Bug fix for ProxySPEX, removed external library label in tests (#484)

Author: landonbutler

src/shapiq/approximator/sparse/base.py

@@ -482,6 +482,8 @@ def _refine(
         """
         n = train_X.shape[1]
         four_items = list(four_dict.items())
+        if len(four_items) <= self.n:
+            return four_dict
         list_keys = [item[0] for item in four_items]
         four_coefs = np.array([item[1] for item in four_items])
 
@@ -492,6 +494,8 @@ def _refine(
             four_coefs_for_energy[nfc_idx] = 0
         four_coefs_sq = four_coefs_for_energy**2
         tot_energy = np.sum(four_coefs_sq)
+        if tot_energy == 0:
+            return four_dict
         sorted_four_coefs_sq = np.sort(four_coefs_sq)[::-1]
         cumulative_energy_ratio = np.cumsum(sorted_four_coefs_sq / tot_energy)
         thresh_idx_95 = np.argmin(cumulative_energy_ratio < 0.95) + 1

tests/shapiq/tests_unit/tests_approximators/test_approximator_base_sparse.py

@@ -225,7 +225,6 @@ def test_approximate_with_spex(
 
 
 @skip_if_no_lightgbm
-@pytest.mark.external_libraries
 def test_approximate_with_proxyspex():
     """Tests the approximate method with the proxyspex decoder."""
     n = 10
@@ -263,3 +262,48 @@ def test_sparse_init_succeeds_lightgbm(monkeypatch):
 
     with pytest.raises(ImportError, match="The 'lightgbm' package is required"):
         Sparse(n=10, index="SII", decoder_type="proxyspex")
+
+
+def test_refine_zero_total_energy():
+    """
+    Tests that the _refine method handles the case when total energy is zero.
+
+    This test ensures that when all Fourier coefficients (excluding the baseline)
+    are zero, the method returns the original dictionary without attempting to
+    divide by zero.
+    """
+    import numpy as np
+
+    n = 5
+    # Use "soft" decoder which doesn't require lightgbm
+    approximator = Sparse(n=n, index="FBII", max_order=2, decoder_type="soft", random_state=42)
+
+    # Create a four_dict where all non-baseline coefficients are zero
+    four_dict = {
+        (): 1.0,  # baseline coefficient
+        (0,): 0.0,
+        (1,): 0.0,
+        (0, 1): 0.0,
+        (2,): 0.0,
+        (1, 2): 0.0,
+    }
+
+    # Create some dummy training data
+    train_X = np.array(
+        [
+            [0, 0, 0, 0, 0],
+            [1, 0, 0, 0, 0],
+            [0, 1, 0, 0, 0],
+            [1, 1, 0, 0, 0],
+            [0, 0, 1, 0, 0],
+            [1, 0, 1, 0, 0],
+        ]
+    )
+    train_y = np.array([1.0, 1.0, 1.0, 1.0, 1.0, 1.0])
+
+    # Call _refine - should not raise a ZeroDivisionError
+    result = approximator._refine(four_dict, train_X, train_y)
+
+    # The result should be the same as the input when total energy is zero
+    assert result == four_dict
+    assert len(result) == len(four_dict)

tests/shapiq/tests_unit/tests_approximators/test_approximator_proxyspex.py

@@ -12,7 +12,6 @@
 
 
 @skip_if_no_lightgbm
-@pytest.mark.external_libraries
 def test_initialization_defaults():
     """Test that ProxySPEX initializes with correct defaults."""
     n = 10
@@ -36,7 +35,6 @@ def test_initialization_defaults():
     ],
 )
 @skip_if_no_lightgbm
-@pytest.mark.external_libraries
 def test_initialization_custom(n, index, max_order, top_order):
     """Test ProxySPEX initialization with custom parameters."""
     proxyspex = ProxySPEX(
@@ -61,7 +59,6 @@ def test_initialization_custom(n, index, max_order, top_order):
     ],
 )
 @skip_if_no_lightgbm
-@pytest.mark.external_libraries
 def test_approximate(n, interactions, budget):
     """Test ProxySPEX approximation functionality."""