LinearRegressor: collect learning data parameter added to default con…

…structor

CHANGELOG.md

@@ -1,9 +1,13 @@
 # Changelog
 
-# 13.9.0
+## 13.10.0
 - `LinearRegressor`:
     - `Default constructor`: `collectLearningData` parameter added
 
+## 13.9.0
+- `LogisticRegressor`:
+    - `Default constructor`: `collectLearningData` parameter added
+
 ## 13.8.1
 - `ml_dataframe` dependency updated
 - `xrange` dependency constrain removed

lib/src/regressor/linear_regressor/linear_regressor.dart

@@ -94,6 +94,11 @@ abstract class LinearRegressor implements Assessable, Serializable, Predictor {
   /// optimization algorithm. [initialCoefficients] should have length that is
   /// equal to the number of features in the [fittingData].
   ///
+  /// [collectLearningData] Whether or not to collect learning data, for
+  /// instance cost function value per each iteration. Affects performance much.
+  /// If [collectLearningData] is true, one may access [costPerIteration]
+  /// getter in order to evaluate learning process more thoroughly.
+  ///
   /// [dtype] A data type for all the numeric values, used by the algorithm. Can
   /// affect performance or accuracy of the computations. Default value is
   /// [DType.float32].
@@ -113,6 +118,7 @@ abstract class LinearRegressor implements Assessable, Serializable, Predictor {
     int batchSize = 1,
     Matrix initialCoefficients,
     bool isFittingDataNormalized = false,
+    bool collectLearningData = false,
     DType dtype = DType.float32,
   }) {
     final optimizer = createSquaredCostOptimizer(
@@ -137,13 +143,16 @@ abstract class LinearRegressor implements Assessable, Serializable, Predictor {
     final coefficients = optimizer.findExtrema(
       initialCoefficients: initialCoefficients,
       isMinimizingObjective: true,
+      collectLearningData: collectLearningData,
     ).getColumn(0);
+    final costPerIteration = optimizer.costPerIteration;
 
     return LinearRegressorImpl(
       coefficients,
       targetName,
       fitIntercept: fitIntercept,
       interceptScale: interceptScale,
+      costPerIteration: costPerIteration,
       dtype: dtype,
     );
   }
@@ -205,4 +214,8 @@ abstract class LinearRegressor implements Assessable, Serializable, Predictor {
   /// A value defining a size of the intercept if [fitIntercept] is
   /// `true`
   num get interceptScale;
+
+  /// Returns a list of cost values per each learning iteration. Returns null
+  /// if the parameter `collectLearningData` of the default constructor is false
+  List<num> get costPerIteration;
 }

lib/src/regressor/linear_regressor/linear_regressor_impl.dart

@@ -25,6 +25,7 @@ class LinearRegressorImpl
   LinearRegressorImpl(this.coefficients, this.targetName, {
     bool fitIntercept = false,
     double interceptScale = 1.0,
+    this.costPerIteration,
     this.dtype = DType.float32,
   }) :
     fitIntercept = fitIntercept,
@@ -56,6 +57,13 @@ class LinearRegressorImpl
   )
   final Vector coefficients;
 
+  @override
+  @JsonKey(
+    name: linearRegressorCostPerIterationJsonKey,
+    includeIfNull: false,
+  )
+  final List<num> costPerIteration;
+
   @override
   @JsonKey(
     name: linearRegressorDTypeJsonKey,

lib/src/regressor/linear_regressor/linear_regressor_json_keys.dart

@@ -3,3 +3,4 @@ const linearRegressorFitInterceptJsonKey = 'FI';
 const linearRegressorInterceptScaleJsonKey = 'IS';
 const linearRegressorCoefficientsJsonKey = 'CS';
 const linearRegressorDTypeJsonKey = 'DT';
+const linearRegressorCostPerIterationJsonKey = 'CPI';

pubspec.yaml

@@ -1,6 +1,6 @@
 name: ml_algo
 description: Machine learning algorithms written in native dart
-version: 13.9.0
+version: 13.10.0
 homepage: https://github.com/gyrdym/ml_algo
 
 environment:

test/regressor/linear_regressor_integration_test.dart

@@ -32,6 +32,7 @@ void main() {
         fitIntercept: fitIntercept,
         interceptScale: interceptScale,
         iterationsLimit: iterationsLimit,
+        collectLearningData: true,
         dtype: dtype,
       );
     });
@@ -53,6 +54,7 @@ void main() {
         linearRegressorInterceptScaleJsonKey: interceptScale,
         linearRegressorCoefficientsJsonKey: regressor.coefficients.toJson(),
         linearRegressorDTypeJsonKey: dTypeToJson(dtype),
+        linearRegressorCostPerIterationJsonKey: regressor.costPerIteration,
       });
     });
 
@@ -75,6 +77,8 @@ void main() {
     test('should restore from json file', () async {
       await regressor.saveAsJson(filePath);
 
+      print(regressor.costPerIteration);
+
       final file = File(filePath);
       final json = await file.readAsString();
       final restoredModel = LinearRegressor.fromJson(json);
@@ -84,6 +88,7 @@ void main() {
       expect(restoredModel.fitIntercept, regressor.fitIntercept);
       expect(restoredModel.coefficients, regressor.coefficients);
       expect(restoredModel.targetName, regressor.targetName);
+      expect(restoredModel.costPerIteration, regressor.costPerIteration);
     });
   });
 }

test/regressor/linear_regressor_test.dart

@@ -20,18 +20,10 @@ import '../mocks.dart';
 
 void main() {
   group('LinearRegressor', () {
-    final initialCoefficients = Matrix.fromList([
-      [1],
-      [2],
-      [3],
-      [4],
-      [5],
-    ]);
-
+    final initialCoefficients = Matrix.column([1, 2, 3, 4, 5]);
     final learnedCoefficients = Matrix.fromColumns([
       Vector.fromList([55, 66, 77, 88, 99]),
     ]);
-
     final observations = DataFrame(
       [
         <num>[10, 20, 30, 40, 200],
@@ -40,17 +32,16 @@ void main() {
       header: ['feature_1', 'feature_2', 'feature_3', 'feature_4', 'target'],
       headerExists: false,
     );
+    final costPerIteration = [1, 2, 3, 100];
 
     CostFunction costFunctionMock;
     CostFunctionFactory costFunctionFactoryMock;
-
     LinearOptimizer linearOptimizerMock;
     LinearOptimizerFactory linearOptimizerFactoryMock;
 
     setUp(() {
       costFunctionMock = CostFunctionMock();
       costFunctionFactoryMock = createCostFunctionFactoryMock(costFunctionMock);
-
       linearOptimizerMock = LinearOptimizerMock();
       linearOptimizerFactoryMock = createLinearOptimizerFactoryMock(
           linearOptimizerMock);
@@ -64,7 +55,9 @@ void main() {
       when(linearOptimizerMock.findExtrema(
         initialCoefficients: anyNamed('initialCoefficients'),
         isMinimizingObjective: anyNamed('isMinimizingObjective'),
+        collectLearningData: anyNamed('collectLearningData'),
       )).thenReturn(learnedCoefficients);
+      when(linearOptimizerMock.costPerIteration).thenReturn(costPerIteration);
     });
 
     tearDownAll(() => injector = null);
@@ -150,27 +143,56 @@ void main() {
         fitIntercept: true,
         interceptScale: 2.0,
       );
-
       final features = Matrix.fromList([
         [55, 44, 33, 22],
         [10, 88, 77, 11],
         [12, 22, 39, 13],
       ]);
-
       final featuresWithIntercept = Matrix.fromColumns([
         Vector.filled(3, 2),
         ...features.columns,
       ]);
-
       final prediction = predictor.predict(
         DataFrame.fromMatrix(features),
       );
 
       expect(prediction.header, equals(['target']));
-
       expect(prediction.toMatrix(), equals(
         featuresWithIntercept * learnedCoefficients,
       ));
     });
+
+    test('should collect cost values per iteration if collectLearningData is '
+        'true', () {
+      final regressor = LinearRegressor(
+        observations,
+        'target',
+        initialCoefficients: initialCoefficients,
+        collectLearningData: true,
+      );
+
+      expect(regressor.costPerIteration, same(costPerIteration));
+      verify(linearOptimizerMock.findExtrema(
+        initialCoefficients: anyNamed('initialCoefficients'),
+        isMinimizingObjective: anyNamed('isMinimizingObjective'),
+        collectLearningData: true,
+      )).called(1);
+    });
+
+    test('should not collect cost values per iteration if collectLearningData is '
+        'false', () {
+      LinearRegressor(
+        observations,
+        'target',
+        initialCoefficients: initialCoefficients,
+        collectLearningData: false,
+      );
+
+      verify(linearOptimizerMock.findExtrema(
+        initialCoefficients: anyNamed('initialCoefficients'),
+        isMinimizingObjective: anyNamed('isMinimizingObjective'),
+        collectLearningData: false,
+      )).called(1);
+    });
   });
 }