Added support for all number of classes in binomial soft labels function

dlordinal/soft_labelling/binomial_distribution.py

@@ -46,20 +46,11 @@ def get_binomial_soft_labels(J):
     if J < 2 or not isinstance(J, int):
         raise ValueError(f"{J=} must be a positive integer greater than 1")
 
-    params = {}
-
-    params["4"] = np.linspace(0.1, 0.9, 4)
-    params["5"] = np.linspace(0.1, 0.9, 5)
-    params["6"] = np.linspace(0.1, 0.9, 6)
-    params["7"] = np.linspace(0.1, 0.9, 7)
-    params["8"] = np.linspace(0.1, 0.9, 8)
-    params["10"] = np.linspace(0.1, 0.9, 10)
-    params["12"] = np.linspace(0.1, 0.9, 12)
-    params["14"] = np.linspace(0.1, 0.9, 14)
+    params = np.linspace(0.1, 0.9, J)
 
     probs = []
 
     for true_class in range(0, J):
-        probs.append(binom.pmf(np.arange(0, J), J - 1, params[str(J)][true_class]))
+        probs.append(binom.pmf(np.arange(0, J), J - 1, params[true_class]))
 
     return np.array(probs)

dlordinal/soft_labelling/tests/test_binomial_distribution.py

@@ -28,3 +28,33 @@ def test_get_binomial_soft_labels():
 
     # Individual probabilities should be within [0, 1]
     assert np.all(result >= 0) and np.all(result <= 1)
+
+
+def test_get_binomial_soft_labels_invalid_input():
+    with pytest.raises(ValueError):
+        get_binomial_soft_labels(1)
+
+    with pytest.raises(ValueError):
+        get_binomial_soft_labels(1.0)
+
+    with pytest.raises(ValueError):
+        get_binomial_soft_labels(0)
+
+    with pytest.raises(ValueError):
+        get_binomial_soft_labels(-1)
+
+
+def test_get_binomial_soft_labels_valid_input():
+    for i in range(2, 11):
+        soft_labels = get_binomial_soft_labels(i)
+
+        # Sum of probabilities in each row should be approximately 1
+        row_sums = np.sum(soft_labels, axis=1)
+        for row_sum in row_sums:
+            assert row_sum == pytest.approx(1.0, abs=1e-6)
+
+        # Check that all the elements in the matrix are less than or equal to the
+        # element in the diagonal
+        diagonal = np.diag(soft_labels)
+        diff = soft_labels - diagonal[:, np.newaxis]
+        assert np.all(diff <= 1e-9)