Class: SVMKit::EvaluationMeasure::NormalizedMutualInformation

Inherits:
Object
  • Object
show all
Includes:
Base::Evaluator
Defined in:
lib/svmkit/evaluation_measure/normalized_mutual_information.rb

Overview

NormalizedMutualInformation is a class that calculates the normalized mutual information of cluatering results.

Reference

  • C D. Manning, P. Raghavan, and H. Schutze, “Introduction to Information Retrieval,” Cambridge University Press., 2008.

  • N X. Vinh, J. Epps, and J. Bailey, “Information Theoretic Measures for Clusterings Comparison: Variants, Properties, Normalization and Correction for Chance,” J. Machine Learning Research, vol. 11, pp. 2837–1854, 2010.

Examples:

evaluator = SVMKit::EvaluationMeasure::NormalizedMutualInformation.new
puts evaluator.score(ground_truth, predicted)

Instance Method Summary collapse

Instance Method Details

#score(y_true, y_pred) ⇒ Float

Calculate noramlzied mutual information

Parameters:

  • y_true (Numo::Int32) —

    (shape: [n_samples]) Ground truth labels.

  • y_pred (Numo::Int32) —

    (shape: [n_samples]) Predicted cluster labels.

Returns:

  • (Float) —

    Normalized mutual information



25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
 
# File 'lib/svmkit/evaluation_measure/normalized_mutual_information.rb', line 25

def score(y_true, y_pred)
  SVMKit::Validation.check_label_array(y_true)
  SVMKit::Validation.check_label_array(y_pred)
  # initiazlie some variables.
  mutual_information = 0.0
  n_samples = y_pred.size
  class_ids = y_true.to_a.uniq
  cluster_ids = y_pred.to_a.uniq
  # calculate entropy.
  class_entropy = -1.0 * class_ids.map do |k|
    ratio = y_true.eq(k).count.fdiv(n_samples)
    ratio * Math.log(ratio)
  end.reduce(:+)
  return 0.0 if class_entropy.zero?
  cluster_entropy = -1.0 * cluster_ids.map do |k|
    ratio = y_pred.eq(k).count.fdiv(n_samples)
    ratio * Math.log(ratio)
  end.reduce(:+)
  return 0.0 if cluster_entropy.zero?
  # calculate mutual information.
  cluster_ids.map do |k|
    pr_sample_ids = y_pred.eq(k).where.to_a
    n_pr_samples = pr_sample_ids.size
    class_ids.map do |j|
      tr_sample_ids = y_true.eq(j).where.to_a
      n_tr_samples = tr_sample_ids.size
      n_intr_samples = (pr_sample_ids & tr_sample_ids).size
      if n_intr_samples > 0
        mutual_information +=
          n_intr_samples.fdiv(n_samples) * Math.log((n_samples * n_intr_samples).fdiv(n_pr_samples * n_tr_samples))
      end
    end
  end
  # return normalized mutual information.
  mutual_information / Math.sqrt(class_entropy * cluster_entropy)
end