rumale 0.17.1 → 0.17.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 54719fad09a9036dbdc1430323272af7f3c3e746aecac9b6199eef70bfe07856
-  data.tar.gz: b9df7b5ff0cc4feebb053955aea4d022e72263be1288578906b67f2100b97160
+  metadata.gz: '09585216da740231269c5675c48a8fa2ee3a8aba3f5c5b7d671429d113aaa4ba'
+  data.tar.gz: 66eb473c718f6a03938db19df06694373c781e098d68e2f2c20162865dc4f9f8
 SHA512:
-  metadata.gz: 42aedd744761fb61e6fcfa32643ee56b267ea163a2376f3eb308b181cd5981ceef4270e88ed1fe322e5f198d94b3b34af277f246e99f0439cfc3699c3838d76e
-  data.tar.gz: '082dda62d97b2655413ef599185b38c88cceacdec3ceecb99aab67b222306b5478963dfa7e216daa56838c6c3a9ac65ced53e51d513ae23ddf0954fb2c5982e5'
+  metadata.gz: d68eaa297116c4b834cbf2745355d07e4e4b0e50b170c631c417f623d7a0e75a515389903b6e2cb6f22355f9cf5eded4968eac5c7f216f5a98698bb5283c2e00
+  data.tar.gz: ddb3be08dc88cc99f5c7086e8bea83496f2644316d382f052a954ce6d1056bc2aabfbd2664ff572a55fdb1017a9bbb0d790d4d83b3788b781d133a97d26bd92e

data/CHANGELOG.md

@@ -1,3 +1,7 @@
+# 0.17.2
+- Add transformer class for kernel approximation with [Nystroem](https://yoshoku.github.io/rumale/doc/Rumale/KernelApproximation/Nystroem.html) method.
+- Delete array validation on [Pipeline](https://yoshoku.github.io/rumale/doc/Rumale/Pipeline/Pipeline.html) class considering that array of hash is given to HashVectorizer.
+
 # 0.17.1
 - Add transformer class for [PolynomialFeatures](https://yoshoku.github.io/rumale/doc/Rumale/Preprocessing/PolynomialFeatures.html)
 - Add verbose and tol parameter to [FactorizationMachineClassifier](https://yoshoku.github.io/rumale/doc/Rumale/PolynomialModel/FactorizationMachineClassifier.html)

data/README.md

@@ -50,22 +50,21 @@ First, let's classify simple xor data.
 require 'rumale'
 
 # Prepare XOR data.
-features = [[0, 0], [0, 1], [1, 0], [1, 1]]
+samples = [[0, 0], [0, 1], [1, 0], [1, 1]]
 labels = [0, 1, 1, 0]
 
 # Train classifier with nearest neighbor rule.
 estimator = Rumale::NearestNeighbors::KNeighborsClassifier.new(n_neighbors: 1)
-estimator.fit(x, y)
+estimator.fit(samples, labels)
 
 # Predict labels.
-p y
-p estimator.predict(x)
+p labels
+p estimator.predict(samples)
 ```
 
 Execution of the above script result in the following.
 
 ```ruby
-Numo::Int32#shape=[4]
 [0, 1, 1, 0]
 Numo::Int32#shape=[4]
 [0, 1, 1, 0]

data/lib/rumale.rb

@@ -26,6 +26,7 @@ require 'rumale/optimizer/nadam'
 require 'rumale/optimizer/yellow_fin'
 require 'rumale/pipeline/pipeline'
 require 'rumale/kernel_approximation/rbf'
+require 'rumale/kernel_approximation/nystroem'
 require 'rumale/linear_model/base_linear_model'
 require 'rumale/linear_model/base_sgd'
 require 'rumale/linear_model/svc'

data/lib/rumale/kernel_approximation/nystroem.rb

@@ -0,0 +1,106 @@
+# frozen_string_literal: true
+
+require 'rumale/base/base_estimator'
+require 'rumale/base/transformer'
+require 'rumale/pairwise_metric'
+
+module Rumale
+  module KernelApproximation
+    # Nystroem is a class that implements feature mapping with Nystroem method.
+    #
+    # @example
+    #   require 'numo/linalg/autoloader'
+    #
+    #   transformer = Rumale::KernelApproximation::Nystroem.new(gamma: 1, n_components: 128, random_seed: 1)
+    #   new_training_samples = transformer.fit_transform(training_samples)
+    #   new_testing_samples = transformer.transform(testing_samples)
+    #
+    # *Reference*
+    # 1. T. Yang, Y. Li, M. Mahdavi, R. Jin, and Z-H. Zhou, "Nystrom Method vs Random Fourier Features: A Theoretical and Empirical Comparison," Advances in NIPS'12, Vol. 1, pp. 476--484, 2012.
+    class Nystroem
+      include Base::BaseEstimator
+      include Base::Transformer
+
+      # Returns the randomly sampled training data for feature mapping.
+      # @return [Numo::DFloat] (shape: n_components, n_features])
+      attr_reader :components
+
+      # Returns the indices sampled training data.
+      # @return [Numo::Int32] (shape: [n_components])
+      attr_reader :component_indices
+
+      # Returns the normalizing factors.
+      # @return [Numo::DFloat] (shape: [n_components, n_components])
+      attr_reader :normalizer
+
+      # Return the random generator for transformation.
+      # @return [Random]
+      attr_reader :rng
+
+      # Create a new transformer for mapping to kernel feature space with Nystrom method.
+      #
+      # @param kernel [String] The type of kernel. This parameter is ignored in the current implementation.
+      # @param gamma [Float] The parameter of RBF kernel: exp(-gamma * x^2).
+      # @param n_components [Integer] The number of dimensions of the RBF kernel feature space.
+      # @param random_seed [Integer] The seed value using to initialize the random generator.
+      def initialize(kernel: 'rbf', gamma: 1, n_components: 100, random_seed: nil)
+        check_params_numeric(gamma: gamma, n_components: n_components)
+        check_params_numeric_or_nil(random_seed: random_seed)
+        @params = method(:initialize).parameters.map { |_t, arg| [arg, binding.local_variable_get(arg)] }.to_h
+        @params[:random_seed] ||= srand
+        @rng = Random.new(@params[:random_seed])
+        @component_indices = nil
+        @components = nil
+        @normalizer = nil
+      end
+
+      # Fit the model with given training data.
+      #
+      # @overload fit(x) -> RBF
+      #   @param x [Numo::NArray] (shape: [n_samples, n_features]) The training data to be used for fitting the model.
+      # @return [Nystroem] The learned transformer itself.
+      def fit(x, _y = nil)
+        x = check_convert_sample_array(x)
+        raise 'Nystroem#fit requires Numo::Linalg but that is not loaded.' unless enable_linalg?
+
+        # initialize some variables.
+        sub_rng = @rng.dup
+        n_samples = x.shape[0]
+        n_components = [1, [@params[:n_components], n_samples].min].max
+
+        # random sampling.
+        @component_indices = Numo::Int32.cast([*0...n_samples].shuffle(random: sub_rng)[0...n_components])
+        @components = x[@component_indices, true]
+
+        # calculate normalizing factor.
+        kernel_mat = Rumale::PairwiseMetric.rbf_kernel(@components, nil, @params[:gamma])
+        eig_vals, eig_vecs = Numo::Linalg.eigh(kernel_mat)
+        la = eig_vals.class.maximum(eig_vals.reverse, 1e-12)
+        u = eig_vecs.reverse(1)
+        @normalizer = u.dot((1.0 / Numo::NMath.sqrt(la)).diag)
+
+        self
+      end
+
+      # Fit the model with training data, and then transform them with the learned model.
+      #
+      # @overload fit_transform(x) -> Numo::DFloat
+      #   @param x [Numo::DFloat] (shape: [n_samples, n_features]) The training data to be used for fitting the model.
+      # @return [Numo::DFloat] (shape: [n_samples, n_components]) The transformed data
+      def fit_transform(x, _y = nil)
+        x = check_convert_sample_array(x)
+        fit(x).transform(x)
+      end
+
+      # Transform the given data with the learned model.
+      #
+      # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The data to be transformed with the learned model.
+      # @return [Numo::DFloat] (shape: [n_samples, n_components]) The transformed data.
+      def transform(x)
+        x = check_convert_sample_array(x)
+        z = Rumale::PairwiseMetric.rbf_kernel(x, @components, @params[:gamma])
+        z.dot(@normalizer)
+      end
+    end
+  end
+end

data/lib/rumale/pipeline/pipeline.rb

@@ -40,7 +40,6 @@ module Rumale
       # @param y [Numo::NArray] (shape: [n_samples, n_outputs]) The target values or labels to be used for fitting the model.
       # @return [Pipeline] The learned pipeline itself.
       def fit(x, y)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x, y, fit: true)
         last_estimator&.fit(trans_x, y)
         self
@@ -52,7 +51,6 @@ module Rumale
       # @param y [Numo::NArray] (shape: [n_samples, n_outputs], default: nil) The target values or labels to be used for fitting the model.
       # @return [Numo::NArray] The predicted results by last estimator.
       def fit_predict(x, y = nil)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x, y, fit: true)
         last_estimator.fit_predict(trans_x)
       end
@@ -63,7 +61,6 @@ module Rumale
       # @param y [Numo::NArray] (shape: [n_samples, n_outputs], default: nil) The target values or labels to be used for fitting the model.
       # @return [Numo::NArray] The predicted results by last estimator.
       def fit_transform(x, y = nil)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x, y, fit: true)
         last_estimator.fit_transform(trans_x, y)
       end
@@ -73,7 +70,6 @@ module Rumale
       # @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)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.decision_function(trans_x)
       end
@@ -83,7 +79,6 @@ module Rumale
       # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to obtain prediction result.
       # @return [Numo::NArray] The predicted results by last estimator.
       def predict(x)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.predict(trans_x)
       end
@@ -93,7 +88,6 @@ module Rumale
       # @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)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.predict_log_proba(trans_x)
       end
@@ -103,7 +97,6 @@ module Rumale
       # @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 = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.predict_proba(trans_x)
       end
@@ -113,7 +106,6 @@ module Rumale
       # @param x [Numo::DFloat] (shape: [n_samples, n_features]) The samples to be transformed.
       # @return [Numo::DFloat] (shape: [n_samples, n_components]) The transformed samples.
       def transform(x)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.nil? ? trans_x : last_estimator.transform(trans_x)
       end
@@ -123,7 +115,6 @@ module Rumale
       # @param z [Numo::DFloat] (shape: [n_samples, n_components]) The transformed samples to be restored into original space.
       # @return [Numo::DFloat] (shape: [n_samples, n_featuress]) The restored samples.
       def inverse_transform(z)
-        z = check_convert_sample_array(z)
         itrans_z = z
         @steps.keys.reverse_each do |name|
           transformer = @steps[name]
@@ -139,26 +130,10 @@ module Rumale
       # @param y [Numo::NArray] (shape: [n_samples, n_outputs]) True target values or labels for testing data.
       # @return [Float] The score of last estimator
       def score(x, y)
-        x = check_convert_sample_array(x)
         trans_x = apply_transforms(x)
         last_estimator.score(trans_x, y)
       end
 
-      # Dump marshal data.
-      # @return [Hash] The marshal data about Pipeline.
-      def marshal_dump
-        { params: @params,
-          steps: @steps }
-      end
-
-      # Load marshal data.
-      # @return [nil]
-      def marshal_load(obj)
-        @params = obj[:params]
-        @steps = obj[:steps]
-        nil
-      end
-
       private
 
       def validate_steps(steps)

data/lib/rumale/version.rb

@@ -3,5 +3,5 @@
 # Rumale is a machine learning library in Ruby.
 module Rumale
   # The version of Rumale you are using.
-  VERSION = '0.17.1'
+  VERSION = '0.17.2'
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: rumale
 version: !ruby/object:Gem::Version
-  version: 0.17.1
+  version: 0.17.2
 platform: ruby
 authors:
 - yoshoku
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2020-01-26 00:00:00.000000000 Z
+date: 2020-02-01 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: numo-narray
@@ -209,6 +209,7 @@ files:
 - lib/rumale/evaluation_measure/silhouette_score.rb
 - lib/rumale/feature_extraction/feature_hasher.rb
 - lib/rumale/feature_extraction/hash_vectorizer.rb
+- lib/rumale/kernel_approximation/nystroem.rb
 - lib/rumale/kernel_approximation/rbf.rb
 - lib/rumale/kernel_machine/kernel_pca.rb
 - lib/rumale/kernel_machine/kernel_ridge.rb