Merge pull request #264 from understandable-machine-intelligence-lab/…

…review-dist-metrics

Review distance metrics and added squared distance as measure

quantus/functions/similarity_func.py

@@ -183,7 +183,7 @@ def lipschitz_constant(
 
 def abs_difference(a: np.array, b: np.array, **kwargs) -> float:
     """
-    Calculate the absolute difference between two images (or explanations).
+    Calculate the mean of the absolute differences between two images (or explanations).
 
     Parameters
     ----------
@@ -202,6 +202,27 @@ def abs_difference(a: np.array, b: np.array, **kwargs) -> float:
     return np.mean(abs(a - b))
 
 
+def squared_difference(a: np.array, b: np.array, **kwargs) -> float:
+    """
+    Calculate the sqaured differences between two images (or explanations).
+
+    Parameters
+    ----------
+    a: np.ndarray
+         The first array to use for similarity scoring.
+    b: np.ndarray
+         The second array to use for similarity scoring.
+    kwargs: optional
+        Keyword arguments.
+
+    Returns
+    -------
+    float
+        The similarity score.
+    """
+    return np.sum((a - b) ** 2)
+
+
 def cosine(a: np.array, b: np.array, **kwargs) -> float:
     """
     Calculate Cosine of two images (or explanations).
@@ -250,7 +271,7 @@ def ssim(a: np.array, b: np.array, **kwargs) -> float:
     )
 
 
-def difference(a: np.array, b: np.array, **kwargs) -> float:
+def difference(a: np.array, b: np.array, **kwargs) -> np.array:
     """
     Calculate the difference between two images (or explanations).
 
@@ -265,7 +286,7 @@ def difference(a: np.array, b: np.array, **kwargs) -> float:
 
     Returns
     -------
-    float
-        The similarity score.
+    np.array
+        The difference in each element.
     """
     return a - b

quantus/helpers/constants.py

@@ -90,6 +90,7 @@
     "distance_chebyshev": distance_chebyshev,
     "lipschitz_constant": lipschitz_constant,
     "abs_difference": abs_difference,
+    "squared_difference": squared_difference,
     "difference": difference,
     "cosine": cosine,
     "ssim": ssim,

requirements_test.txt

@@ -13,7 +13,7 @@ tf-explain>=0.3.1
 zennit>=0.4.5; python_version >= '3.7'
 tensorflow>=2.5.0; python_version == '3.7'
 tensorflow>=2.12.0; sys_platform != 'darwin' and python_version > '3.7'
-tensorflow_macos>=2.12.0; sys_platform == 'darwin' and python_version > '3.7'
+tensorflow_macos>=2.9.0; sys_platform == 'darwin' and python_version > '3.7'
 torch<=1.11.0; python_version == '3.7'
 torch>=1.13.1; sys_platform != 'linux' and python_version > '3.7'
 torch>=1.13.1, <2.0.0; sys_platform == 'linux' and python_version > '3.7' and python_version <= '3.10'

tests/conftest.py

@@ -2,7 +2,8 @@
 import pickle
 import torch
 import numpy as np
-from keras.datasets import cifar10
+import tensorflow
+from tensorflow.keras.datasets import cifar10
 import pandas as pd
 from sklearn.model_selection import train_test_split
 

tests/helpers/test_similarity_func.py

@@ -54,6 +54,20 @@ def atts_ssim_diff():
     return {"a": np.zeros((16, 16)), "b": np.ones((16, 16))}
 
 
+@pytest.fixture
+def atts_sq_diff_1():
+    return {"a": np.array([1, 2, 3]), "b": np.array([1, 2, 3])}
+
+@pytest.fixture
+def atts_sq_diff_2():
+    return {"a": np.array([1, 2, 3]), "b": np.array([4, 5, 6])}
+
+@pytest.fixture
+def atts_sq_diff_3():
+    return {"a": np.array([1, 2, 3]), "b": np.array([4, 5])}
+
+
+
 @pytest.mark.similar_func
 @pytest.mark.parametrize(
     "data,params,expected",
@@ -270,3 +284,21 @@ def test_mse(data: np.ndarray, params: dict, expected: Union[float, dict, bool])
 def test_difference(data: np.ndarray, params: dict, expected: Union[float, dict, bool]):
     out = difference(a=data["a"], b=data["b"])
     assert all(out == expected), "Test failed."
+
+
+@pytest.mark.similar_func
+@pytest.mark.parametrize(
+    "data,params,expected",
+    [
+        (lazy_fixture("atts_sq_diff_1"), {}, 0),
+        (lazy_fixture("atts_sq_diff_2"), {}, 27),
+        (lazy_fixture("atts_sq_diff_3"), {}, ValueError),
+    ],
+)
+def test_squared_difference(data: np.ndarray, params: dict, expected: Union[int, ValueError]):
+    try:
+        out = squared_difference(a=data["a"], b=data["b"])
+        assert out == expected, "Test failed."
+    except ValueError:
+        pass
+