ruby-em_algorithm 0.0.2

data/example/ex1.rb

@@ -0,0 +1,20 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+
+data_array = YAML.load_file(ARGV[0])
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 9.0)],
+#  :weights => [1.0]
+#)
+model = Mixture.new(
+  :models => [Gaussian.new(0.0, 9.0), Gaussian.new(10.0, 9.0)],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array)
+em.run!
+puts em.model.to_gnuplot

data/example/ex1.rb~

@@ -0,0 +1,20 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+
+data_array = YAML.load_file(ARGV[0])
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 9.0)],
+#  :weights => [1.0]
+#)
+model = Mixture.new(
+  :models => [Gaussian.new(0.0, 9.0), Gaussian.new(10.0, 9.0)],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array)
+em.run!
+em.model.to_gnuplot

data/example/ex2.rb

@@ -0,0 +1,33 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file(ARGV[0]).map {|v| GSL::Vector[v]}
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 3.0)],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[0.0, 0.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[10.0, 10.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[-10.0, -10.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[18.0, 18.0], Matrix[[1.0, 0.0], [0.0, 1.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[-2.0, -2.0], Matrix[[1.0, 0.0], [0.0, 1.0]], 1.0), MdGaussian.new(Vector[2.0, 2.0], Matrix[[1.0, 0.0], [0.0, 1.0]], 1.0)],
+#  :weights => [0.5, 0.5]
+#)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array)
+em.run!
+puts em.model.to_gnuplot

data/example/ex2.rb~

@@ -0,0 +1,33 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file(ARGV[0]).map {|v| GSL::Vector[v]}
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 3.0)],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[0.0, 0.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[10.0, 10.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[-10.0, -10.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[18.0, 18.0], Matrix[[1.0, 0.0], [0.0, 1.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[-2.0, -2.0], Matrix[[1.0, 0.0], [0.0, 1.0]], 1.0), MdGaussian.new(Vector[2.0, 2.0], Matrix[[1.0, 0.0], [0.0, 1.0]], 1.0)],
+#  :weights => [0.5, 0.5]
+#)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array)
+em.run!
+puts em.model.to_gnuplot

data/example/ex3-tmp.rb

@@ -0,0 +1,23 @@
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file("example/data/2dim-gmm-test.txt").map {|v| GSL::Vector[v]}
+
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[-5.0, -5.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[5.0, 5.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array,
+                           #:convergence_check => "ChiSquare",
+                           :convergence_check => "Likelihood",
+                           :estimate_as_frequency => 2,
+                           :debug => false)
+em.run!
+puts em.model.to_gnuplot(:mixture_only)
+p em.distribution_to_value_ratio
+chi_square = ChiSquare.new(data_array)
+chi_square.calculate(em.model, em.distribution_to_value_ratio[0])
+

data/example/ex3-tmp.rb~

@@ -0,0 +1,25 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file("example/data/2dim-gmm-test.txt").map {|v| GSL::Vector[v]}
+
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[-5.0, -5.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[5.0, 5.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array,
+                           #:convergence_check => "ChiSquare",
+                           :convergence_check => "Likelihood",
+                           :estimate_as_frequency => 2,
+                           :debug => false)
+em.run!
+puts em.model.to_gnuplot(:mixture_only)
+p em.distribution_to_value_ratio
+chi_square = ChiSquare.new(data_array)
+chi_square.calculate(em.model, em.distribution_to_value_ratio[0])
+

data/example/ex3.rb

@@ -0,0 +1,43 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file(ARGV[0]).map {|v| GSL::Vector[v]}
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 3.0)],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[10.0, 10.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[1.0, 1.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[1.24102564102564, 1.3843589743589746], Matrix[[4.114512820512822, 1.095717948717949], [1.095717948717949, 3.558512820512825]]), MdGaussian.new(Vector[-0.9518604651162781, -2.0293023255813956], Matrix[[3.4144651162790747, 1.764395348837209], [1.764395348837209, 6.793999999999991]])],
+#  :weights => [0.5, 0.5]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[-3.0, -4.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[2.0, 2.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [0.5, 0.5]
+#)
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[-5.0, -5.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[5.0, 5.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array,
+                           :value_distribution_estimation => true,
+                           :debug => true
+                          )
+em.run!
+#puts em.model.to_gnuplot(:mixture_only)
+#puts em.model.to_gnuplot
+puts em.model.value_distribution_to_gnuplot(em.const)
+

data/example/ex3.rb~

@@ -0,0 +1,43 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'ruby-em_algorithm'
+include EMAlgorithm
+include GSL
+
+data_array = YAML.load_file(ARGV[0]).map {|v| GSL::Vector[v]}
+
+#model = Mixture.new(
+#  :models => [Gaussian.new(0.0, 3.0)],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[10.0, 10.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[1.0, 1.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [1.0]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[1.24102564102564, 1.3843589743589746], Matrix[[4.114512820512822, 1.095717948717949], [1.095717948717949, 3.558512820512825]]), MdGaussian.new(Vector[-0.9518604651162781, -2.0293023255813956], Matrix[[3.4144651162790747, 1.764395348837209], [1.764395348837209, 6.793999999999991]])],
+#  :weights => [0.5, 0.5]
+#)
+#model = Mixture.new(
+#  :models => [MdGaussian.new(Vector[-3.0, -4.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[2.0, 2.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+#  :weights => [0.5, 0.5]
+#)
+model = Mixture.new(
+  :models => [MdGaussian.new(Vector[-5.0, -5.0], Matrix[[9.0, 0.0], [0.0, 9.0]]), MdGaussian.new(Vector[5.0, 5.0], Matrix[[9.0, 0.0], [0.0, 9.0]])],
+  :weights => [0.5, 0.5]
+)
+
+em = EMAlgorithm::Base.new(:model => model, :data_array => data_array,
+                           :estimate_as_frequency => true,
+                           :debug => true
+                          )
+em.run!
+#puts em.model.to_gnuplot(:mixture_only)
+#puts em.model.to_gnuplot
+puts em.model.value_distribution_to_gnuplot(em.const)
+

data/example/tools/2dim.rb

@@ -0,0 +1,69 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'gsl'
+include GSL
+
+def bivariate_gaussian(number, mu_x, mu_y, sigma_x, sigma_y, rho, seed)
+  r = Rng.alloc(Rng::TAUS, seed)
+  array_x = []
+  array_y = []
+  array_xy = []
+  number.times do
+    x, y = r.bivariate_gaussian(sigma_x, sigma_y, rho)
+    array_x << x + mu_x
+    array_y << y + mu_y
+    array_xy << [x + mu_x, y + mu_y]
+  end
+
+  v_x = Vector.alloc(array_x)
+  v_y = Vector.alloc(array_y)
+
+  if ARGV[0] == "without_weight"
+    return array_xy
+  end
+
+  #h = Histogram2d.alloc(30, [-6, 6], 30, [-6, 6])
+  h = Histogram2d.alloc(100, [-10, 10], 100, [-10, 10])
+  #h = Histogram2d.alloc(256, [-6, 6], 256, [-6, 6])
+  h.fill(v_x, v_y)
+  data_array = []
+  (0..(h.nx-1)).each do |x|
+    (0..(h.ny-1)).each do |y|
+      data_array << [(h.xrange[x]+h.xrange[x+1])/2.0, (h.yrange[y]+h.yrange[y+1])/2.0, h[x,y].to_f, (h.xrange[x]-h.xrange[x+1]).abs*(h.yrange[y]-h.yrange[y+1]).abs]
+      #p [(h.xrange[x]+h.xrange[x+1])/2.0, (h.yrange[y]+h.yrange[y+1])/2.0, h[x,y].to_f, (h.xrange[x]-h.xrange[x+1]).abs*(h.yrange[y]-h.yrange[y+1]).abs]
+    end
+  end
+  data_array
+end
+
+#data_array = bivariate_gaussian(24000, 1.0, 1.0, 1.0, 1.0, 0, 1)
+
+data_array = bivariate_gaussian(24000, 1, 1, 1, 1, 0, 1)
+data_array += bivariate_gaussian(12000, -1, -2, 1, 1, 0, 1)
+
+#data_array = bivariate_gaussian(2400, 1, 1, 1, 1, 0, 1)
+#data_array += bivariate_gaussian(1200, -1, -2, 1, 1, 0, 1)
+
+#data_array = bivariate_gaussian(6000, 1, 1, 1, 1, 0, 1)
+#data_array += bivariate_gaussian(3000, -1, -2, 1, 1, 0, 1)
+if ARGV[0] == "without_weight"
+  puts data_array.to_yaml
+  exit(0)
+end
+data_array = data_array.map do |datum|
+  if rand > 1.0
+    [datum[0], datum[1], 0.0, 0.0]
+  else
+    [datum[0], datum[1], datum[2], datum[3]]
+  end
+end
+if ARGV[0] == "gnuplot"
+  data_array.each do |datum|
+    next if datum[2] == 0.0
+    puts "#{datum[0]} #{datum[1]} #{datum[2]}"
+  end
+elsif ARGV[0].nil?
+  data_array = data_array.reject {|datum| datum[2] == 0.0}
+  puts data_array.to_yaml
+end

data/example/tools/2dim.rb~

@@ -0,0 +1,69 @@
+#!/usr/bin/env ruby
+
+require 'yaml'
+require 'gsl'
+include GSL
+
+def bivariate_gaussian(number, mu_x, mu_y, sigma_x, sigma_y, rho, seed)
+  r = Rng.alloc(Rng::TAUS, seed)
+  array_x = []
+  array_y = []
+  array_xy = []
+  number.times do
+    x, y = r.bivariate_gaussian(sigma_x, sigma_y, rho)
+    array_x << x + mu_x
+    array_y << y + mu_y
+    array_xy << [x + mu_x, y + mu_y]
+  end
+
+  v_x = Vector.alloc(array_x)
+  v_y = Vector.alloc(array_y)
+
+  if ARGV[0] == "without_weight"
+    return array_xy
+  end
+
+  #h = Histogram2d.alloc(30, [-6, 6], 30, [-6, 6])
+  h = Histogram2d.alloc(100, [-10, 10], 100, [-10, 10])
+  #h = Histogram2d.alloc(256, [-6, 6], 256, [-6, 6])
+  h.fill(v_x, v_y)
+  data_array = []
+  (0..(h.nx-1)).each do |x|
+    (0..(h.ny-1)).each do |y|
+      data_array << [(h.xrange[x]+h.xrange[x+1])/2.0, (h.yrange[y]+h.yrange[y+1])/2.0, h[x,y].to_f, (h.xrange[x]-h.xrange[x+1]).abs*(h.yrange[y]-h.yrange[y+1]).abs]
+      #p [(h.xrange[x]+h.xrange[x+1])/2.0, (h.yrange[y]+h.yrange[y+1])/2.0, h[x,y].to_f, (h.xrange[x]-h.xrange[x+1]).abs*(h.yrange[y]-h.yrange[y+1]).abs]
+    end
+  end
+  data_array
+end
+
+#data_array = bivariate_gaussian(24000, 1.0, 1.0, 1.0, 1.0, 0, 1)
+
+data_array = bivariate_gaussian(24000, 1, 1, 1, 1, 0, 1)
+data_array += bivariate_gaussian(12000, -1, -2, 1, 1, 0, 1)
+
+#data_array = bivariate_gaussian(2400, 1, 1, 1, 1, 0, 1)
+#data_array += bivariate_gaussian(1200, -1, -2, 1, 1, 0, 1)
+
+#data_array = bivariate_gaussian(6000, 1, 1, 1, 1, 0, 1)
+#data_array += bivariate_gaussian(3000, -1, -2, 1, 1, 0, 1)
+if ARGV[0] == "without_weight"
+  puts data_array.to_yaml
+  exit(0)
+end
+data_array = data_array.map do |datum|
+  if rand > 1.0
+    [datum[0], datum[1], 0.0, 0.0]
+  else
+    [datum[0], datum[1], datum[2], datum[3]]
+  end
+end
+if ARGV[0] == "gnuplot"
+  data_array.each do |datum|
+    next if datum[2] == 0.0
+    puts "#{datum[0]} #{datum[1]} #{datum[2]}"
+  end
+elsif ARGV[0].nil?
+  data_array = data_array.reject {|datum| datum[2] == 0.0}
+  puts data_array.to_yaml
+end

data/example/tools/boxmuller.rb

@@ -0,0 +1,28 @@
+#!/usr/bin/env ruby
+# -*- coding: utf-8 -*-
+##ボックスミュラー法によるガウス分布データ生成
+##偶数個しか作れない仕様であることに注意
+
+require 'yaml'
+
+include Math
+def boxmuller(n,mu,sigma)
+  j=0
+  data_array = []
+  while j < n
+    a=rand()
+    b=rand()
+    r1=sqrt(-2*Math.log(a.to_f)).to_f*sin(2*Math::PI*b.to_f).to_f
+    r2=sqrt(-2*Math.log(a.to_f)).to_f*cos(2*Math::PI*b.to_f).to_f
+    r3=r1.to_f*sigma.to_f+mu.to_f
+    r4=r2.to_f*sigma.to_f+mu.to_f
+    data_array << r3
+    data_array << r4
+    j=j+2
+  end
+  data_array
+end
+
+data_array = boxmuller(7000, 5, 4.0)
+data_array = data_array + boxmuller(3000, -5, 1.0)
+puts data_array.to_yaml

data/example/tools/boxmuller.rb~

@@ -0,0 +1,28 @@
+#!/usr/bin/env ruby
+# -*- coding: utf-8 -*-
+##ボックスミュラー法によるガウス分布データ生成
+##偶数個しか作れない仕様であることに注意
+
+require 'yaml'
+
+include Math
+def boxmuller(n,mu,sigma)
+  j=0
+  data_array = []
+  while j < n
+    a=rand()
+    b=rand()
+    r1=sqrt(-2*Math.log(a.to_f)).to_f*sin(2*Math::PI*b.to_f).to_f
+    r2=sqrt(-2*Math.log(a.to_f)).to_f*cos(2*Math::PI*b.to_f).to_f
+    r3=r1.to_f*sigma.to_f+mu.to_f
+    r4=r2.to_f*sigma.to_f+mu.to_f
+    data_array << r3
+    data_array << r4
+    j=j+2
+  end
+  data_array
+end
+
+data_array = boxmuller(7000, 5, 4.0)
+#data_array = data_array + boxmuller(3000, -5, 1.0)
+puts data_array.to_yaml