Completed first draft

.coverage

@@ -0,0 +1 @@
+!coverage.py: This is a private format, don't read it directly!{"lines":{"/Users/lucacappelletti/github/miur_daad_balancing/tests/__init__.py":[1],"/Users/lucacappelletti/github/miur_daad_balancing/tests/test_balanced.py":[1,2,3,4,6,7,8,9,10,11,12,13,14],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/__init__.py":[1,2,3,4,7],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/__init__.py":[1,2,3,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/load_balanced.py":[1,2,4,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/load_data.py":[1,2,3,5,6,7],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/current_path.py":[1,3,5,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/load_full_balanced.py":[1,2,4,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/utils/get_classes.py":[1,2,4,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/umbalanced.py":[1,2,4,6],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/balanced.py":[1,2,3,4,6,8,9],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/originals/__init__.py":[1,2,5],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/originals/sampling_class_portion.py":[1,4,9,10,11,12,15,16,17,18,19,20,22,23,26,27,28,29,30,31,32,35,36,37,39],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/originals/truncate_sample_size.py":[1,4,34,35,36,37,40,41,42,43,45,47,49,51,52,54,55,56,57,58,61,62,63,65],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/full_balanced.py":[1,2,3,4,5,7,9,10,11],"/Users/lucacappelletti/github/miur_daad_balancing/tests/utils/__init__.py":[1,2,5],"/Users/lucacappelletti/github/miur_daad_balancing/tests/utils/sample_data.py":[1,2,3,4,6,8,9,10,11],"/Users/lucacappelletti/github/miur_daad_balancing/tests/utils/compare_tuples.py":[1,2,4,6,7],"/Users/lucacappelletti/github/miur_daad_balancing/tests/test_full_balanced.py":[1,2,3,4,6,7,8,9,10,11,12,13,14,15,16],"/Users/lucacappelletti/github/miur_daad_balancing/tests/test_import.py":[1,3,4],"/Users/lucacappelletti/github/miur_daad_balancing/tests/test_umbalanced.py":[1,2,3,5,6,7,8,9],"/Users/lucacappelletti/github/miur_daad_balancing/tests/test_version.py":[1,2,4,5],"/Users/lucacappelletti/github/miur_daad_balancing/miur_daad_balancing/__version__.py":[1,2]}}

.coverage.MBP-di-Luca.8199.029523

@@ -0,0 +1 @@
+!coverage.py: This is a private format, don't read it directly!{"lines":{}}

README.rst

@@ -46,7 +46,7 @@ Testing dataset balancing techniques from previous works.
     :alt: Pypi project
 
 .. |downloads| image:: https://pepy.tech/badge/miur_daad_balancing
-    :target: https://pepy.tech/badge/miur_daad_balancing
+    :target: https://pepy.tech/badge/miur-daad-balancing
     :alt: Pypi total project downloads 
 
 .. |codacy|  image:: https://api.codacy.com/project/badge/Grade/b4a7c72f058b433597426fa696d71539

miur_daad_balancing/__init__.py

@@ -1,6 +1,8 @@
 from .utils import get_classes, load_balanced, load_full_balanced
 from .umbalanced import umbalanced
+from .balanced import balanced
+from .full_balanced import full_balanced
 
 __all__ = [
-    "load_balanced", "load_full_balanced", "get_classes", "umbalanced"
+    "load_balanced", "load_full_balanced", "get_classes", "umbalanced", "balanced", "full_balanced"
 ]

miur_daad_balancing/balanced.py

@@ -1,8 +1,9 @@
 from typing import Tuple
+from .utils import load_balanced
+from .originals import truncate_sample_size
 import numpy as np
 
 def balanced(training:np.ndarray, testing:np.ndarray)->Tuple[np.ndarray, np.ndarray]:
     """Return balanced training data to the given maximum, leaving testing untouched."""
-    X_train, y_train = training
-    unique, counts = np.unique(y_train, return_counts=True)
-
+    X_train, y_train, _, _ = truncate_sample_size(*training, max_size_given=load_balanced()["max"])
+    return (X_train, y_train), testing

miur_daad_balancing/full_balanced.py

@@ -0,0 +1,11 @@
+from typing import Tuple
+from .utils import load_full_balanced
+from .originals import sampling_class_portion
+from .balanced import balanced
+import numpy as np
+
+def full_balanced(training:np.ndarray, testing:np.ndarray)->Tuple[np.ndarray, np.ndarray]:
+    """Return full_balanced training data to the given maximum and testing data with updated proportions."""
+    training, _ = balanced(training, testing)
+    X_test, y_test, _, _ = sampling_class_portion(*testing, class_portion=load_full_balanced())
+    return training, (X_test, y_test)

miur_daad_balancing/originals/__init__.py

@@ -0,0 +1,6 @@
+from .sampling_class_portion import sampling_class_portion
+from .truncate_sample_size import truncate_sample_size
+
+__all__ = [
+    "sampling_class_portion", "truncate_sample_size"
+]

tests/utils/sampling_class_portion.py → miur_daad_balancing/originals/sampling_class_portion.py

@@ -1,7 +1,7 @@
 import numpy as np
 
 
-def sampling_class_portion(data, classes, others=None, class_portion=None, rng=np.random.RandomState(100)):
+def sampling_class_portion(data, classes, others=None, class_portion=None):
     """
     Sampling data points in each class to keep a given portion among classes.
     class_portion: dict, the portion for each class, each value should be at least 1, e.g. class_portion={"class0":5,"class1":1,"class3":2}
@@ -27,7 +27,7 @@ def sampling_class_portion(data, classes, others=None, class_portion=None, rng=n
         ind_this_num = indices_range[indices == i]
         replacetf = True if sample_sizes[u[i]] < (
             size_min*class_portion[u[i]]) else False
-        ind_this_reduced = ind_this_num[rng.choice(
+        ind_this_reduced = ind_this_num[np.random.choice(
             sample_sizes[u[i]], size=size_min*class_portion[u[i]], replace=replacetf)]
         indices_all = np.append(indices_all, ind_this_reduced)
 

tests/utils/truncate_sample_size.py → miur_daad_balancing/originals/truncate_sample_size.py

@@ -1,7 +1,7 @@
 import numpy as np
 
 
-def truncate_sample_size(data, classes, others=None, max_size_given=None, rng=np.random.RandomState(100)):
+def truncate_sample_size(data, classes, others=None, max_size_given=None):
     """
     Balance sample size of a data set among classes.
 
@@ -42,7 +42,6 @@ def truncate_sample_size(data, classes, others=None, max_size_given=None, rng=np
         sample_sizes.append(sample_size_this)
     sample_sizes = np.array(sample_sizes, dtype=int)
 
-    size_min = np.amin(sample_sizes)  # smallest sample size
     size_max = np.amax(sample_sizes)  # largest sample size
 
     if size_max < max_size_given:
@@ -54,7 +53,7 @@ def truncate_sample_size(data, classes, others=None, max_size_given=None, rng=np
 
     for i in range(num_u):
         ind_this_num = indices_range[indices == i]
-        ind_this_reduced = ind_this_num[rng.choice(
+        ind_this_reduced = ind_this_num[np.random.choice(
             len(ind_this_num), size=sample_sizes[i], replace=False)]
         indices_all = np.append(indices_all, ind_this_reduced)
 

tests/test_balanced.py

@@ -1,10 +1,14 @@
 from miur_daad_balancing import balanced, load_balanced
 import numpy as np
-from .utils import sample_data, compare_tuples, truncate_sample_size
+from miur_daad_balancing.originals import truncate_sample_size, sampling_class_portion
+from .utils import sample_data, compare_tuples
 
-def test_umbalanced():
+def test_balanced():
     training, testing = sample_data()
+    np.random.seed(42)
     X_train, y_train, _, _ = truncate_sample_size(*training, max_size_given=load_balanced()["max"])
-    balanced_training = (X_train, y_train)
-    #assert compare_tuples(training, balanced_training)
-    #assert compare_tuples(testing, balanced_testing)
+    original_balanced_training = (X_train, y_train)
+    np.random.seed(42)
+    balanced_training, balanced_testing = balanced(training, testing)
+    assert compare_tuples(original_balanced_training, balanced_training)
+    assert compare_tuples(testing, balanced_testing)

tests/test_full_balanced.py

@@ -1,12 +1,16 @@
-from miur_daad_balancing import full_balanced, load_full_balanced, load_balanced
+from miur_daad_balancing import full_balanced, load_balanced, load_full_balanced
 import numpy as np
-from .utils import sample_data, compare_tuples, sampling_class_portion, truncate_sample_size
+from miur_daad_balancing.originals import truncate_sample_size, sampling_class_portion
+from .utils import sample_data, compare_tuples
 
-def test_umbalanced():
+def test_full_balanced():
     training, testing = sample_data()
+    np.random.seed(42)
     X_train, y_train, _, _ = truncate_sample_size(*training, max_size_given=load_balanced()["max"])
-    balanced_training = (X_train, y_train)
+    original_balanced_training = (X_train, y_train)
+    np.random.seed(42)
     X_test, y_test, _, _ = sampling_class_portion(*testing, class_portion=load_full_balanced())
-    balanced_testing = (X_test, y_test)
-    #assert compare_tuples(training, balanced_training)
-    #assert compare_tuples(testing, balanced_testing)
+    original_balanced_testing = (X_test, y_test)
+    np.random.seed(42)
+    balanced_training, balanced_testing = full_balanced(training, testing)
+    assert compare_tuples(original_balanced_training, balanced_training)

tests/utils/__init__.py

@@ -1,8 +1,6 @@
-from .sampling_class_portion import sampling_class_portion
-from .truncate_sample_size import truncate_sample_size
 from .sample_data import sample_data
 from .compare_tuples import compare_tuples
 
 __all__ = [
-    "sampling_class_portion", "truncate_sample_size", "compare_tuples"
+    "compare_tuples", "sample_data"
 ]