simple simplification of counting, only need to pass the totals, not …

…the whole classifier

lib/simple_bayes/bayes.rb

@@ -52,7 +52,7 @@ def untrain(name, text)
     def classifications text, default_prob = 0.005
       doc = Document.new text
       categories.values.map do |cat|
-        prob_cat = cat.probability self
+        prob_cat = cat.probability total_unique.to_f
         prob_doc = cat.probability_of_document(doc, default_prob)
         [prob_cat * prob_doc, cat]
       end
@@ -61,7 +61,7 @@ def classifications text, default_prob = 0.005
     def log_classifications text, default_prob = 0.005
       doc = Document.new text
       categories.values.map do |cat|
-        prob_cat = cat.log_probability self
+        prob_cat = cat.log_probability total_unique.to_f
         prob_doc = cat.log_probability_of_document(doc, default_prob)
         [prob_cat + prob_doc, cat]
       end

lib/simple_bayes/category.rb

@@ -8,6 +8,7 @@ module SimpleBayes
   # Categories are going to be a shared idea, so why not break out the
   # common category functionality into its own class.
   class Category
+    MIN_LOG_PROBABILITY = -Float::MAX
     include TermOccurrence
 
     attr_reader :name, :term_occurrences
@@ -17,18 +18,13 @@ def initialize name
       @term_occurrences = Hash.new 0
     end
 
-    def log_probability classifier
-      uniqs = classifier.total_unique.to_f
-      cat_uniqs = total_unique
-      if uniqs > 0 && cat_uniqs > 0
-        Math.log(cat_uniqs) - Math.log(uniqs)
-      else
-        -Float::MAX
-      end
+    def log_probability uniqs
+      tot = total_unique
+      (uniqs > 0 && tot > 0) ? Math.log(tot/uniqs) : MIN_LOG_PROBABILITY
     end
 
-    def probability classifier
-      Math.exp log_probability(classifier)
+    def probability uniqs
+      Math.exp log_probability(uniqs)
     end
 
     # Calculates the probability of a document given this category, ie:
@@ -44,13 +40,12 @@ def probability_of_document doc, default_prob = 0.005
     end
 
     def log_probability_of_document doc, default_prob = 0.005
-      all_occurs = total_occurrences.to_f
-      return -Float::MAX if all_occurs < 1
+      all = total_occurrences.to_f
+      return MIN_LOG_PROBABILITY if all < 1
 
       doc.inject(0) do |sum, (t,_)|
-        t_occurs = occurrences_of(t)
-        log_prob_t = (t_occurs > 0 ? (Math.log(t_occurs) - Math.log(all_occurs)) : Math.log(default_prob) )
-        sum + log_prob_t
+        term = occurrences_of(t)
+        sum + Math.log(term > 0 ? term/all : default_prob)
       end
     end
   end

spec/simple_bayes/category_spec.rb

@@ -2,40 +2,33 @@
 
 describe SimpleBayes::Category do
   let(:category) { SimpleBayes::Category.new 'unnamed' }
-  let(:classifier) { mock('classifier') }
 
   it "should include TermOccurrence" do
     category.should be_a_kind_of(SimpleBayes::TermOccurrence)
   end
 
   it "should calculate its log probability based on the unique terms of itself and a classifier" do
     category.stub(:total_unique => 8)
-    classifier.stub(:total_unique => 32)
-    category.log_probability(classifier).should be_within(1.0e-10).of(Math.log 0.25)
+    category.log_probability(32.0).should be_within(1.0e-10).of(Math.log 0.25)
   end
 
   it "should return the 'biggest' negative float if either it or the classifier lack unique terms" do
     category.stub(:total_unique => 0)
-    classifier.stub(:total_unique => 32)
-    category.log_probability(classifier).should == -Float::MAX
+    category.log_probability(32.0).should == -Float::MAX
     category.stub(:total_unique => 8)
-    classifier.stub(:total_unique => 0)
-    category.log_probability(classifier).should == -Float::MAX
+    category.log_probability(0.0).should == -Float::MAX
   end
 
   it "should calculate its probability based on the unique terms of itself and a classifier" do
     category.stub(:total_unique => 8)
-    classifier.stub(:total_unique => 32)
-    category.probability(classifier).should be_within(1.0e-10).of(0.25)
+    category.probability(32.0).should be_within(1.0e-10).of(0.25)
   end
 
   it "should be close to 0 if either it or the classifier lack unique terms" do
     category.stub(:total_unique => 0)
-    classifier.stub(:total_unique => 32)
-    category.probability(classifier).should be_within(1.0e-10).of(0.0)
+    category.probability(32.0).should be_within(1.0e-10).of(0.0)
     category.stub(:total_unique => 8)
-    classifier.stub(:total_unique => 0)
-    category.probability(classifier).should be_within(1.0e-10).of(0.0)
+    category.probability(0.0).should be_within(1.0e-10).of(0.0)
   end
 
   it "should calculate the log probability of a document given the category" do