Merge b3edad7 into 50ab81e

CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+## 13.3.0
+- `KnnClassifier` class added
+
 ## 13.2.0
 - `KNN algorithm`: standardization for distance added
 - `KnnRegressor`: 

README.md

@@ -20,16 +20,20 @@ the lib, please, do not use it in a browser.
     assessing prediction quality on different parts of a dataset. 
 
 - #### Classification algorithms
-    - [LogisticRegressor](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/logistic_regressor.dart). 
+    - [LogisticRegressor](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/logistic_regressor/logistic_regressor.dart). 
     A class, that performs linear binary classification of data. To use this kind of classifier your data have to be 
     [linearly separable](https://en.wikipedia.org/wiki/Linear_separability).
 
-    - [SoftmaxRegressor](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/softmax_regressor.dart). 
+    - [SoftmaxRegressor](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/softmax_regressor/softmax_regressor.dart). 
     A class, that performs linear multiclass classification of data. To use this kind of classifier your data have to be 
     [linearly separable](https://en.wikipedia.org/wiki/Linear_separability).
 
-    - [DecisionTreeClassifier](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/decision_tree_classifier.dart)
+    - [DecisionTreeClassifier](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/decision_tree_classifier/decision_tree_classifier.dart)
     A class, that performs classification, using decision trees. May work with data with non-linear patterns.
+
+    - [KnnClassifier](https://github.com/gyrdym/ml_algo/blob/master/lib/src/classifier/knn_classifier/knn_classifier.dart)
+    A class, that performs classification, using `k nearest neighbours algorithm` - it makes prediction basing on 
+    first `k` closest observations to the given one.
 
 - #### Regression algorithms
     - [LinearRegressor](https://github.com/gyrdym/ml_algo/blob/master/lib/src/regressor/linear_regressor.dart). A 

benchmark/knn_regressor.dart

@@ -1,7 +1,6 @@
 // 10.0 sec (MacBook Air mid 2017)
 import 'package:benchmark_harness/benchmark_harness.dart';
 import 'package:ml_algo/ml_algo.dart';
-import 'package:ml_algo/src/regressor/knn_regressor_impl.dart';
 import 'package:ml_dataframe/ml_dataframe.dart';
 import 'package:ml_linalg/matrix.dart';
 import 'package:ml_linalg/vector.dart';
@@ -12,10 +11,7 @@ const featuresNum = 20;
 class KnnRegressorBenchmark extends BenchmarkBase {
   KnnRegressorBenchmark() : super('Knn regression benchmark');
 
-  Matrix features;
   DataFrame testFeatures;
-  Matrix labels;
-  Matrix testLabels;
   KnnRegressor regressor;
 
 
@@ -30,9 +26,16 @@ class KnnRegressorBenchmark extends BenchmarkBase {
 
   @override
   void setup() {
-    features = Matrix.fromRows(List.generate(observationsNum * 2,
+    final featureMatrix = Matrix.fromRows(List.generate(observationsNum * 2,
             (i) => Vector.randomFilled(featuresNum)));
-    labels = Matrix.fromColumns([Vector.randomFilled(observationsNum * 2)]);
+
+    final labelMatrix = Matrix
+        .fromColumns([Vector.randomFilled(observationsNum * 2)]);
+
+    final observations = DataFrame.fromMatrix(Matrix.fromColumns([
+      ...featureMatrix.columns,
+      ...labelMatrix.columns,
+    ]));
 
     testFeatures = DataFrame.fromMatrix(
         Matrix.fromRows(
@@ -42,9 +45,8 @@ class KnnRegressorBenchmark extends BenchmarkBase {
             ),
         ),
     );
-    testLabels = Matrix.fromColumns([Vector.randomFilled(observationsNum)]);
 
-    regressor = KnnRegressorImpl(features, labels, 'target', k: 7);
+    regressor = KnnRegressor(observations, observations.header.last, 7);
   }
 
   void tearDown() {}

benchmark/knn_solver.dart

@@ -0,0 +1,45 @@
+// MacBook Air 13.3 mid 2017: ~ 5 sec
+import 'package:benchmark_harness/benchmark_harness.dart';
+import 'package:ml_algo/src/knn_solver/knn_solver.dart';
+import 'package:ml_algo/src/knn_solver/knn_solver_impl.dart';
+import 'package:ml_linalg/distance.dart';
+import 'package:ml_linalg/matrix.dart';
+import 'package:ml_linalg/vector.dart';
+
+const k = 10;
+const trainObservationsNum = 2000;
+const observationsNum = 100;
+const featuresNum = 100;
+
+class KnnSolverBenchmark extends BenchmarkBase {
+  KnnSolverBenchmark() : super('KnnSolver benchmark');
+
+  KnnSolver solver;
+  Matrix features;
+
+  static void main() {
+    KnnSolverBenchmark().report();
+  }
+
+  @override
+  void run() {
+    solver.findKNeighbours(features).toList(growable: false);
+  }
+
+  @override
+  void setup() {
+    final trainFeatures = Matrix.fromRows(List.generate(trainObservationsNum,
+            (i) => Vector.randomFilled(featuresNum)));
+    final trainLabels = Matrix.fromColumns([Vector.randomFilled(trainObservationsNum)]);
+
+    solver = KnnSolverImpl(trainFeatures, trainLabels, k, Distance.euclidean,
+        false);
+
+    features = Matrix.fromRows(List.generate(observationsNum,
+            (i) => Vector.randomFilled(featuresNum)));
+  }
+}
+
+void main() {
+  KnnSolverBenchmark.main();
+}

benchmark/main.dart

@@ -1,6 +1,6 @@
 import 'linear_regressor.dart' as gradient_descent_regression_benchmark;
 import 'logistic_regressor.dart' as logistic_regression_benchmark;
-import 'algorithms/knn.dart' as knn_regressor_benchmark;
+import 'knn_solver.dart' as knn_regressor_benchmark;
 
 Future main() async {
   //  (MacBook Air mid 2017)

lib/ml_algo.dart

@@ -1,13 +1,14 @@
-export 'package:ml_algo/src/algorithms/knn/kernel_type.dart';
-export 'package:ml_algo/src/classifier/decision_tree_classifier.dart';
-export 'package:ml_algo/src/classifier/logistic_regressor.dart';
-export 'package:ml_algo/src/classifier/softmax_regressor.dart';
+export 'package:ml_algo/src/classifier/decision_tree_classifier/decision_tree_classifier.dart';
+export 'package:ml_algo/src/classifier/knn_classifier/knn_classifier.dart';
+export 'package:ml_algo/src/classifier/logistic_regressor/logistic_regressor.dart';
+export 'package:ml_algo/src/classifier/softmax_regressor/softmax_regressor.dart';
+export 'package:ml_algo/src/knn_kernel/kernel_type.dart';
 export 'package:ml_algo/src/linear_optimizer/gradient_optimizer/learning_rate_generator/learning_rate_type.dart';
 export 'package:ml_algo/src/linear_optimizer/linear_optimizer_type.dart';
 export 'package:ml_algo/src/linear_optimizer/regularization_type.dart';
 export 'package:ml_algo/src/metric/classification/type.dart';
 export 'package:ml_algo/src/metric/metric_type.dart';
 export 'package:ml_algo/src/metric/regression/type.dart';
 export 'package:ml_algo/src/model_selection/cross_validator/cross_validator.dart';
-export 'package:ml_algo/src/regressor/knn_regressor.dart';
-export 'package:ml_algo/src/regressor/linear_regressor.dart';
+export 'package:ml_algo/src/regressor/knn_regressor/knn_regressor.dart';
+export 'package:ml_algo/src/regressor/linear_regressor/linear_regressor.dart';

lib/src/classifier/_helpers/log_likelihood_optimizer_factory.dart

@@ -1,6 +1,6 @@
 import 'package:ml_algo/src/cost_function/cost_function_factory.dart';
 import 'package:ml_algo/src/cost_function/cost_function_type.dart';
-import 'package:ml_algo/src/di/injector.dart';
+import 'package:ml_algo/src/di/dependencies.dart';
 import 'package:ml_algo/src/helpers/add_intercept_if.dart';
 import 'package:ml_algo/src/helpers/features_target_split.dart';
 import 'package:ml_algo/src/linear_optimizer/gradient_optimizer/learning_rate_generator/learning_rate_type.dart';
@@ -43,8 +43,6 @@ LinearOptimizer createLogLikelihoodOptimizer(
   final points = splits[0].toMatrix();
   final labels = splits[1].toMatrix();
 
-  final dependencies = getDependencies();
-
   final optimizerFactory = dependencies
       .getDependency<LinearOptimizerFactory>();
 

lib/src/classifier/decision_tree_classifier.dart → lib/src/classifier/decision_tree_classifier/decision_tree_classifier.dart

@@ -1,5 +1,5 @@
 import 'package:ml_algo/src/classifier/classifier.dart';
-import 'package:ml_algo/src/classifier/decision_tree_classifier_impl.dart';
+import 'package:ml_algo/src/classifier/decision_tree_classifier/decision_tree_classifier_impl.dart';
 import 'package:ml_algo/src/decision_tree_solver/solver_factory/greedy_solver.dart';
 import 'package:ml_algo/src/model_selection/assessable.dart';
 import 'package:ml_dataframe/ml_dataframe.dart';

lib/src/classifier/decision_tree_classifier_impl.dart → lib/src/classifier/decision_tree_classifier/decision_tree_classifier_impl.dart

@@ -1,4 +1,4 @@
-import 'package:ml_algo/src/classifier/decision_tree_classifier.dart';
+import 'package:ml_algo/src/classifier/decision_tree_classifier/decision_tree_classifier.dart';
 import 'package:ml_algo/src/predictor/assessable_predictor_mixin.dart';
 import 'package:ml_algo/src/decision_tree_solver/decision_tree_solver.dart';
 import 'package:ml_dataframe/ml_dataframe.dart';

lib/src/classifier/knn_classifier/knn_classifier.dart

@@ -0,0 +1,55 @@
+import 'package:ml_algo/src/classifier/classifier.dart';
+import 'package:ml_algo/src/classifier/knn_classifier/knn_classifier_factory.dart';
+import 'package:ml_algo/src/di/dependencies.dart';
+import 'package:ml_algo/src/knn_kernel/kernel_type.dart';
+import 'package:ml_algo/src/model_selection/assessable.dart';
+import 'package:ml_dataframe/ml_dataframe.dart';
+import 'package:ml_linalg/distance.dart';
+import 'package:ml_linalg/dtype.dart';
+
+/// A class that performs classification basing on `k nearest neighbours` (KNN)
+/// algorithm
+///
+/// K nearest neighbours algorithm is an algorithm that is targeted to search
+/// most similar labelled observations (number of these observations equals `k`)
+/// for the given unlabelled one.
+///
+/// It is possible to use majority class among the k found observations as a
+/// prediction for the given unlabelled observation, but it may lead to the
+/// imprecise result. Thus the weighted version of KNN algorithm is used in the
+/// classifier. To get weight of a particular observation one may use a kernel
+/// function.
+abstract class KnnClassifier implements Assessable, Classifier {
+  /// Parameters:
+  ///
+  /// [fittingData] Labelled observations, among which will be searched [k]
+  /// nearest to the given unlabelled observations neighbours. Must contain
+  /// [targetName] column.
+  ///
+  /// [targetName] A string, that serves as a name of the column, that contains
+  /// labels (or outcomes).
+  ///
+  /// [k] a number of nearest neighbours to be found among [fittingData]
+  ///
+  /// [kernel] a type of a kernel function, that will be used to predict an
+  /// outcome for a new observation
+  ///
+  /// [distance] a distance type, that will be used to measure a distance
+  /// between two observation vectors
+  ///
+  /// [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]
+  factory KnnClassifier(
+      DataFrame fittingData,
+      String targetName,
+      int k,
+      {
+        KernelType kernel = KernelType.gaussian,
+        Distance distance = Distance.euclidean,
+        DType dtype = DType.float32,
+      }
+  ) => dependencies
+      .getDependency<KnnClassifierFactory>()
+      .create(fittingData, targetName, k, kernel, distance, dtype);
+}