k_means 0.0.5 → 0.0.7

data/README.rdoc

@@ -4,18 +4,32 @@ Attempting to build a fast, memory efficient K-Means program.
 
 == Install
 
-  gem sources -a http://gems.github.com
-  sudo gem install reddavis-k_means
+  gem sources -a http://rubygems.org
+  sudo gem install k_means
 
 == How To Use
   require 'rubygems'
   require 'k_means'
-  
+
   data = [[1,1], [1,2], [1,1], [1000, 1000], [500, 500]]
-  kmeans = KMeans.new(@data, :centroids => 2)
+  kmeans = KMeans.new(data, :centroids => 2)
   kmeans.inspect  # Use kmeans.view to get hold of the un-inspected array
   => [[3, 4], [0, 1, 2]]
-  
+
+== Custom Centroids
+  require 'rubygems'
+  require 'k_means'
+
+  # Your custom centroid needs to have #position and #reposition methods
+  class CustomCentroid
+    attr_accessor :position
+    def initialize(position); @position = position; end
+    def reposition(nodes, centroid_positions); end
+  end
+
+  data = [[1,1], [1,2], [1,1], [1000, 1000], [500, 500]]
+  kmeans = KMeans.new(data, :custom_centroids => @specified_centroids)
+
 == Distance Measurements
 
 KMeans uses the Distance Measures Gem (http://github.com/reddavis/Distance-Measures) so we get quite a range of distance measurements.
@@ -35,19 +49,19 @@ The measurements currently available are:
   binary_jaccard_distance
 
   tanimoto_coefficient
-  
+
 To specify a particular one to use in the KMeans algorithm, just provide it as an option:
 
   KMeans.new(@data, :distance_measure => :jaccard_index)
   KMeans.new(@data, :distance_measure => :cosine_similarity)
   KMeans.new(@data, :distance_measure => :tanimoto_coefficient)
-  
+
 You get the idea...
-  
+
 == Benchmarks
 
   # 1000 records with 50 dimensions
-  data = Array.new(1000) {Array.new(50) {rand(10)}} 
+  data = Array.new(1000) {Array.new(50) {rand(10)}}
   ai4r_data = Ai4r::Data::DataSet.new(:data_items=> data)
 
   # Clustering can happen in magical ways

data/VERSION

@@ -1 +1 @@
-0.0.5
+0.0.7

data/benchmark/benchmark_ai4r.rb

@@ -3,7 +3,7 @@ require 'rubygems'
 require 'ai4r'
 require File.dirname(__FILE__) + '/../lib/k_means'
 
-data = Array.new(1000) {Array.new(50) {rand(10)}} 
+data = Array.new(100) {Array.new(50) {rand(10)}}
 
 ai4r_data = Ai4r::Data::DataSet.new(:data_items=> data)
 
@@ -16,7 +16,7 @@ Benchmark.bm do |x|
   x.report('Mine') do
     n.times { KMeans.new(data) }
   end
-  
+
   x.report("Ai4R") do
     n.times do
       b = Ai4r::Clusterers::KMeans.new

data/k_means.gemspec

@@ -5,11 +5,11 @@
 
 Gem::Specification.new do |s|
   s.name = %q{k_means}
-  s.version = "0.0.5"
+  s.version = "0.0.7"
 
   s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
   s.authors = ["reddavis"]
-  s.date = %q{2010-01-25}
+  s.date = %q{2010-07-11}
   s.description = %q{Attempting to create a fast, memory efficient KMeans}
   s.email = %q{reddavis@gmail.com}
   s.extra_rdoc_files = [
@@ -40,7 +40,7 @@ Gem::Specification.new do |s|
   s.homepage = %q{http://github.com/reddavis/k_means}
   s.rdoc_options = ["--charset=UTF-8"]
   s.require_paths = ["lib"]
-  s.rubygems_version = %q{1.3.5}
+  s.rubygems_version = %q{1.3.6}
   s.summary = %q{K Means algorithm}
   s.test_files = [
     "test/ext/test_object.rb",

data/lib/k_means/centroid.rb

@@ -35,7 +35,7 @@ class Centroid
   
   # Finds the average distance of all the nodes assigned to
   # the centroid and then moves the centroid to that position
-  def reposition(nodes)
+  def reposition(nodes, centroids)
     return if nodes.empty?
     averages = [0.0] * nodes[0].position.size
     nodes.each do |node|
@@ -46,4 +46,4 @@ class Centroid
     @position = averages.map {|x| x / nodes.size}
   end
   
-end
+end

data/lib/k_means/k_means.rb

@@ -1,42 +1,41 @@
 require 'ext/object'
 
 class KMeans
-  
+
   attr_reader :centroids, :nodes
-  
+
   def initialize(data, options={})
-    k = options[:centroids] || 4
-    @verbose = options[:verbose]
-    
     distance_measure = options[:distance_measure] || :euclidean_distance
     @nodes = Node.create_nodes(data, distance_measure)
-    @centroids = Centroid.create_centroids(k, @nodes)
-    
+    @centroids = options[:custom_centroids] ||
+      Centroid.create_centroids(options[:centroids] || 4, @nodes)
+    @verbose = options[:verbose]
+
     perform_cluster_process
   end
-  
+
   def inspect
     @centroid_pockets.inspect
   end
-  
+
   def view
     @centroid_pockets
   end
-  
+
   private
-  
+
   def perform_cluster_process
     iterations, updates = 0, 1
-    while updates > 0 && iterations < 100 
+    while updates > 0 && iterations < 100
       iterations += 1
       verbose_message("Iteration #{iterations}")
-      updates = 0  
+      updates = 0
       updates += update_nodes
       reposition_centroids
     end
     place_nodes_into_pockets
   end
-  
+
   # This creates an array of arrays
   # Each internal array represents a centroid
   # and each in the array represents the nodes index
@@ -51,7 +50,7 @@ class KMeans
     end
     @centroid_pockets = centroid_pockets
   end
-    
+
   def update_nodes
     sum = 0
     @nodes.each do |node|
@@ -59,17 +58,18 @@ class KMeans
     end
     sum
   end
-  
+
   def reposition_centroids
+    centroid_positions = @centroids.map(&:position)
     @centroids.each do |centroid|
-      nodes = [] 
+      nodes = []
       @nodes.each {|n| nodes << n if n.closest_centroid == centroid}
-      centroid.reposition(nodes)
+      centroid.reposition(nodes, centroid_positions)
     end
   end
-        
+
   def verbose_message(message)
     puts message if @verbose
   end
-      
-end
+
+end

data/lib/k_means/node.rb

@@ -1,5 +1,5 @@
 class Node
-  
+
   class << self
     def create_nodes(data, similarity_measure)
       nodes = []
@@ -9,41 +9,59 @@ class Node
       nodes
     end
   end
-  
+
   attr_accessor :position, :best_distance, :closest_centroid
-  
+
   def initialize(position, similarity_measure)
     @position = position
     @similarity_measure = similarity_measure
   end
-  
+
   def update_closest_centroid(centroids)
+    # If we haven't processed this node we need to give it an initial centroid
+    # so that we have something to compare distances against
     calculate_initial_centroid(centroids.first) unless @closest_centroid
+
     updated = false
     centroids.each do |centroid|
+      # Check if they are in the same position
+      if centroid.position == @position
+        updated = update_attributes(centroid, 0.0)
+        break
+      end
+
       distance = calculate_distance(centroid)
       if distance < best_distance
-        updated = true
-        @closest_centroid = centroid
-        @best_distance = distance
+        updated = update_attributes(centroid, distance)
       end
     end
+
     updated == true ? 1 : 0
   end
-    
+
+  def reset!
+    @closest_centroid = nil
+    @best_distance    = nil
+  end
+
   private
-  
+
+  def update_attributes(closest_centroid, best_distance)
+    @closest_centroid, @best_distance = closest_centroid, best_distance
+    true
+  end
+
   def calculate_initial_centroid(centroid)
     @closest_centroid = centroid
     @best_distance = calculate_distance(centroid)
   end
-  
+
   def calculate_distance(centroid)
     begin
       @position.send(@similarity_measure, centroid.position)
     rescue NoMethodError
-      raise "Hey, that's not a measurement. Read the REAdME for available measurements"
+      raise "Hey, '#{@similarity_measure}' is not a measurement. Read the REAdME for available measurements"
     end
   end
-  
-end
+
+end

data/test/k_means/test_centroid.rb

@@ -31,7 +31,7 @@ class TestCentroid < Test::Unit::TestCase
         end
       end
       average_position.map! {|x| x / 2}
-      @centroid.reposition(create_nodes)
+      @centroid.reposition(create_nodes, [@centroid.position])
       assert_equal average_position, @centroid.position
     end
       

data/test/k_means/test_k_means.rb

@@ -2,23 +2,49 @@ require 'helper'
 
 class TestKMeans < Test::Unit::TestCase
   context "A KMeans Instance" do
-    
+
     setup do
-      @data = Array.new(200) {Array.new(2) {rand}}
+      @data = Array.new(3) {Array.new(2) {rand}}
       @kmeans = KMeans.new(@data, :centroids => 2, :distance_measure => :cosine_similarity)
     end
-  
+
     should "return an array" do
       assert_kind_of String, @kmeans.inspect
     end
-    
+
     should "have 2 centroids" do
       assert_equal 2, @kmeans.centroids.size
     end
-    
+
     should "have 200 nodes" do
-      assert_equal 200, @kmeans.nodes.size
+      assert_equal 3, @kmeans.nodes.size
+    end
+
+  end
+
+  context "A KMeans Instance with specified initial centroids" do
+    setup do
+      @data = Array.new(3) {Array.new(2) {rand}}
+      class CustomCentroid
+        attr_accessor :position
+        def initialize(position); @position = position; end
+        def reposition(nodes, centroid_positions); end
+      end
+
+      @specified_centroids = @data[0..2].map { |d| CustomCentroid.new(d) }
+      @kmeans = KMeans.new(@data, :custom_centroids => @specified_centroids, :distance_measure => :cosine_similarity)
+    end
+
+    should "return an inspected array" do
+      assert_kind_of String, @kmeans.inspect
+    end
+
+    should "have 3 centroids" do
+      assert_equal 3, @kmeans.centroids.size
+   end
+
+    should "have 3 nodes" do
+      assert_equal 3, @kmeans.nodes.size
     end
-        
   end
 end

data/test/k_means/test_node.rb

@@ -2,49 +2,63 @@ require 'helper'
 
 class TestNode < Test::Unit::TestCase
   context "A Data Instance" do
-    
+
     setup do
       @node = Node.new([4, 4], :euclidean_distance)
     end
-  
+
     should "return an array" do
       assert_kind_of Array, @node.position
     end
-    
+
     should "create an array of nodes" do
       data = Array.new(10) {Array.new(2) {rand}}
       nodes = Node.create_nodes(data, :euclidean_distance)
       assert_kind_of Array, nodes
     end
-    
+
     should "create 10 nodes" do
       data = Array.new(10) {Array.new(2) {rand}}
       nodes = Node.create_nodes(data, :euclidean_distance)
       assert_equal 10, nodes.size
     end
-    
+
     should "initialize closest centroid" do
       a = @node.closest_centroid
       centroids = [Centroid.new([4, 4]), Centroid.new([5, 4])]
       @node.update_closest_centroid(centroids)
       assert_not_equal nil, @node.closest_centroid
     end
-    
+
     should "update closest centroid" do
       centroids = [Centroid.new([5, 4])]
       @node.update_closest_centroid(centroids)
       a = @node.closest_centroid
       @node.update_closest_centroid([Centroid.new([4,4])])
-      
+
       assert_not_equal a, @node.closest_centroid
       assert_equal 0, @node.best_distance
     end
-    
+
+    context "Reset the node" do
+      should "update closest centroid" do
+        centroids = [Centroid.new([4, 4])]
+        @node.update_closest_centroid(centroids)
+        a = @node.closest_centroid
+        @node.reset!
+
+        @node.update_closest_centroid([Centroid.new([5,4])])
+
+        assert_not_equal a, @node.closest_centroid
+        assert_equal 1, @node.best_distance
+      end
+    end
+
     should "raise error if a false measure is specified" do
       assert_raise NoMethodError do
         Node.new([9,9], :fakey).calculate_distance([1,1])
       end
     end
-        
+
   end
-end
+end

metadata

@@ -1,7 +1,12 @@
 --- !ruby/object:Gem::Specification 
 name: k_means
 version: !ruby/object:Gem::Version 
-  version: 0.0.5
+  prerelease: false
+  segments: 
+  - 0
+  - 0
+  - 7
+  version: 0.0.7
 platform: ruby
 authors: 
 - reddavis
@@ -9,19 +14,23 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2010-01-25 00:00:00 +00:00
+date: 2010-07-11 00:00:00 +01:00
 default_executable: 
 dependencies: 
 - !ruby/object:Gem::Dependency 
   name: distance_measures
-  type: :runtime
-  version_requirement: 
-  version_requirements: !ruby/object:Gem::Requirement 
+  prerelease: false
+  requirement: &id001 !ruby/object:Gem::Requirement 
     requirements: 
     - - ">="
       - !ruby/object:Gem::Version 
+        segments: 
+        - 0
+        - 0
+        - 0
         version: 0.0.0
-    version: 
+  type: :runtime
+  version_requirements: *id001
 description: Attempting to create a fast, memory efficient KMeans
 email: reddavis@gmail.com
 executables: []
@@ -64,18 +73,20 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements: 
   - - ">="
     - !ruby/object:Gem::Version 
+      segments: 
+      - 0
       version: "0"
-  version: 
 required_rubygems_version: !ruby/object:Gem::Requirement 
   requirements: 
   - - ">="
     - !ruby/object:Gem::Version 
+      segments: 
+      - 0
       version: "0"
-  version: 
 requirements: []
 
 rubyforge_project: 
-rubygems_version: 1.3.5
+rubygems_version: 1.3.6
 signing_key: 
 specification_version: 3
 summary: K Means algorithm