require 'rumale/probabilistic_output'
+
+module Rumale
+  # This module consists of the classes that implement generalized linear models.
+  module LinearModel
+    # SVC is a class that implements Support Vector Classifier
+    # with mini-batch stochastic gradient descent optimization.
+    # For multiclass classification problem, it uses one-vs-the-rest strategy.
+    #
+    # @example
+    #   estimator =
+    #     Rumale::LinearModel::SVC.new(reg_param: 1.0, max_iter: 1000, batch_size: 20, random_seed: 1)
+    #   estimator.fit(training_samples, traininig_labels)
+    #   results = estimator.predict(testing_samples)
+    #
+    # *Reference*
+    # - S. Shalev-Shwartz and Y. Singer, "Pegasos: Primal Estimated sub-GrAdient SOlver for SVM," Proc. ICML'07, pp. 807--814, 2007.
+    class SVC < BaseLinearModel
+      include Base::Classifier
+
+      # Return the weight vector for SVC.
+      # @return [Numo::DFloat] (shape: [n_classes, n_features])
+      attr_reader :weight_vec
+
+      # Return the bias term (a.k.a. intercept) for SVC.
+      # @return [Numo::DFloat] (shape: [n_classes])
+      attr_reader :bias_term
+
+      # 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 classifier with Support Vector Machine by the SGD optimization.
+      #
+      # @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 iterations.
+      # @param batch_size [Integer] The size of the mini batches.
+      # @param probability [Boolean] The flag indicating whether to perform probability estimation.
+      # @param optimizer [Optimizer] The optimizer to calculate adaptive learning rate.
+      #   If nil is given, Nadam is used.
+      # @param random_seed [Integer] The seed value using to initialize the random generator.
+      def initialize(reg_param: 1.0, fit_bias: false, bias_scale: 1.0,
+                     max_iter: 1000, batch_size: 20, probability: false, optimizer: nil, random_seed: nil)
+        check_params_float(reg_param: reg_param, bias_scale: bias_scale)
+        check_params_integer(max_iter: max_iter, batch_size: batch_size)
+        check_params_boolean(fit_bias: fit_bias, probability: probability)
+        check_params_type_or_nil(Integer, random_seed: random_seed)
+        check_params_positive(reg_param: reg_param, bias_scale: bias_scale, max_iter: max_iter, batch_size: batch_size)
+        keywd_args = method(:initialize).parameters.map { |_t, arg| [arg, binding.local_variable_get(arg)] }.to_h
+        keywd_args.delete(:probability)
+        super(keywd_args)
+        @params[:probability] = probability
+        @prob_param = nil
+        @classes = nil
+      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 [SVC] The learned classifier itself.
+      def fit(x, y)
+        check_sample_array(x)
+        check_label_array(y)
+        check_sample_label_size(x, y)
+
+        @classes = Numo::Int32[*y.to_a.uniq.sort]
+        n_classes = @classes.size
+        n_features = x.shape[1]
+
+        if n_classes > 2
+          @weight_vec = Numo::DFloat.zeros(n_classes, n_features)
+          @bias_term = Numo::DFloat.zeros(n_classes)
+          @prob_param = Numo::DFloat.zeros(n_classes, 2)
+          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)
+            @prob_param[n, true] = if @params[:probability]
+                                     Rumale::ProbabilisticOutput.fit_sigmoid(x.dot(@weight_vec[n, true].transpose) + @bias_term[n], bin_y)
+                                   else
+                                     Numo::DFloat[1, 0]
+                                   end
+          end
+        else
+          negative_label = y.to_a.uniq.min
+          bin_y = Numo::Int32.cast(y.ne(negative_label)) * 2 - 1
+          @weight_vec, @bias_term = partial_fit(x, bin_y)
+          @prob_param = if @params[:probability]
+                          Rumale::ProbabilisticOutput.fit_sigmoid(x.dot(@weight_vec.transpose) + @bias_term, bin_y)
+                        else
+                          Numo::DFloat[1, 0]
+                        end
+        end
+
+        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)
+        check_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)
+        check_sample_array(x)
+
+        return Numo::Int32.cast(decision_function(x).ge(0.0)) * 2 - 1 if @classes.size <= 2
+
+        n_samples, = x.shape
+        decision_values = decision_function(x)
+        Numo::Int32.asarray(Array.new(n_samples) { |n| @classes[decision_values[n, true].max_index] })
+      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)
+        check_sample_array(x)
+
+        if @classes.size > 2
+          probs = 1.0 / (Numo::NMath.exp(@prob_param[true, 0] * decision_function(x) + @prob_param[true, 1]) + 1.0)
+          return (probs.transpose / probs.sum(axis: 1)).transpose
+        end
+
+        n_samples, = x.shape
+        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
+
+      # Dump marshal data.
+      # @return [Hash] The marshal data about SVC.
+      def marshal_dump
+        { params: @params,
+          weight_vec: @weight_vec,
+          bias_term: @bias_term,
+          prob_param: @prob_param,
+          classes: @classes,
+          rng: @rng }
+      end
+
+      # Load marshal data.
+      # @return [nil]
+      def marshal_load(obj)
+        @params = obj[:params]
+        @weight_vec = obj[:weight_vec]
+        @bias_term = obj[:bias_term]
+        @prob_param = obj[:prob_param]
+        @classes = obj[:classes]
+        @rng = obj[:rng]
+        nil
+      end
+
+      private
+
+      def calc_loss_gradient(x, y, weight)
+        target_ids = (x.dot(weight) * y).lt(1.0).where
+        grad = Numo::DFloat.zeros(@params[:batch_size])
+        grad[target_ids] = -y[target_ids]
+        grad
+      end
+    end
+  end
+end

data/lib/rumale/linear_model/svr.rb

@@ -0,0 +1,122 @@
+# frozen_string_literal: true
+
+require 'rumale/linear_model/base_linear_model'
+require 'rumale/base/regressor'
+
+module Rumale
+  module LinearModel
+    # SVR is a class that implements Support Vector Regressor
+    # with mini-batch stochastic gradient descent optimization.
+    #
+    # @example
+    #   estimator =
+    #     Rumale::LinearModel::SVR.new(reg_param: 1.0, epsilon: 0.1, max_iter: 1000, batch_size: 20, random_seed: 1)
+    #   estimator.fit(training_samples, traininig_target_values)
+    #   results = estimator.predict(testing_samples)
+    #
+    # *Reference*
+    # 1. S. Shalev-Shwartz and Y. Singer, "Pegasos: Primal Estimated sub-GrAdient SOlver for SVM," Proc. ICML'07, pp. 807--814, 2007.
+    class SVR < BaseLinearModel
+      include Base::Regressor
+
+      # Return the weight vector for SVR.
+      # @return [Numo::DFloat] (shape: [n_outputs, n_features])
+      attr_reader :weight_vec
+
+      # Return the bias term (a.k.a. intercept) for SVR.
+      # @return [Numo::DFloat] (shape: [n_outputs])
+      attr_reader :bias_term
+
+      # Return the random generator for performing random sampling.
+      # @return [Random]
+      attr_reader :rng
+
+      # Create a new regressor with Support Vector Machine by the SGD optimization.
+      #
+      # @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 epsilon [Float] The margin of tolerance.
+      # @param max_iter [Integer] The maximum number of iterations.
+      # @param batch_size [Integer] The size of the mini batches.
+      # @param optimizer [Optimizer] The optimizer to calculate adaptive learning rate.
+      #   If nil is given, Nadam is used.
+      # @param random_seed [Integer] The seed value using to initialize the random generator.
+      def initialize(reg_param: 1.0, fit_bias: false, bias_scale: 1.0, epsilon: 0.1,
+                     max_iter: 1000, batch_size: 20, optimizer: nil, random_seed: nil)
+        check_params_float(reg_param: reg_param, bias_scale: bias_scale, epsilon: epsilon)
+        check_params_integer(max_iter: max_iter, batch_size: batch_size)
+        check_params_boolean(fit_bias: fit_bias)
+        check_params_type_or_nil(Integer, random_seed: random_seed)
+        check_params_positive(reg_param: reg_param, bias_scale: bias_scale, epsilon: epsilon,
+                              max_iter: max_iter, batch_size: batch_size)
+        keywd_args = method(:initialize).parameters.map { |_t, arg| [arg, binding.local_variable_get(arg)] }.to_h
+        keywd_args.delete(:epsilon)
+        super(keywd_args)
+        @params[:epsilon] = epsilon
+      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::DFloat] (shape: [n_samples, n_outputs]) The target values to be used for fitting the model.
+      # @return [SVR] The learned regressor itself.
+      def fit(x, y)
+        check_sample_array(x)
+        check_tvalue_array(y)
+        check_sample_tvalue_size(x, y)
+
+        n_outputs = y.shape[1].nil? ? 1 : y.shape[1]
+        n_features = x.shape[1]
+
+        if n_outputs > 1
+          @weight_vec = Numo::DFloat.zeros(n_outputs, n_features)
+          @bias_term = Numo::DFloat.zeros(n_outputs)
+          n_outputs.times { |n| @weight_vec[n, true], @bias_term[n] = partial_fit(x, y[true, n]) }
+        else
+          @weight_vec, @bias_term = partial_fit(x, y)
+        end
+
+        self
+      end
+
+      # Predict values for samples.
+      #
+      # @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)
+        check_sample_array(x)
+        x.dot(@weight_vec.transpose) + @bias_term
+      end
+
+      # Dump marshal data.
+      # @return [Hash] The marshal data about SVR.
+      def marshal_dump
+        { params: @params,
+          weight_vec: @weight_vec,
+          bias_term: @bias_term,
+          rng: @rng }
+      end
+
+      # Load marshal data.
+      # @return [nil]
+      def marshal_load(obj)
+        @params = obj[:params]
+        @weight_vec = obj[:weight_vec]
+        @bias_term = obj[:bias_term]
+        @rng = obj[:rng]
+        nil
+      end
+
+      private
+
+      def calc_loss_gradient(x, y, weight)
+        z = x.dot(weight)
+        grad = Numo::DFloat.zeros(@params[:batch_size])
+        grad[(z - y).gt(@params[:epsilon]).where] = 1
+        grad[(y - z).gt(@params[:epsilon]).where] = -1
+        grad
+      end
+    end
+  end
+end

data/lib/rumale/model_selection/cross_validation.rb

@@ -0,0 +1,123 @@
+# frozen_string_literal: true
+
+require 'rumale/validation'
+require 'rumale/base/base_estimator'
+require 'rumale/base/classifier'
+require 'rumale/base/regressor'
+require 'rumale/base/splitter'
+require 'rumale/base/evaluator'
+require 'rumale/evaluation_measure/log_loss'
+
+module Rumale
+  # This module consists of the classes for model validation techniques.
+  module ModelSelection
+    # CrossValidation is a class that evaluates a given classifier with cross-validation method.
+    #
+    # @example
+    #   svc = Rumale::LinearModel::SVC.new
+    #   kf = Rumale::ModelSelection::StratifiedKFold.new(n_splits: 5)
+    #   cv = Rumale::ModelSelection::CrossValidation.new(estimator: svc, splitter: kf)
+    #   report = cv.perform(samples, lables)
+    #   mean_test_score = report[:test_score].inject(:+) / kf.n_splits
+    #
+    class CrossValidation
+      include Validation
+
+      # Return the classifier of which performance is evaluated.
+      # @return [Classifier]
+      attr_reader :estimator
+
+      # Return the splitter that divides dataset.
+      # @return [Splitter]
+      attr_reader :splitter
+
+      # Return the evaluator that calculates score.
+      # @return [Evaluator]
+      attr_reader :evaluator
+
+      # Return the flag indicating whether to caculate the score of training dataset.
+      # @return [Boolean]
+      attr_reader :return_train_score
+
+      # Create a new evaluator with cross-validation method.
+      #
+      # @param estimator [Classifier] The classifier of which performance is evaluated.
+      # @param splitter [Splitter] The splitter that divides dataset to training and testing dataset.
+      # @param evaluator [Evaluator] The evaluator that calculates score of estimator results.
+      # @param return_train_score [Boolean] The flag indicating whether to calculate the score of training dataset.
+      def initialize(estimator: nil, splitter: nil, evaluator: nil, return_train_score: false)
+        check_params_type(Rumale::Base::BaseEstimator, estimator: estimator)
+        check_params_type(Rumale::Base::Splitter, splitter: splitter)
+        check_params_type_or_nil(Rumale::Base::Evaluator, evaluator: evaluator)
+        check_params_boolean(return_train_score: return_train_score)
+        @estimator = estimator
+        @splitter = splitter
+        @evaluator = evaluator
+        @return_train_score = return_train_score
+      end
+
+      # Perform the evalution of given classifier with cross-validation method.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features])
+      #   The dataset to be used to evaluate the estimator.
+      # @param y [Numo::Int32 / Numo::DFloat] (shape: [n_samples] / [n_samples, n_outputs])
+      #   The labels to be used to evaluate the classifier / The target values to be used to evaluate the regressor.
+      # @return [Hash] The report summarizing the results of cross-validation.
+      #   * :fit_time (Array<Float>) The calculation times of fitting the estimator for each split.
+      #   * :test_score (Array<Float>) The scores of testing dataset for each split.
+      #   * :train_score (Array<Float>) The scores of training dataset for each split. This option is nil if
+      #     the return_train_score is false.
+      def perform(x, y)
+        check_sample_array(x)
+        if @estimator.is_a?(Rumale::Base::Classifier)
+          check_label_array(y)
+          check_sample_label_size(x, y)
+        end
+        if @estimator.is_a?(Rumale::Base::Regressor)
+          check_tvalue_array(y)
+          check_sample_tvalue_size(x, y)
+        end
+        # Initialize the report of cross validation.
+        report = { test_score: [], train_score: nil, fit_time: [] }
+        report[:train_score] = [] if @return_train_score
+        # Evaluate the estimator on each split.
+        @splitter.split(x, y).each do |train_ids, test_ids|
+          # Split dataset into training and testing dataset.
+          feature_ids = !kernel_machine? || train_ids
+          train_x = x[train_ids, feature_ids]
+          train_y = y.shape[1].nil? ? y[train_ids] : y[train_ids, true]
+          test_x = x[test_ids, feature_ids]
+          test_y = y.shape[1].nil? ? y[test_ids] : y[test_ids, true]
+          # Fit the estimator.
+          start_time = Time.now.to_i
+          @estimator.fit(train_x, train_y)
+          # Calculate scores and prepare the report.
+          report[:fit_time].push(Time.now.to_i - start_time)
+          if @evaluator.nil?
+            report[:test_score].push(@estimator.score(test_x, test_y))
+            report[:train_score].push(@estimator.score(train_x, train_y)) if @return_train_score
+          elsif log_loss?
+            report[:test_score].push(@evaluator.score(test_y, @estimator.predict_proba(test_x)))
+            report[:train_score].push(@evaluator.score(train_y, @estimator.predict_proba(train_x))) if @return_train_score
+          else
+            report[:test_score].push(@evaluator.score(test_y, @estimator.predict(test_x)))
+            report[:train_score].push(@evaluator.score(train_y, @estimator.predict(train_x))) if @return_train_score
+          end
+        end
+        report
+      end
+
+      private
+
+      def kernel_machine?
+        class_name = @estimator.class.to_s
+        class_name = @estimator.params[:estimator].class.to_s if class_name.include?('Multiclass')
+        class_name.include?('KernelMachine')
+      end
+
+      def log_loss?
+        @evaluator.is_a?(Rumale::EvaluationMeasure::LogLoss)
+      end
+    end
+  end
+end

data/lib/rumale/model_selection/grid_search_cv.rb

@@ -0,0 +1,247 @@
+# frozen_string_literal: true
+
+require 'rumale/validation'
+require 'rumale/base/base_estimator'
+require 'rumale/base/evaluator'
+require 'rumale/base/splitter'
+require 'rumale/pipeline/pipeline'
+
+module Rumale
+  module ModelSelection
+    # GridSearchCV is a class that performs hyperparameter optimization with grid search method.
+    #
+    # @example
+    #   rfc = Rumale::Ensemble::RandomForestClassifier.new(random_seed: 1)
+    #   pg = { n_estimators: [5, 10], max_depth: [3, 5], max_leaf_nodes: [15, 31] }
+    #   kf = Rumale::ModelSelection::StratifiedKFold.new(n_splits: 5)
+    #   gs = Rumale::ModelSelection::GridSearchCV.new(estimator: rfc, param_grid: pg, splitter: kf)
+    #   gs.fit(samples, labels)
+    #   p gs.cv_results
+    #   p gs.best_params
+    #
+    # @example
+    #   rbf = Rumale::KernelApproximation::RBF.new(random_seed: 1)
+    #   svc = Rumale::LinearModel::SVC.new(random_seed: 1)
+    #   pipe = Rumale::Pipeline::Pipeline.new(steps: { rbf: rbf, svc: svc })
+    #   pg = { rbf__gamma: [32.0, 1.0], rbf__n_components: [4, 128], svc__reg_param: [16.0, 0.1] }
+    #   kf = Rumale::ModelSelection::StratifiedKFold.new(n_splits: 5)
+    #   gs = Rumale::ModelSelection::GridSearchCV.new(estimator: pipe, param_grid: pg, splitter: kf)
+    #   gs.fit(samples, labels)
+    #   p gs.cv_results
+    #   p gs.best_params
+    #
+    class GridSearchCV
+      include Base::BaseEstimator
+      include Validation
+
+      # Return the result of cross validation for each parameter.
+      # @return [Hash]
+      attr_reader :cv_results
+
+      # Return the score of the estimator learned with the best parameter.
+      # @return [Float]
+      attr_reader :best_score
+
+      # Return the best parameter set.
+      # @return [Hash]
+      attr_reader :best_params
+
+      # Return the index of the best parameter.
+      # @return [Integer]
+      attr_reader :best_index
+
+      # Return the estimator learned with the best parameter.
+      # @return [Estimator]
+      attr_reader :best_estimator
+
+      # Create a new grid search method.
+      #
+      # @param estimator [Classifier/Regresor] The estimator to be searched for optimal parameters with grid search method.
+      # @param param_grid [Array<Hash>] The parameter sets is represented with array of hash that
+      #   consists of parameter names as keys and array of parameter values as values.
+      # @param splitter [Splitter] The splitter that divides dataset to training and testing dataset on cross validation.
+      # @param evaluator [Evaluator] The evaluator that calculates score of estimator results on cross validation.
+      #   If nil is given, the score method of estimator is used to evaluation.
+      # @param greater_is_better [Boolean] The flag that indicates whether the estimator is better as
+      #   evaluation score is larger.
+      def initialize(estimator: nil, param_grid: nil, splitter: nil, evaluator: nil, greater_is_better: true)
+        check_params_type(Rumale::Base::BaseEstimator, estimator: estimator)
+        check_params_type(Rumale::Base::Splitter, splitter: splitter)
+        check_params_type_or_nil(Rumale::Base::Evaluator, evaluator: evaluator)
+        check_params_boolean(greater_is_better: greater_is_better)
+        @params = {}
+        @params[:param_grid] = valid_param_grid(param_grid)
+        @params[:estimator] = Marshal.load(Marshal.dump(estimator))
+        @params[:splitter] = Marshal.load(Marshal.dump(splitter))
+        @params[:evaluator] = Marshal.load(Marshal.dump(evaluator))
+        @params[:greater_is_better] = greater_is_better
+        @cv_results = nil
+        @best_score = nil
+        @best_params = nil
+        @best_index = nil
+        @best_estimator = nil
+      end
+
+      # Fit the model with given training data and all sets of parameters.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The training data to be used for fitting the model.
+      # @param y [Numo::NArray] (shape: [n_samples, n_outputs]) The target values or labels to be used for fitting the model.
+      # @return [GridSearchCV] The learned estimator with grid search.
+      def fit(x, y)
+        check_sample_array(x)
+
+        init_attrs
+
+        param_combinations.each do |prm_set|
+          prm_set.each do |prms|
+            report = perform_cross_validation(x, y, prms)
+            store_cv_result(prms, report)
+          end
+        end
+
+        find_best_params
+
+        @best_estimator = configurated_estimator(@best_params)
+        @best_estimator.fit(x, y)
+        self
+      end
+
+      # Call the decision_function method of learned estimator with the best parameter.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to compute the scores.
+      # @return [Numo::DFloat] (shape: [n_samples]) Confidence score per sample.
+      def decision_function(x)
+        check_sample_array(x)
+        @best_estimator.decision_function(x)
+      end
+
+      # Call the predict method of learned estimator with the best parameter.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to obtain prediction result.
+      # @return [Numo::NArray] Predicted results.
+      def predict(x)
+        check_sample_array(x)
+        @best_estimator.predict(x)
+      end
+
+      # Call the predict_log_proba method of learned estimator with the best parameter.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to predict the log-probailities.
+      # @return [Numo::DFloat] (shape: [n_samples, n_classes]) Predicted log-probability of each class per sample.
+      def predict_log_proba(x)
+        check_sample_array(x)
+        @best_estimator.predict_log_proba(x)
+      end
+
+      # Call the predict_proba method of learned estimator with the best parameter.
+      #
+      # @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)
+        check_sample_array(x)
+        @best_estimator.predict_proba(x)
+      end
+
+      # Call the score method of learned estimator with the best parameter.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) Testing data.
+      # @param y [Numo::NArray] (shape: [n_samples, n_outputs]) True target values or labels for testing data.
+      # @return [Float] The score of estimator.
+      def score(x, y)
+        check_sample_array(x)
+        @best_estimator.score(x, y)
+      end
+
+      # Dump marshal data.
+      # @return [Hash] The marshal data about GridSearchCV.
+      def marshal_dump
+        { params: @params,
+          cv_results: @cv_results,
+          best_score: @best_score,
+          best_params: @best_params,
+          best_index: @best_index,
+          best_estimator: @best_estimator }
+      end
+
+      # Load marshal data.
+      # @return [nil]
+      def marshal_load(obj)
+        @params = obj[:params]
+        @cv_results = obj[:cv_results]
+        @best_score = obj[:best_score]
+        @best_params = obj[:best_params]
+        @best_index = obj[:best_index]
+        @best_estimator = obj[:best_estimator]
+        nil
+      end
+
+      private
+
+      def valid_param_grid(grid)
+        raise TypeError, 'Expect class of param_grid to be Hash or Array' unless grid.is_a?(Hash) || grid.is_a?(Array)
+        grid = [grid] if grid.is_a?(Hash)
+        grid.each do |h|
+          raise TypeError, 'Expect class of elements in param_grid to be Hash' unless h.is_a?(Hash)
+          raise TypeError, 'Expect class of parameter values in param_grid to be Array' unless h.values.all? { |v| v.is_a?(Array) }
+        end
+        grid
+      end
+
+      def param_combinations
+        @param_combinations ||= @params[:param_grid].map do |prm|
+          x = Hash[prm.sort].map { |k, v| [k].product(v) }
+          x[0].product(*x[1...x.size]).map { |v| Hash[v] }
+        end
+      end
+
+      def perform_cross_validation(x, y, prms)
+        est = configurated_estimator(prms)
+        cv = CrossValidation.new(estimator: est, splitter: @params[:splitter],
+                                 evaluator: @params[:evaluator], return_train_score: true)
+        cv.perform(x, y)
+      end
+
+      def configurated_estimator(prms)
+        estimator = Marshal.load(Marshal.dump(@params[:estimator]))
+        if @params[:estimator].is_a?(Rumale::Pipeline::Pipeline)
+          prms.each do |k, v|
+            est_name, prm_name = k.to_s.split('__')
+            estimator.steps[est_name.to_sym].params[prm_name.to_sym] = v
+          end
+        else
+          prms.each { |k, v| estimator.params[k] = v }
+        end
+        estimator
+      end
+
+      def init_attrs
+        @cv_results = %i[mean_test_score std_test_score
+                         mean_train_score std_train_score
+                         mean_fit_time std_fit_time params].map { |v| [v, []] }.to_h
+        @best_score = nil
+        @best_params = nil
+        @best_index = nil
+        @best_estimator = nil
+      end
+
+      def store_cv_result(prms, report)
+        test_scores = Numo::DFloat[*report[:test_score]]
+        train_scores = Numo::DFloat[*report[:train_score]]
+        fit_times = Numo::DFloat[*report[:fit_time]]
+        @cv_results[:mean_test_score].push(test_scores.mean)
+        @cv_results[:std_test_score].push(test_scores.stddev)
+        @cv_results[:mean_train_score].push(train_scores.mean)
+        @cv_results[:std_train_score].push(train_scores.stddev)
+        @cv_results[:mean_fit_time].push(fit_times.mean)
+        @cv_results[:std_fit_time].push(fit_times.stddev)
+        @cv_results[:params].push(prms)
+      end
+
+      def find_best_params
+        @best_score = @params[:greater_is_better] ? @cv_results[:mean_test_score].max : @cv_results[:mean_test_score].min
+        @best_index = @cv_results[:mean_test_score].index(@best_score)
+        @best_params = @cv_results[:params][@best_index]
+      end
+    end
+  end
+end