GetSummaryDataView/Row implementation for Pca and Linear Predictors (dotnet#185)

* Implement `ICanGetSummaryAsIDataView` on `PcaPredictor` class
* Implement `ICanGetSummaryAsIRow` on `LinearPredictor` class

Author: ganik

src/Microsoft.ML.PCA/PcaTrainer.cs

@@ -307,6 +307,7 @@ public static CommonOutputs.AnomalyDetectionOutput TrainPcaAnomaly(IHostEnvironm
     // REVIEW: move the predictor to a different file and fold EigenUtils.cs to this file.
     public sealed class PcaPredictor : PredictorBase<Float>,
         IValueMapper,
+        ICanGetSummaryAsIDataView,
         ICanSaveInTextFormat, ICanSaveModel, ICanSaveSummary
     {
         public const string LoaderSignature = "pcaAnomExec";
@@ -468,6 +469,26 @@ public void SaveAsText(TextWriter writer, RoleMappedSchema schema)
             }
         }
 
+        public IDataView GetSummaryDataView(RoleMappedSchema schema)
+        {
+            var bldr = new ArrayDataViewBuilder(Host);
+
+            var cols = new VBuffer<Float>[_rank + 1];
+            var names = new string[_rank + 1];
+            for (var i = 0; i < _rank; ++i)
+            {
+                names[i] = "EigenVector" + i;
+                cols[i] = _eigenVectors[i];
+            }
+            names[_rank] = "MeanVector";
+            cols[_rank] = _mean;
+
+            bldr.AddColumn("VectorName", names);
+            bldr.AddColumn("VectorData", NumberType.R4, cols);
+
+            return bldr.GetDataView();
+        }
+
         public ColumnType InputType
         {
             get { return _inputType; }

src/Microsoft.ML.StandardLearners/Standard/LinearPredictor.cs

@@ -44,6 +44,7 @@ public abstract class LinearPredictor : PredictorBase<Float>,
         ICanSaveInTextFormat,
         ICanSaveInSourceCode,
         ICanSaveModel,
+        ICanGetSummaryAsIRow,
         ICanSaveSummary,
         IPredictorWithFeatureWeights<Float>,
         IWhatTheFeatureValueMapper,
@@ -343,6 +344,30 @@ public void SaveAsCode(TextWriter writer, RoleMappedSchema schema)
 
         public abstract void SaveSummary(TextWriter writer, RoleMappedSchema schema);
 
+        public virtual IRow GetSummaryIRowOrNull(RoleMappedSchema schema)
+        {
+            var cols = new List<IColumn>();
+
+            var names = default(VBuffer<DvText>);
+            MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, Weight.Length, ref names);
+            var slotNamesCol = RowColumnUtils.GetColumn(MetadataUtils.Kinds.SlotNames,
+                new VectorType(TextType.Instance, Weight.Length), ref names);
+            var slotNamesRow = RowColumnUtils.GetRow(null, slotNamesCol);
+            var colType = new VectorType(NumberType.R4, Weight.Length);
+
+            // Add the bias and the weight columns.
+            var bias = Bias;
+            cols.Add(RowColumnUtils.GetColumn("Bias", NumberType.R4, ref bias));
+            var weights = Weight;
+            cols.Add(RowColumnUtils.GetColumn("Weights", colType, ref weights, slotNamesRow));
+            return RowColumnUtils.GetRow(null, cols.ToArray());
+        }
+
+        public virtual IRow GetStatsIRowOrNull(RoleMappedSchema schema)
+        {
+            return null;
+        }
+
         public abstract void SaveAsIni(TextWriter writer, RoleMappedSchema schema, ICalibrator calibrator = null);
 
         public virtual void GetFeatureWeights(ref VBuffer<Float> weights)
@@ -366,8 +391,7 @@ public ValueMapper<TSrc, VBuffer<Float>> GetWhatTheFeatureMapper<TSrc, TDstContr
 
     public sealed partial class LinearBinaryPredictor : LinearPredictor,
         ICanGetSummaryInKeyValuePairs,
-        IParameterMixer<Float>,
-        ICanGetSummaryAsIRow
+        IParameterMixer<Float>
     {
         public const string LoaderSignature = "Linear2CExec";
         public const string RegistrationName = "LinearBinaryPredictor";
@@ -503,26 +527,7 @@ public IList<KeyValuePair<string, object>> GetSummaryInKeyValuePairs(RoleMappedS
             return results;
         }
 
-        public IRow GetSummaryIRowOrNull(RoleMappedSchema schema)
-        {
-            var cols = new List<IColumn>();
-
-            var names = default(VBuffer<DvText>);
-            MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, Weight.Length, ref names);
-            var slotNamesCol = RowColumnUtils.GetColumn(MetadataUtils.Kinds.SlotNames,
-                new VectorType(TextType.Instance, Weight.Length), ref names);
-            var slotNamesRow = RowColumnUtils.GetRow(null, slotNamesCol);
-            var colType = new VectorType(NumberType.R4, Weight.Length);
-
-            // Add the bias and the weight columns.
-            var bias = Bias;
-            cols.Add(RowColumnUtils.GetColumn("Bias", NumberType.R4, ref bias));
-            var weights = Weight;
-            cols.Add(RowColumnUtils.GetColumn("Weights", colType, ref weights, slotNamesRow));
-            return RowColumnUtils.GetRow(null, cols.ToArray());
-        }
-
-        public IRow GetStatsIRowOrNull(RoleMappedSchema schema)
+        public override IRow GetStatsIRowOrNull(RoleMappedSchema schema)
         {
             if (_stats == null)
                 return null;
@@ -582,8 +587,7 @@ public override void SaveAsIni(TextWriter writer, RoleMappedSchema schema, ICali
 
     public sealed class LinearRegressionPredictor : RegressionPredictor,
         IParameterMixer<Float>,
-        ICanGetSummaryInKeyValuePairs,
-        ICanGetSummaryAsIRow
+        ICanGetSummaryInKeyValuePairs
     {
         public const string LoaderSignature = "LinearRegressionExec";
         public const string RegistrationName = "LinearRegressionPredictor";
@@ -663,30 +667,6 @@ public IList<KeyValuePair<string, object>> GetSummaryInKeyValuePairs(RoleMappedS
 
             return results;
         }
-
-        public IRow GetSummaryIRowOrNull(RoleMappedSchema schema)
-        {
-            var cols = new List<IColumn>();
-
-            var names = default(VBuffer<DvText>);
-            MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, Weight.Length, ref names);
-            var slotNamesCol = RowColumnUtils.GetColumn(MetadataUtils.Kinds.SlotNames,
-                new VectorType(TextType.Instance, Weight.Length), ref names);
-            var slotNamesRow = RowColumnUtils.GetRow(null, slotNamesCol);
-            var colType = new VectorType(NumberType.R4, Weight.Length);
-
-            // Add the bias and the weight columns.
-            var bias = Bias;
-            cols.Add(RowColumnUtils.GetColumn("Bias", NumberType.R4, ref bias));
-            var weights = Weight;
-            cols.Add(RowColumnUtils.GetColumn("Weights", colType, ref weights, slotNamesRow));
-            return RowColumnUtils.GetRow(null, cols.ToArray());
-        }
-
-        public IRow GetStatsIRowOrNull(RoleMappedSchema schema)
-        {
-            return null;
-        }
     }
 
     public sealed class PoissonRegressionPredictor : RegressionPredictor, IParameterMixer<Float>

test/BaselineOutput/Common/EntryPoints/lr-weights.txt

@@ -0,0 +1,8 @@
+#@ TextLoader{
+#@   header+
+#@   sep=tab
+#@   col=Bias:R4:0
+#@   col=Weights:R4:1-9
+#@ }
+Bias	thickness	uniform_size	uniform_shape	adhesion	epit_size	bare_nuclei	bland_chromatin	normal_nucleoli	mitoses
+-6.186806	2.65800762	1.68089855	1.944068	1.42514718	0.8536965	2.9325006	1.74816787	1.58165014	0.595681

test/BaselineOutput/Common/EntryPoints/mc-lr-stats.txt

@@ -0,0 +1,10 @@
+#@ TextLoader{
+#@   header+
+#@   sep=tab
+#@   col={name={Count of training examples} type=I8 src=0}
+#@   col={name={Residual Deviance} type=R4 src=1}
+#@   col={name={Null Deviance} type=R4 src=2}
+#@   col=AIC:R4:3
+#@ }
+Count of training examples	Residual Deviance	Null Deviance	AIC
+683	119.098892	884.3502	159.098892

test/BaselineOutput/Common/EntryPoints/mc-lr-weights.txt

@@ -0,0 +1,10 @@
+#@ TextLoader{
+#@   header+
+#@   sep=tab
+#@   col=Bias:R4:0
+#@   col=Weights:R4:1-9
+#@   col=ClassNames:TX:10
+#@ }
+Bias	thickness	uniform_size	uniform_shape	adhesion	epit_size	bare_nuclei	bland_chromatin	normal_nucleoli	mitoses	ClassNames
+3.36404228	-1.579712	-0.8266232	-1.051891	-0.79305464	-0.386733949	-1.59106934	-1.01550019	-0.8356989	-0.332574666	Class_0
+-3.36404562	1.57971311	0.826623559	1.051891	0.7930542	0.386735022	1.59107041	1.015499	0.8356983	0.332574	Class_1