Fixed base dataset class and adjusted the rest to use new gridsearch function

Author: Michal-Fularz

decision_trees/datasets/dataset_base.py

@@ -3,7 +3,8 @@
 
 import numpy as np
 
-from decision_trees import dataset_tester
+from decision_trees.dataset_tester import test_dataset
+from decision_trees.utils.constants import ClassifierType, GridSearchType
 
 
 class DatasetBase(metaclass=abc.ABCMeta):
@@ -16,18 +17,19 @@ def load_data(self) -> Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
     def _normalise(data: np.ndarray) -> np.ndarray:
         pass
 
-    def run(self):
+    def test_as_classifier(self, number_of_bits_per_feature: int):
         train_data, train_target, test_data, test_target = self.load_data()
 
-        dataset_tester.test_dataset(8,
-                                    train_data, train_target, test_data, test_target,
-                                    dataset_tester.ClassifierType.DECISION_TREE,
-                                    )
-
-    def run_grid_search(self):
-        train_data, train_target, test_data, test_target = self.load_data()
-
-        dataset_tester.grid_search(train_data, train_target,
-                                   test_data, test_target,
-                                   dataset_tester.ClassifierType.DECISION_TREE
-                                   )
+        print('Testing decision tree classifier')
+        test_dataset(
+            number_of_bits_per_feature,
+            train_data, train_target, test_data, test_target,
+            ClassifierType.DECISION_TREE
+        )
+
+        print('Testing random forest classifier')
+        test_dataset(
+            number_of_bits_per_feature,
+            train_data, train_target, test_data, test_target,
+            ClassifierType.RANDOM_FOREST
+        )

decision_trees/datasets/emg_raw.py

@@ -4,7 +4,6 @@
 import numpy as np
 
 from decision_trees.datasets.dataset_base import DatasetBase
-from decision_trees.dataset_tester import test_dataset
 from decision_trees.gridsearch import perform_gridsearch
 from decision_trees.utils.constants import ClassifierType, GridSearchType
 
@@ -95,16 +94,13 @@ def _normalise(self, data: np.ndarray):
     print(f"np.unique(train_target): {np.unique(train_target)}")
     print(f"np.unique(test_target): {np.unique(test_target)}")
 
-    test_dataset(8,
-                 train_data, train_target,
-                 test_data, test_target,
-                 ClassifierType.RANDOM_FOREST,
-                 )
+    d.test_as_classifier(16)
 
     perform_gridsearch(train_data, train_target,
                        test_data, test_target,
-                       10 - 1,
+                       [16, 12, 8, 6, 4, 2, 1],
                        ClassifierType.RANDOM_FOREST,
-                       GridSearchType.NONE,
-                       './../../data/gridsearch_results/'
+                       GridSearchType.PARFIT,
+                       './../../data/gridsearch_results/',
+                       d.__class__.__name__
                        )

decision_trees/datasets/fashion_mnist_raw.py

@@ -32,7 +32,7 @@ def _normalise(data: np.ndarray):
         return data
 
 
-def test_mnist_raw():
+def test_fashion_mnist_raw():
     #####################################
     # SET THE FOLLOWING PARAMETERS
     # MNIST FASHION DATABASE
@@ -42,7 +42,7 @@ def test_mnist_raw():
     #####################################
 
     d = FashionMnistRaw()
-    d.run()
+    d.test_as_classifier(8)
 
     assert True
 
@@ -56,19 +56,27 @@ def main():
     print(f"np.unique(test_target): {np.unique(test_target)}")
 
     from decision_trees import dataset_tester
-
-    dataset_tester.perform_gridsearch(train_data[:60000], train_target[:60000],
-                                      test_data[:10000], test_target[:10000],
-                                      10 - 1,
-                                      dataset_tester.ClassifierType.DECISION_TREE,
-                                      dataset_tester.GridSearchType.NONE,
-                                      "./../../data/gridsearch_results/"
-                                      )
-
-    # dataset_tester.test_dataset(4,
-    #                             train_data[:60000], train_target[:60000], test_data[:10000], test_target[:10000],
-    #                             dataset_tester.ClassifierType.DECISION_TREE,
-    #                             )
+    from decision_trees.gridsearch import perform_gridsearch
+    from decision_trees.utils.constants import ClassifierType, GridSearchType
+
+    perform_gridsearch(
+        train_data[:60000], train_target[:60000],
+        test_data[:10000], test_target[:10000],
+        [16, 12, 8, 6, 4, 2, 1],
+        ClassifierType.DECISION_TREE,
+        GridSearchType.NONE,
+        "./../../data/gridsearch_results/",
+        d.__class__.__name__
+    )
+
+    # this is the same as the code below, but on the whole dataset
+    d.test_as_classifier(8)
+
+    dataset_tester.test_dataset(
+        8,
+        train_data[:60000], train_target[:60000], test_data[:10000], test_target[:10000],
+        ClassifierType.DECISION_TREE
+    )
 
 
 if __name__ == "__main__":

decision_trees/datasets/mnist_raw.py

@@ -49,10 +49,10 @@ def test_mnist_raw():
     number_of_test_samples = 100  # 70000 - number_of_train_samples
     # END OF PARAMETERS SETTING
     if (number_of_train_samples + number_of_test_samples) > 70000:
-        print("ERROR, too much samples set!")
+        print("ERROR, too many samples set!")
     #####################################
 
     d = MnistRaw(number_of_train_samples, number_of_test_samples)
-    d.run()
+    d.test_as_classifier(8)
 
     assert True

decision_trees/datasets/terrain.py

@@ -6,7 +6,6 @@
 import numpy as np
 
 from decision_trees.datasets.dataset_base import DatasetBase
-from decision_trees.dataset_tester import test_dataset
 from decision_trees.gridsearch import perform_gridsearch
 from decision_trees.utils.constants import ClassifierType, GridSearchType
 
@@ -118,16 +117,14 @@ def main():
     print(f'np.shape(train_data): {np.shape(train_data)}')
     print(f'np.unique(test_target): {np.unique(test_target)}')
 
-    test_dataset(32,
-                 train_data, train_target, test_data, test_target,
-                 ClassifierType.RANDOM_FOREST
-                 )
+    d.test_as_classifier(16)
 
     perform_gridsearch(train_data, train_target, test_data, test_target,
-                       10 - 1,
+                       [16, 12, 8, 6, 4, 2, 1],
                        ClassifierType.RANDOM_FOREST,
                        GridSearchType.NONE,
-                       './../../data/gridsearch_results/'
+                       './../../data/gridsearch_results/',
+                       d.__class__.__name__
                        )