t_learn 0.1.1.3 → 0.1.1.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 4997f379b3ef4527598a83ceb94ab6620d61b886
-  data.tar.gz: cca4c6643f596a70d69b1a79a86ca94189bcce0d
+  metadata.gz: b6becf413fe48eb9531615a4f25124b195f2d398
+  data.tar.gz: 574db0c50bbb483001f7e4679c1677c1a9f91990
 SHA512:
-  metadata.gz: 1951ea181fb690965373c0444af8c3d3682b2bdf17889baf7a4aa6d84f57701a5a4211916a1034f83aef801b6101ba64a1da1a5c2f43fa67f3b605dc8c807dd7
-  data.tar.gz: fc1ddcaa6a905309de2e8729fd96f44c11ac04dc8c652287d8c5e805957af42af3206c627276767557b4e1f6ddd1889ae4568834fcb0023d87743942033cbbd4
+  metadata.gz: ea796e2c06c35b66da56160029464d675164767e97f1b5454b63330304e4f9c799a66758fdca5aec5d5c93b90136e4c8e38e8e8e6764706ba4e8c66f899fa80e
+  data.tar.gz: 6978a68150c212043a9c09988a209eca37473040bbe3f534f53ef02dc35874e9a1a795e9f59a93735a58b5a806e983a15c7018558aed1a24c566e81317b5f71f

data/.rake_tasks~

@@ -0,0 +1,7 @@
+build
+clean
+clobber
+install
+install:local
+release[remote]
+spec

data/Gemfile

@@ -2,3 +2,8 @@ source 'https://rubygems.org'
 
 # Specify your gem's dependencies in t_learn.gemspec
 gemspec
+
+group :development, :test do
+  gem 'pry-rails'
+  gem 'matrix'
+end

data/lib/t_learn/em.rb

@@ -2,117 +2,235 @@
 # -*- encoding: utf-8 -*-
 
 require "yaml"
+require "json"
+require "pry"
 
 module TLearn
-  class EM
-    attr_accessor :mu_list, :var_list, :pi_list, :datas, :log_likelihood, :burden_ratio, :k_num_
-    , :n, :r_list
-    
-    def initialize
-      @data = Array.new(10, 0.0)  # load_data
-      @mu_list = Array.new(10, 0.0)
-      @var_list = Array.new(10, 0.0)
-      @r_list = Array.new(10, 0.0)
-      @n_list = Array.new(10, 0.0)
-      @k_num = 2
+
+  class EM_Gaussian
+    attr_accessor :mu_list, :conv_list, :pi_list, :log_likelihood, :k_num,:data_list, :real_data_list
+
+    def init(data_list, k_num)
+      @k_num = k_num # ガウス分布の数
+      @data_list = data_list
+      @dim = @data_list[0].size
+      @data_list = scale(@data_list)
+      @real_data_list = Marshal.load(Marshal.dump(@data_list))
+      @mu_list = Array.new(@k_num).map{Array.new(@dim, rand())}
+      @conv_list = Array.new(@k_num).map{ini_conv()}
+      @pi_list = @k_num.times.map{rand()}
+      @gamma = Array.new(@data_list.size).map{Array.new(@k_num, 0)}
     end
- 
-    #
-    # === make datas for test
-    #
-    def make_data
 
+    def ini_conv
+      conv = []
+      @dim.times {|i|
+        conv.push(make_array(i))
+      }
+      return conv
     end
 
-    def run
-      # loop until likelihood is converged
-        # calc burden_ratio
-        new_r_list = []
-        nn = @r_list.inject(:x)
-        @r_list.each_with_index do |r, i|
-          new_r_list[i] = calc_burden_ratio(r)
-          @mu_list[i] = update_mu(r)
-          @var_list[i] = update_var(r)
-          @pi_list = update_pi(r, nn)
+    def make_array(i)
+      array = []
+      @dim.times {|x|
+        if i == x 
+          array.push(1.0)
+        else
+          array.push(0.0)
         end
-        @r_list = new_r_list
-        # calc log_likelihood
+      }
+      return array
+    end
 
+    def create_log(cycle)
+      log = {:cycle => cycle, 
+             :mu => @mu_list.clone, 
+             :conv => @conv_list.clone, 
+             :pi_list => @pi_list.clone}
+      return log
     end
 
-    
- 
-    def calc_burden_ratio(k)
-      denominator = 0.0
-      @pi_list.each_with_index do |pi, i|
-        denominator += normal_rand(@mu_list[i], @var_list[i]) 
+
+    def fit(data_list, k_num)
+      init(data_list, k_num)
+      result = []
+      cycle = 0
+      last_likelihood = calc_log_likelihood()
+      loop do 
+        e_step()
+        m_step()
+        likelihood = calc_log_likelihood()
+        diff = (likelihood - last_likelihood).abs
+        last_likelihood = likelihood
+        puts "likelihood: #{likelihood}"
+        result.push(create_log(cycle))
+        cycle += 1
+        break if diff < 0.000001
       end
+      puts "===================================="
+      puts "pi : #{ @pi_list }"
+      puts "mu : #{ @mu_list}"
+      puts "conv : #{ @conv_list}"
+      return result
+    end
 
-      molecule = @pi_list[k] * gauusian(x, @mu_list[k], @var_list[k])
-      
-      return molecule/denominator
-    end    
 
-    def log_likelihood
-       
-    end 
+    def e_step()
+      @data_list.each_with_index{|data, n|
+        denominator = 0.0
+        @k_num.times{|k|
+          denominator += @pi_list[k] * gauusian(data, @mu_list[k], @conv_list[k])
+        }
+        @k_num.times { |k|
+          @gamma[n][k] = @pi_list[k] * gauusian(data, @mu_list[k], @conv_list[k]) / denominator
+        }
+      }
+    end
 
-    def update_mu(k, burden_ratio)
-      sum = 0.0
-      @datas.each_with_index do |data, i|
-        sum += burden_ratio[i] * data
-      end
-      return sum/@n_list[k]
+    def m_step()
+      @k_num.times {|k|
+        nk = 0.0
+        @data_list.each_with_index{|data,  n| 
+          nk += @gamma[n][k] 
+        }
+
+        @mu_list[k] = calc_ave(k, nk) 
+        @conv_list[k] = calc_conv(k, nk)
+        @pi_list[k] = nk/@data_list.size
+      }
     end
 
-    def update_var(k, burden_ratio)
-      sum = 0.0
-      @datas.each_with_index do |data, i|
-        sum += burden_ratio[i] * ( data - @mu_list[k]) * ( data - @mu_list[k])
-      end
-      return sum/@n_list[k]
+    def calc_ave(k, nk) 
+      mu = Array.new(@dim)
+      @dim.times{|i|
+        mu[i] = @data_list.each_with_index.inject(0.0){|sum,(data, n)| 
+          sum += @gamma[n][k] * data[i] 
+        } / nk
 
+      }
+      return mu
     end
 
-    def update_pi(k)
-      return @n_list[k]/@k_num
-    end
     
+    def calc_conv(k, nk)
+      conv = Array.new(@dim).map{Array.new(@dim, 0)}
+      @dim.times{|i|
+        @dim.times{|j|
+          @data_list.each_with_index{|data, n|
+            conv[i][j] += @gamma[n][k] * (data[i]-@mu_list[k][i]) * (data[j]-@mu_list[k][j])
+          } 
+        }
+      }
+      conv = conv.map{|arr| arr.map{|v| v/nk}}
+      return conv
+    end
+
+    def calc_log_likelihood
+      log_likelihood = 0.0
+      @data_list.each_with_index{|data, i|
+        sum = 0.0
+        @k_num.times{|k|
+          sum += @pi_list[k] * gauusian(data, @mu_list[k], @conv_list[k]) 
+        }
+        log_likelihood += Math.log(sum)
+      }
+      return log_likelihood
+    end 
+
+
     #
     # === 混合正規分布を生成する
     #
     def gaussian_mix()
-      r1 = normal_rand()
-      r2 = normal_rand(5.0, 2.0)
+      r1_x = normal_rand()
+      r1_y = normal_rand()
+      r2_x = normal_rand(10.0, 50.0)
+      r2_y = normal_rand(10.0, 50.0)
 
-      if (rand() < 0.5)
-        return r1
+      if (rand() < 0.15)
+        return [r1_x, r1_y]
       else
-        return r2
+        return [r2_x, r2_y]
       end
     end
 
-
     #
     # === gauusian distribution 
     #
     def gauusian(x, mu, sigma)
-      f1 = (1.0/Math.sqrt(2.0*Math::PI))
-      f2 = Math.exp(-((x-mu)^2 / 2*sigma^2))
-      return f1 * f2
+      if @dim <= 1
+        x = x[0]
+        mu = mu[0]
+        sigma = sigma[0][0]
+        f1 = 1.0/(Math.sqrt(2.0*Math::PI)*Math.sqrt(sigma))
+        f2 = Math.exp(-(((x-mu)**2)/((2.0*sigma))))
+        return f1 * f2
+      else
+        return gauusian2dim(x, mu, sigma)
+      end
     end
 
+
     #
-    # ===ボックス―ミューラー法をよる正規分布乱数発生
-    # @param mu flout 平均
-    # @param sigma flout 標準偏差
-    # @return ボックスミューラー法に従う正規分布に従う乱数を生成
+    # === gauusian distribution .2 dim version
     #
-    def normal_rand(mu = 0,sigma = 1.0)
-      a, b = rand(), rand() ;
-      return (Math.sqrt(-2*Math.log(rand()))*Math.sin(2*Math::PI*rand()) * sigma) + mu
+    def gauusian2dim(x, mu, conv)
+      x = Matrix[x]
+      mu = Matrix[mu]
+      conv = Matrix[*conv]
+      begin
+        f1 = 1.0/(2.0 * Math::PI * ( conv.det**(0.5) ))
+        f2 = Math.exp((-1.0/2.0)*((x-mu) * conv.inverse * (x-mu).transpose)[0, 0])
+      rescue
+        binding.pry ;
+      end
+
+      return (f1 * f2)
     end
 
 
+    def scale(x)
+      if x[0].instance_of?(Array)  # check whether x's factor is 1dim or over 2dim
+        sum_each_vec = []
+        ave_list = []
+        std_list = []
+        x.each{|vec| 
+          vec.each_with_index{|data, i|
+            sum_each_vec[i] = (sum_each_vec[i] == nil) ? data : sum_each_vec[i]+data
+          }
+        }
+        x[0].size.times{|i|
+          ave_list.push(sum_each_vec[i]/x.size)
+        }
+
+        sum_each_vec = []
+        x.each{|vec| 
+          vec.each_with_index{|data, i|
+            sum_each_vec[i] = (sum_each_vec[i] == nil) ? (ave_list[i]-data)**2 : (sum_each_vec[i]+(ave_list[i]-data)**2)
+          }
+        }
+        x[0].size.times{|i|
+          std_list.push(Math.sqrt(sum_each_vec[i]/x.size))
+        }
+
+        scaled_x = []
+        x.each_with_index{|vec, i| 
+          scaled_x[i] ||= []
+          vec.each_with_index{|data, j|
+            scaled_x[i][j] ||= (data-ave_list[j])/std_list[j]
+          }
+        }
+        return scaled_x
+      else  # if 1dim
+        mu = x.each.inject(0.0){|sum, data| sum+=data}/x.size
+        var = (x.each.inject(0.0){|sum, data| sum+=(data-mu)**2}/x.size) 
+        std = Math.sqrt(var)
+        scaled_x = []
+        x.each{|sum,data| 
+          scaled_x.push((mu-data)/std)
+        } 
+        return scaled_x
+      end
+    end
   end
 end

data/lib/t_learn/text.json

@@ -0,0 +1,272 @@
+[[3.600000,  79.000000], 
+[1.800000,  54.000000], 
+[3.333000,  74.000000], 
+[2.283000,  62.000000], 
+[4.533000,  85.000000], 
+[2.883000,  55.000000], 
+[4.700000,  88.000000], 
+[3.600000,  85.000000], 
+[1.950000,  51.000000], 
+[4.350000,  85.000000], 
+[1.833000,  54.000000], 
+[3.917000,  84.000000], 
+[4.200000,  78.000000], 
+[1.750000,  47.000000], 
+[4.700000,  83.000000], 
+[2.167000,  52.000000], 
+[1.750000,  62.000000], 
+[4.800000,  84.000000], 
+[1.600000,  52.000000], 
+[4.250000,  79.000000], 
+[1.800000,  51.000000], 
+[1.750000,  47.000000], 
+[3.450000,  78.000000], 
+[3.067000,  69.000000], 
+[4.533000,  74.000000], 
+[3.600000,  83.000000], 
+[1.967000,  55.000000], 
+[4.083000,  76.000000], 
+[3.850000,  78.000000], 
+[4.433000,  79.000000], 
+[4.300000,  73.000000], 
+[4.467000,  77.000000], 
+[3.367000,  66.000000], 
+[4.033000,  80.000000], 
+[3.833000,  74.000000], 
+[2.017000,  52.000000], 
+[1.867000,  48.000000], 
+[4.833000,  80.000000], 
+[1.833000,  59.000000], 
+[4.783000,  90.000000], 
+[4.350000,  80.000000], 
+[1.883000,  58.000000], 
+[4.567000,  84.000000], 
+[1.750000,  58.000000], 
+[4.533000,  73.000000], 
+[3.317000,  83.000000], 
+[3.833000,  64.000000], 
+[2.100000,  53.000000], 
+[4.633000,  82.000000], 
+[2.000000,  59.000000], 
+[4.800000,  75.000000], 
+[4.716000,  90.000000], 
+[1.833000,  54.000000], 
+[4.833000,  80.000000], 
+[1.733000,  54.000000], 
+[4.883000,  83.000000], 
+[3.717000,  71.000000], 
+[1.667000,  64.000000], 
+[4.567000,  77.000000], 
+[4.317000,  81.000000], 
+[2.233000,  59.000000], 
+[4.500000,  84.000000], 
+[1.750000,  48.000000], 
+[4.800000,  82.000000], 
+[1.817000,  60.000000], 
+[4.400000,  92.000000], 
+[4.167000,  78.000000], 
+[4.700000,  78.000000], 
+[2.067000,  65.000000], 
+[4.700000,  73.000000], 
+[4.033000,  82.000000], 
+[1.967000,  56.000000], 
+[4.500000,  79.000000], 
+[4.000000,  71.000000], 
+[1.983000,  62.000000], 
+[5.067000,  76.000000], 
+[2.017000,  60.000000], 
+[4.567000,  78.000000], 
+[3.883000,  76.000000], 
+[3.600000,  83.000000], 
+[4.133000,  75.000000], 
+[4.333000,  82.000000], 
+[4.100000,  70.000000], 
+[2.633000,  65.000000], 
+[4.067000,  73.000000], 
+[4.933000,  88.000000], 
+[3.950000,  76.000000], 
+[4.517000,  80.000000], 
+[2.167000,  48.000000], 
+[4.000000,  86.000000], 
+[2.200000,  60.000000], 
+[4.333000,  90.000000], 
+[1.867000,  50.000000], 
+[4.817000,  78.000000], 
+[1.833000,  63.000000], 
+[4.300000,  72.000000], 
+[4.667000,  84.000000], 
+[3.750000,  75.000000], 
+[1.867000,  51.000000], 
+[4.900000,  82.000000], 
+[2.483000,  62.000000], 
+[4.367000,  88.000000], 
+[2.100000,  49.000000], 
+[4.500000,  83.000000], 
+[4.050000,  81.000000], 
+[1.867000,  47.000000], 
+[4.700000,  84.000000], 
+[1.783000,  52.000000], 
+[4.850000,  86.000000], 
+[3.683000,  81.000000], 
+[4.733000,  75.000000], 
+[2.300000,  59.000000], 
+[4.900000,  89.000000], 
+[4.417000,  79.000000], 
+[1.700000,  59.000000], 
+[4.633000,  81.000000], 
+[2.317000,  50.000000], 
+[4.600000,  85.000000], 
+[1.817000,  59.000000], 
+[4.417000,  87.000000], 
+[2.617000,  53.000000], 
+[4.067000,  69.000000], 
+[4.250000,  77.000000], 
+[1.967000,  56.000000], 
+[4.600000,  88.000000], 
+[3.767000,  81.000000], 
+[1.917000,  45.000000], 
+[4.500000,  82.000000], 
+[2.267000,  55.000000], 
+[4.650000,  90.000000], 
+[1.867000,  45.000000], 
+[4.167000,  83.000000], 
+[2.800000,  56.000000], 
+[4.333000,  89.000000], 
+[1.833000,  46.000000], 
+[4.383000,  82.000000], 
+[1.883000,  51.000000], 
+[4.933000,  86.000000], 
+[2.033000,  53.000000], 
+[3.733000,  79.000000], 
+[4.233000,  81.000000], 
+[2.233000,  60.000000], 
+[4.533000,  82.000000], 
+[4.817000,  77.000000], 
+[4.333000,  76.000000], 
+[1.983000,  59.000000], 
+[4.633000,  80.000000], 
+[2.017000,  49.000000], 
+[5.100000,  96.000000], 
+[1.800000,  53.000000], 
+[5.033000,  77.000000], 
+[4.000000,  77.000000], 
+[2.400000,  65.000000], 
+[4.600000,  81.000000], 
+[3.567000,  71.000000], 
+[4.000000,  70.000000], 
+[4.500000,  81.000000], 
+[4.083000,  93.000000], 
+[1.800000,  53.000000], 
+[3.967000,  89.000000], 
+[2.200000,  45.000000], 
+[4.150000,  86.000000], 
+[2.000000,  58.000000], 
+[3.833000,  78.000000], 
+[3.500000,  66.000000], 
+[4.583000,  76.000000], 
+[2.367000,  63.000000], 
+[5.000000,  88.000000], 
+[1.933000,  52.000000], 
+[4.617000,  93.000000], 
+[1.917000,  49.000000], 
+[2.083000,  57.000000], 
+[4.583000,  77.000000], 
+[3.333000,  68.000000], 
+[4.167000,  81.000000], 
+[4.333000,  81.000000], 
+[4.500000,  73.000000], 
+[2.417000,  50.000000], 
+[4.000000,  85.000000], 
+[4.167000,  74.000000], 
+[1.883000,  55.000000], 
+[4.583000,  77.000000], 
+[4.250000,  83.000000], 
+[3.767000,  83.000000], 
+[2.033000,  51.000000], 
+[4.433000,  78.000000], 
+[4.083000,  84.000000], 
+[1.833000,  46.000000], 
+[4.417000,  83.000000], 
+[2.183000,  55.000000], 
+[4.800000,  81.000000], 
+[1.833000,  57.000000], 
+[4.800000,  76.000000], 
+[4.100000,  84.000000], 
+[3.966000,  77.000000], 
+[4.233000,  81.000000], 
+[3.500000,  87.000000], 
+[4.366000,  77.000000], 
+[2.250000,  51.000000], 
+[4.667000,  78.000000], 
+[2.100000,  60.000000], 
+[4.350000,  82.000000], 
+[4.133000,  91.000000], 
+[1.867000,  53.000000], 
+[4.600000,  78.000000], 
+[1.783000,  46.000000], 
+[4.367000,  77.000000], 
+[3.850000,  84.000000], 
+[1.933000,  49.000000], 
+[4.500000,  83.000000], 
+[2.383000,  71.000000], 
+[4.700000,  80.000000], 
+[1.867000,  49.000000], 
+[3.833000,  75.000000], 
+[3.417000,  64.000000], 
+[4.233000,  76.000000], 
+[2.400000,  53.000000], 
+[4.800000,  94.000000], 
+[2.000000,  55.000000], 
+[4.150000,  76.000000], 
+[1.867000,  50.000000], 
+[4.267000,  82.000000], 
+[1.750000,  54.000000], 
+[4.483000,  75.000000], 
+[4.000000,  78.000000], 
+[4.117000,  79.000000], 
+[4.083000,  78.000000], 
+[4.267000,  78.000000], 
+[3.917000,  70.000000], 
+[4.550000,  79.000000], 
+[4.083000,  70.000000], 
+[2.417000,  54.000000], 
+[4.183000,  86.000000], 
+[2.217000,  50.000000], 
+[4.450000,  90.000000], 
+[1.883000,  54.000000], 
+[1.850000,  54.000000], 
+[4.283000,  77.000000], 
+[3.950000,  79.000000], 
+[2.333000,  64.000000], 
+[4.150000,  75.000000], 
+[2.350000,  47.000000], 
+[4.933000,  86.000000], 
+[2.900000,  63.000000], 
+[4.583000,  85.000000], 
+[3.833000,  82.000000], 
+[2.083000,  57.000000], 
+[4.367000,  82.000000], 
+[2.133000,  67.000000], 
+[4.350000,  74.000000], 
+[2.200000,  54.000000], 
+[4.450000,  83.000000], 
+[3.567000,  73.000000], 
+[4.500000,  73.000000], 
+[4.150000,  88.000000], 
+[3.817000,  80.000000], 
+[3.917000,  71.000000], 
+[4.450000,  83.000000], 
+[2.000000,  56.000000], 
+[4.283000,  79.000000], 
+[4.767000,  78.000000], 
+[4.533000,  84.000000], 
+[1.850000,  58.000000], 
+[4.250000,  83.000000], 
+[1.983000,  43.000000], 
+[2.250000,  60.000000], 
+[4.750000,  75.000000], 
+[4.117000,  81.000000], 
+[2.150000,  46.000000], 
+[4.417000,  90.000000], 
+[1.817000,  46.000000], 
+[4.467000,  74.000000]]

data/lib/t_learn/version.rb

@@ -1,3 +1,3 @@
 module TLearn
-  VERSION = "0.1.1.3"
+  VERSION = "0.1.1.4"
 end

data/lib/t_learn.rb

@@ -1,6 +1,7 @@
 require "t_learn/version"
 require "t_learn/hop_field_net"
 require "t_learn/feedforward_neural_network"
+require "t_learn/em"
 
 module TLearn
   # Your code goes here...

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: t_learn
 version: !ruby/object:Gem::Version
-  version: 0.1.1.3
+  version: 0.1.1.4
 platform: ruby
 authors:
 - Tcom242242
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2017-05-03 00:00:00.000000000 Z
+date: 2017-05-31 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -60,6 +60,7 @@ extensions: []
 extra_rdoc_files: []
 files:
 - ".gitignore"
+- ".rake_tasks~"
 - ".rspec"
 - ".travis.yml"
 - CODE_OF_CONDUCT.md
@@ -70,10 +71,12 @@ files:
 - bin/console
 - bin/setup
 - lib/t_learn.rb
+- lib/t_learn/.em.rb.swp
 - lib/t_learn/em.rb
 - lib/t_learn/feedforward_neural_network.rb
 - lib/t_learn/hop_field_net.rb
 - lib/t_learn/k_means.rb
+- lib/t_learn/text.json
 - lib/t_learn/version.rb
 - t_learn.gemspec
 homepage: https://github.com/Tcom242242/t_learn