lib/disco/recommender.rb in disco-0.2.3 vs lib/disco/recommender.rb in disco-0.2.4

@@ -1,34 +1,34 @@
 module Disco
   class Recommender
-    attr_reader :global_mean, :item_factors, :user_factors
+    attr_reader :global_mean
 
     def initialize(factors: 8, epochs: 20, verbose: nil)
       @factors = factors
       @epochs = epochs
       @verbose = verbose
+      @user_map = {}
+      @item_map = {}
     end
 
     def fit(train_set, validation_set: nil)
       train_set = to_dataset(train_set)
       validation_set = to_dataset(validation_set) if validation_set
 
-      @implicit = !train_set.any? { |v| v[:rating] }
+      check_training_set(train_set)
 
+      @implicit = !train_set.any? { |v| v[:rating] }
       unless @implicit
-        ratings = train_set.map { |o| o[:rating] }
-        check_ratings(ratings)
-        @min_rating = ratings.min
-        @max_rating = ratings.max
+        check_ratings(train_set)
+        @min_rating, @max_rating = train_set.minmax_by { |o| o[:rating] }.map { |o| o[:rating] }
 
         if validation_set
-          check_ratings(validation_set.map { |o| o[:rating] })
+          check_ratings(validation_set)
         end
       end
 
-      check_training_set(train_set)
-      create_maps(train_set)
+      update_maps(train_set)
 
       @rated = Hash.new { |hash, key| hash[key] = {} }
       input = []
       value_key = @implicit ? :value : :rating
       train_set.each do |v|
@@ -141,26 +141,55 @@
       @user_index = create_index(@user_factors)
     end
 
     def similar_items(item_id, count: 5)
       check_fit
-      similar(item_id, @item_map, @item_factors, item_norms, count, @item_index)
+      similar(item_id, @item_map, @item_factors, @item_index ? nil : item_norms, count, @item_index)
     end
     alias_method :item_recs, :similar_items
 
     def similar_users(user_id, count: 5)
       check_fit
-      similar(user_id, @user_map, @user_factors, user_norms, count, @user_index)
+      similar(user_id, @user_map, @user_factors, @user_index ? nil : user_norms, count, @user_index)
     end
 
+    def user_ids
+      @user_map.keys
+    end
+
+    def item_ids
+      @item_map.keys
+    end
+
+    def user_factors(user_id = nil)
+      if user_id
+        u = @user_map[user_id]
+        @user_factors[u, true] if u
+      else
+        @user_factors
+      end
+    end
+
+    def item_factors(item_id = nil)
+      if item_id
+        i = @item_map[item_id]
+        @item_factors[i, true] if i
+      else
+        @item_factors
+      end
+    end
+
     private
 
     def create_index(factors)
       require "ngt"
 
+      # could speed up search with normalized cosine
+      # https://github.com/yahoojapan/NGT/issues/36
       index = Ngt::Index.new(factors.shape[1], distance_type: "Cosine")
-      index.batch_insert(factors)
+      ids = index.batch_insert(factors)
+      raise "Unexpected ids. Please report a bug." if ids.first != 1 || ids.last != factors.shape[0]
       index
     end
 
     def user_norms
       @user_norms ||= norms(@user_factors)
@@ -189,45 +218,46 @@
               # convert cosine distance to cosine similarity
               score: 1 - v[:distance]
             }
           end
         else
-          predictions = factors.dot(factors[i, true]) / norms
+          # cosine similarity without norms[i]
+          # otherwise, denominator would be (norms[i] * norms)
+          predictions = factors.inner(factors[i, true]) / norms
 
           predictions =
             map.keys.zip(predictions).map do |item_id, pred|
               {item_id: item_id, score: pred}
             end
 
-          max_score = predictions.delete_at(i)[:score]
+          predictions.delete_at(i)
           predictions.sort_by! { |pred| -pred[:score] } # already sorted by id
           predictions = predictions.first(count) if count
-          # divide by max score to get cosine similarity
+          # divide by norms[i] to get cosine similarity
           # only need to do for returned records
-          predictions.each { |pred| pred[:score] /= max_score }
+          predictions.each { |pred| pred[:score] /= norms[i] }
           predictions
         end
       else
         []
       end
     end
 
-    def create_maps(train_set)
-      user_ids = train_set.map { |v| v[:user_id] }.uniq.sort
-      item_ids = train_set.map { |v| v[:item_id] }.uniq.sort
+    def update_maps(train_set)
+      raise ArgumentError, "Missing user_id" if train_set.any? { |v| v[:user_id].nil? }
+      raise ArgumentError, "Missing item_id" if train_set.any? { |v| v[:item_id].nil? }
 
-      raise ArgumentError, "Missing user_id" if user_ids.any?(&:nil?)
-      raise ArgumentError, "Missing item_id" if item_ids.any?(&:nil?)
-
-      @user_map = user_ids.zip(user_ids.size.times).to_h
-      @item_map = item_ids.zip(item_ids.size.times).to_h
+      train_set.each do |v|
+        @user_map[v[:user_id]] ||= @user_map.size
+        @item_map[v[:item_id]] ||= @item_map.size
+      end
     end
 
     def check_ratings(ratings)
-      unless ratings.all? { |r| !r.nil? }
+      unless ratings.all? { |r| !r[:rating].nil? }
         raise ArgumentError, "Missing ratings"
       end
-      unless ratings.all? { |r| r.is_a?(Numeric) }
+      unless ratings.all? { |r| r[:rating].is_a?(Numeric) }
         raise ArgumentError, "Ratings must be numeric"
       end
     end
 
     def check_training_set(train_set)