Fixed cosine distance. Cosine similarity was implemented, rather than distance.

Author: Sergio Fierens

lib/ai4r/data/proximity.rb

@@ -112,7 +112,7 @@ def self.cosine_distance(a,b)
         end
 
         magnitude = Math.sqrt(norm_a) * Math.sqrt(norm_b)
-        dot_product / magnitude
+        return 1 - (dot_product / magnitude)
       end
     end
 

test/data/proximity_test.rb

@@ -14,7 +14,7 @@ module Ai4r
   module Data
     class ProximityTest < Test::Unit::TestCase
 
-      @@delta = 0.0001    
+      @@delta = 0.0000001    
       @@data1 = [rand*10, rand*10, rand*-10]
       @@data2 = [rand*10, rand*-10, rand*10]
 
@@ -77,11 +77,10 @@ def test_simple_matching_distance
       end       
 
       def test_cosine_distance
-          assert_equal 0, Proximity.cosine_distance(@@data1, @@data1)
+          assert_in_delta 0.0, Proximity.cosine_distance(@@data1, @@data1), @@delta
           assert_equal  Proximity.cosine_distance(@@data1, @@data2), 
                         Proximity.cosine_distance(@@data2, @@data1)
-          assert 1 == Proximity.cosine_distance(@@data1, @@data1)
-          assert_equal 1, Proximity.cosine_distance([3,0], [4,0])
+          assert_in_delta 0.0, Proximity.cosine_distance(@@data1, @@data1), @@delta
       end
     end
   end