Better Handling of zero median values in Kernel Width (#160)

* Filter zeros out of median computation
* removing IDE files from commit

---------

Signed-off-by: Nicholas Parente <parentenickj@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: nparent1

.gitignore

@@ -123,6 +123,11 @@ venv.bak/
 .spyderproject
 .spyproject
 
+# PyCharm
+.idea/
+*.iml
+*.iws
+
 # Rope project settings
 .ropeproject
 

doc/whats_new/v0.1.rst

@@ -50,6 +50,7 @@ Changelog
 - |Feature| Add a suite of general categorical data CI tests, by `Adam Li`_ (:pr:`128`)
 - |Feature| Implement CAM, SCORE, DAS, NoGAM algorithms in ``dodiscover.toporder`` submodule (:pr:`129`)
 - |Feature| Add Psi-FCI and I-FCI algorithm for handling soft-interventional data, :class:`dodiscover.constraint.PsiFCI` by `Adam Li`_ (:pr:`111`)
+- |Fix| Update the kernel_width method to filter out zeros before computing the median pairwise distance,  by `Nick Parente`_ (:pr:`160`)
 
 Code and Documentation Contributors
 -----------------------------------

dodiscover/toporder/utils.py

@@ -19,8 +19,9 @@ def kernel_width(X: NDArray):
         Matrix of the data.
     """
     X_diff = np.expand_dims(X, axis=1) - X  # Gram matrix of the data
-    D = np.linalg.norm(X_diff, axis=2)
-    s = np.median(D.flatten())
+    D = np.linalg.norm(X_diff, axis=2).flatten()
+    D_nonzeros = D[D > 0]  # Remove zeros
+    s = np.median(D_nonzeros) if np.any(D_nonzeros) else 1
     return s
 
 

tests/unit_tests/toporder/test_utils.py

@@ -0,0 +1,14 @@
+import numpy as np
+
+from dodiscover.toporder.utils import kernel_width
+
+
+def test_kernel_width_when_zero_median_pairwise_distances():
+    arr = np.zeros((100, 1), dtype=np.int64)
+    arr[1] = 1
+    assert kernel_width(arr) == 1
+
+
+def test_kernel_width_when_all_zero_pairwise_distances():
+    arr = np.ones((100, 1), dtype=np.int64)
+    assert kernel_width(arr) == 1