aprendizaje_maquina 0.1.2 → 0.1.3a

data/lib/aprendizaje_maquina.rb

@@ -2,14 +2,12 @@ require "aprendizaje_maquina/version"
 require "aprendizaje_maquina/cargar"
 require "aprendizaje_maquina/regresion_lineal"
 require "aprendizaje_maquina/matrixx"
+require "aprendizaje_maquina/clasificacion_logistica"
+require "aprendizaje_maquina/clustering"
 
 module AprendizajeMaquina
-	#class RegresionLogistica
-		# Coming soon...
-	#end
-
 	#class RedNeuronal
-			# Coming soon...
+		# Coming soon...
 	#end
 
 	#class ArbolDecision

data/lib/aprendizaje_maquina/cargar.rb

@@ -1,46 +1,48 @@
 require 'csv'   
 require 'matrix'
+
 module AprendizajeMaquina
+  class Cargar
+    def initialize(path_file)
+	    @path_file = path_file
+	    @csv_data = CSV.read(@path_file)
+	    @largo_colum = @csv_data[0].length
+    end
 
-	class Cargar
+    def to_matrix(columnas = nil)
+  	  if columnas == nil
+  	    array = @csv_data.map{ |e| e.map{ |o| o.include?(".") ? o.to_f : o.to_i } }
+  		  matrix = Matrix.rows(array,copy=true)
+  		  matrix
+  	  elsif columnas.is_a?(Range)
+  	    if columnas.last >= @largo_colum
+  		    raise ArgumentError, "Number of columns don't exist"
+  		  else
+  		    array = @csv_data.map{ |e| e[columnas].map{ |i| i.include?(".") ? i.to_f : i.to_i} }
+  		    matrix = Matrix.rows(array,copy=true)
+  		    matrix
+  		  end
+  	  elsif columnas.is_a?(Integer)
+  		  if columnas >= @largo_colum
+  		    raise ArgumentError, "Number of columns don't exist"
+  		  else
+  		    array = @csv_data.map { |e| e[columnas].include?(".") ? e[columnas].to_f : e[columnas].to_i }
+  		    matrix = Matrix[array].transpose
+  		    matrix
+  		  end
+  	  else
+  		  raise ArgumentError, "Must be nil, range or integer"
+  	  end
+	  end
 		
-		def initialize(path_file)
-			@path_file = path_file
-			@csv_data = CSV.read(@path_file)
-			#@csv_data = CSV.read(File.join(Dir.pwd,@path_file))
-			@largo_colum = @csv_data[0].length
-		end
-			
-		def to_matrix(columnas = nil)
-			if columnas == nil
-				array = @csv_data.map{|e| e.map{|o| o.to_i } }
-				matrix = Matrix.rows(array,copy=true)
-				matrix
-			elsif columnas.is_a?(Range)
-				if columnas.last >= @largo_colum
-					raise ArgumentError, "Number of columns don't exist"
-				else
-					array = @csv_data.map{|e| e[columnas].map{|i| i.to_i} }
-					matrix = Matrix.rows(array,copy=true)
-					matrix
-				end
-			elsif columnas.is_a?(Integer)
-				array = @csv_data.map { |e| e[columnas].to_i }
-				matrix = Matrix[array].transpose
-				matrix
-			else
-				raise ArgumentError, "Must be nil, range or integer"
-			end
-		end
-
-		def to_vector(columna)
-			if columna >= @largo_colum
-				raise ArgumentError, "Column don't exist"
-			else
-				array = @csv_data.map { |e| e[columna].to_i }
-				vector = Vector.elements(array,copy=true)
-				vector
-			end
-		end
-	end
+    def to_vector(columna)
+	    if columna >= @largo_colum
+		    raise ArgumentError, "Column don't exist"
+	    else
+		    array = @csv_data.map { |e| e[columna].include?(".") ? e[columna].to_f : e[columna].to_i }
+		    vector = Vector.elements(array,copy = true)
+		    vector
+	    end
+	  end
+  end
 end

data/lib/aprendizaje_maquina/clasificacion_logistica.rb

@@ -0,0 +1,104 @@
+module AprendizajeMaquina
+  class ClasificacionLogistica
+    def initialize(x,y,theta)
+      @x = x
+      @y = y
+      @theta = theta
+      if y.is_a? Matrix
+        @m = y.row_count
+      elsif y.is_a? Vector
+        @m = y.size
+      else
+        @m = y.length
+      end
+    end
+    
+    def train(iterations,alpha = nil,type_of_train)
+      case type_of_train
+      when 'Grad' then
+        # gradiente de descenso
+        @cost_history = []
+        for i in 0..iterations
+          x = @x * @theta
+          hx = x.map { |e| sigmoid(e) }
+          @theta = @theta - alpha / @m * @x.transpose * (hx - @y)
+          costo = 0
+          cost.to_a.map{ |e| costo = e }
+          @cost_history << ["iteracion: #{i}",costo]
+        end
+        @cost_history
+        "theta values => #{@theta} | cost => #{costo}"
+      when 'Newm' then
+        # metodo de newton
+        @cost_history = []
+        for i in 0..iterations
+          x = @x * @theta
+          hx = x.map { |e| sigmoid(e) }
+          uno_menos_hx = hx.map{ |e| (1-e) }
+          escalar = []
+          for u in 0...hx.size
+            escalar << hx[u] * uno_menos_hx[u]
+          end
+          gradiente = (1.0/@m) * @x.transpose * (hx - @y)
+          hessian = (1.0/@m) * @x.transpose * sumatoria(escalar) * @x
+          inversa = (1.0/hessian.det) * (hessian.adjugate)
+          @theta = @theta - inversa * gradiente
+          costo = 0
+          cost.to_a.map{ |e| costo = e }
+          @cost_history << ["iteracion: #{i}",costo]
+        end
+        @cost_history
+        "theta values => #{@theta} | cost => #{costo}"
+      when 'SGD' then
+        # Stochastic Gradient Descent
+        @cost_history = []
+        for i in 0..iterations
+          for i in 0..i
+            x = matrix(@x.to_a.map{|e| e.shuffle })*@theta
+            hx = x.map {|e| sigmoid(e) }
+            @theta = @theta - alpha / @m * @x.transpose * (hx - @y)
+            costo = 0
+            cost.to_a.map{|e| costo = e }
+            @cost_history << ["iteracion: #{i}",costo]
+          end
+        end
+        @cost_history
+        "theta values => #{@theta} | cost => #{costo}"
+      end     
+    end
+    
+    def predict(x)
+      hipo = x * @theta 
+      var = 0
+      hipo.map { |x| var = x.is_a?(Integer) ? x.to_i : x.to_f }
+      if sigmoid(var) >= 0.5 
+        1
+      else
+        0
+      end
+    end
+    
+    private
+        
+    def sumatoria(array)
+      array.inject(0) { |elem1, elem2| elem1 + elem2 }
+    end
+    
+    def sigmoid(x)
+      1.0 / (1.0 + Math.exp(-x))
+    end
+        
+    def cost
+      x = @x*@theta
+      hx = x.map { |e| sigmoid(e) }
+      log_hx = hx.map{ |e| Math.log(e) }
+      log_uno_menos_hx = hx.map{ |e| Math.log(1 - e) }
+      costo = -1.0/@m * (Matrix[@y.to_a] * log_hx + (Matrix[@y.to_a].map{ |i| 1 - i }) * log_uno_menos_hx)
+      costo
+    end
+                    
+    def matrix(columns)
+      Matrix.rows(columns, false)
+    end
+  end
+end

data/lib/aprendizaje_maquina/clustering.rb

@@ -0,0 +1,96 @@
+module AprendizajeMaquina
+	class KmeansClustering
+		def initialize(num_of_cluster_centroids,dataset_matrix)
+			@num_of_cluster_centroids = num_of_cluster_centroids
+			@dataset_matrix = dataset_matrix
+			@num_columns = @dataset_matrix.column_count
+			@num_rows = @dataset_matrix.row_count
+			@cluster_centroids = init_cluster_centroids
+		end
+
+		def fit(iterations)
+			clustering(iterations)
+		end
+
+		def cluster(num)
+			get("@cluster_#{num}")
+		end
+
+		def predict(vector)	
+			array = []			
+			@cluster_centroids.each do |cluster|
+			  array << (vector-cluster).r
+			end
+			cluster = array.index(array.min)
+			cluster
+		end
+
+		private
+
+		def array_to_vector(array)
+			vector = Vector.elements(array, copy = true)
+			vector
+		end
+
+		def media(array)
+			if array.empty?
+		    	array#raise ArgumentError.new("array is empty")
+		    else
+		    	1.0/array.length * array.inject { |mem, var| mem + var }
+		  	end
+		end
+
+		def init_cluster_centroids
+			cluster_centroids = Array.new(@num_of_cluster_centroids) { 
+				min_max_rand = []
+				for i in 0...@num_columns
+					min_max_rand << rand(@dataset_matrix.column(i).min..@dataset_matrix.column(i).max)
+				end
+				array_to_vector(min_max_rand) 
+			}
+			cluster_centroids
+		end
+
+		def set(instance_variable_name,instance_variable_value)
+			instance_variable_set(instance_variable_name,instance_variable_value)
+		end
+
+		def get(instance_variable_name)
+			instance_variable_get(instance_variable_name)
+		end
+		
+		def clustering(iterations)
+			iterations.times do
+				array2 = []
+				for i in 0...@num_rows
+					array = []
+					@cluster_centroids.each do |cluster|
+						array << (@dataset_matrix.row(i)-cluster).r
+					end
+					array2 << array
+				end
+
+				hash = {}
+				for i in 0...@num_rows
+					hash[@dataset_matrix.row(i)] = array2[i].index(array2[i].min)
+				end
+
+				@cluster_centroids.each_index do |index|
+					set("@cluster_#{index}", Array.new)
+				end
+
+				@cluster_centroids.each_index do |index|
+					hash.each do |key,value|
+						if value == index
+							get("@cluster_#{index}") << key
+						end
+					end
+				end
+
+				@cluster_centroids.each_index do |index|
+					@cluster_centroids[index] = media(get("@cluster_#{index}"))
+				end
+			end
+		end
+	end
+end

data/lib/aprendizaje_maquina/matrixx.rb

@@ -1,7 +1,7 @@
 class Matrix
 	def add_ones
-		matrix = self.to_a.map{|i| i.insert(0,1)}
-		matrix = Matrix.rows(matrix,copy=true)
+		matrix = self.to_a.map{ |i| i.insert(0,1) }
+		matrix = Matrix.rows(matrix,copy = true)
 		matrix
 	end
 
@@ -10,7 +10,7 @@ class Matrix
 		self.column_count.times do |i|
 			array << self.column(i).normalize
 		end
-		matrix_normal = Matrix.rows(array,copy=true).transpose
+		matrix_normal = Matrix.rows(array,copy = true).transpose
 		matrix_normal
 	end	
 end

data/lib/aprendizaje_maquina/regresion_lineal.rb

@@ -1,8 +1,6 @@
 module AprendizajeMaquina
-
-	class RegresionLineal
-		
-		attr_reader :m,:b, :ecuacion, :theta
+  class RegresionLineal
+	  attr_reader :m,:b, :ecuacion, :theta
 
 		def initialize(x,y)
 			@x = x
@@ -44,7 +42,7 @@ module AprendizajeMaquina
 					raise ArgumentError, "Must be a number or matrix 1xN"
 				end
 			else
-				return "There is not a equation to make predictions (first, run encontrar_ecuacion method)"
+				return "There is not a equation to make predictions (first, run find_ecuation method)"
 			end
 		end
 
@@ -67,7 +65,7 @@ module AprendizajeMaquina
 			iter = @n - 1
 			xy = []
 			for i in 0..iter
-				xy << array_1[i]*array_2[i]
+				xy << array_1[i] * array_2[i]
 			end
 			xy
 		end

data/lib/aprendizaje_maquina/version.rb

@@ -1,3 +1,3 @@
 module AprendizajeMaquina
-  VERSION = "0.1.2"
+  VERSION = "0.1.3a"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: aprendizaje_maquina
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.1.3a
 platform: ruby
 authors:
 - Erickson Morales
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2017-11-12 00:00:00.000000000 Z
+date: 2018-05-11 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -72,12 +72,20 @@ files:
 - aprendizaje_maquina.gemspec
 - bin/console
 - bin/setup
+- examples/clasificacion_logistica_example.rb
+- examples/clustering_data.csv
+- examples/clustering_example.rb
+- examples/data_lg.csv
+- examples/regresion_lineal_example.rb
+- examples/train.csv
 - lib/aprendizaje_maquina.rb
 - lib/aprendizaje_maquina/cargar.rb
+- lib/aprendizaje_maquina/clasificacion_logistica.rb
+- lib/aprendizaje_maquina/clustering.rb
 - lib/aprendizaje_maquina/matrixx.rb
 - lib/aprendizaje_maquina/regresion_lineal.rb
 - lib/aprendizaje_maquina/version.rb
-homepage: ''
+homepage: 
 licenses:
 - MIT
 metadata: {}
@@ -92,12 +100,12 @@ required_ruby_version: !ruby/object:Gem::Requirement
       version: '0'
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
-  - - ">="
+  - - ">"
     - !ruby/object:Gem::Version
-      version: '0'
+      version: 1.3.1
 requirements: []
 rubyforge_project: 
-rubygems_version: 2.5.2
+rubygems_version: 2.6.14
 signing_key: 
 specification_version: 4
 summary: Machine learning gem / Una gema para el aprendizaje de maquinas.