GH-55: Adding support for computing TF-IDF vector representations.

Sources/Accord.MachineLearning/Accord.MachineLearning.csproj

@@ -51,6 +51,7 @@
     <Compile Include="Clustering\KMeans\BalancedKMeans.cs" />
     <Compile Include="DecisionTrees\DecisionTreeHelper.cs" />
     <Compile Include="BaseKNearestNeighbors`2.cs" />
+    <Compile Include="Representations\TFIDF.cs" />
     <Compile Include="Representations\BagOfWords.cs" />
     <Compile Include="Rules\AssociationRule.cs" />
     <Compile Include="Rules\Apriori.cs" />
@@ -174,6 +175,8 @@
     <Compile Include="VectorMachines\Learning\Base\BaseSupportVectorRegression.cs" />
     <Compile Include="VectorMachines\Learning\Base\ISupportVectorMachineLearning.cs" />
     <Compile Include="VectorMachines\Learning\LeastSquaresLearning.cs" />
+    <Compile Include="VectorMachines\Learning\AveragedStochasticGradientDescent.cs" />
+    <Compile Include="VectorMachines\Learning\StochasticGradientDescent.cs" />
     <Compile Include="VectorMachines\Learning\LinearCoordinateDescent.cs" />
     <Compile Include="VectorMachines\Learning\LinearDualCoordinateDescent.cs" />
     <Compile Include="VectorMachines\Learning\LinearNewtonMethod.cs" />

Sources/Accord.MachineLearning/Representations/BagOfWords.cs

@@ -51,6 +51,7 @@ namespace Accord.MachineLearning
     /// 
     [Serializable]
     public class BagOfWords : ParallelLearningBase, IBagOfWords<string[]>,
+        ITransform<string[], Sparse<double>>,
         IUnsupervisedLearning<BagOfWords, string[], int[]>
     {
         // TODO: Replace by TwoWayDictionary
@@ -125,6 +126,7 @@ public IDictionary<int, string> CodeToString
         /// 
         /// <param name="texts">The texts to build the bag of words model from.</param>
         /// 
+        [Obsolete("Please use the default constructor and pass the texts to the Learn() method.")]
         public BagOfWords(params string[][] texts)
         {
             if (texts == null)
@@ -139,6 +141,7 @@ public BagOfWords(params string[][] texts)
         /// 
         /// <param name="texts">The texts to build the bag of words model from.</param>
         /// 
+        [Obsolete("Please use the default constructor and pass the texts to the Learn() method.")]
         public BagOfWords(params string[] texts)
         {
             if (texts == null)
@@ -256,6 +259,31 @@ public double[] Transform(string[] input, double[] result)
             return result;
         }
 
+        /// <summary>
+        /// Applies the transformation to an input, producing an associated output.
+        /// </summary>
+        /// <param name="input">The input data to which
+        /// the transformation should be applied.</param>
+        /// <param name="result">The location to where to store the
+        /// result of this transformation.</param>
+        /// <returns>The output generated by applying this
+        /// transformation to the given input.</returns>
+        public Sparse<double> Transform(string[] input, Sparse<double> result)
+        {
+            // Detect all feature words
+            foreach (string word in input)
+            {
+                int j;
+                if (!stringToCode.TryGetValue(word, out j))
+                    continue;
+
+                if (result[j] < MaximumOccurance)
+                    result[j]++;
+            }
+
+            return result;
+        }
+
         /// <summary>
         /// Applies the transformation to an input, producing an associated output.
         /// </summary>
@@ -388,5 +416,39 @@ public BagOfWords Learn(string[][] x, double[] weights = null)
 
             return this;
         }
+
+
+
+
+
+        Sparse<double> ITransform<string[], Sparse<double>>.Transform(string[] input)
+        {
+            return Sparse.FromDense(Transform(input));
+        }
+
+        Sparse<double>[] ITransform<string[], Sparse<double>>.Transform(string[][] input)
+        {
+            return Transform(input, new Sparse<double>[input.Length]);
+        }
+
+        /// <summary>
+        /// Applies the transformation to a set of input vectors,
+        /// producing an associated set of output vectors.
+        /// </summary>
+        /// <param name="input">The input data to which
+        /// the transformation should be applied.</param>
+        /// <param name="result">The location to where to store the
+        /// result of this transformation.</param>
+        /// <returns>The output generated by applying this
+        /// transformation to the given input.</returns>
+        public Sparse<double>[] Transform(string[][] input, Sparse<double>[] result)
+        {
+            var t = ((ITransform<string[], Sparse<double>>)this);
+            Parallel.For(0, input.Length, ParallelOptions, i =>
+            {
+                result[i] = t.Transform(input[i]);
+            });
+            return result;
+        }
     }
 }