updates

utils/clustering_utils.py

@@ -45,12 +45,16 @@ def _calculate_silhouette_score(X: np.ndarray, labels: np.ndarray, metric: str)
     unique_labels_set = set(labels)
     n_clusters_ = len(unique_labels_set) - (1 if -1 in unique_labels_set else 0)
 
+    # The silhouette score is only defined if there is more than 1 cluster.
+    # Outliers (label -1) are excluded from the score calculation.
     if n_clusters_ > 1:
+        # Create a mask to select only points that are part of a cluster (not noise)
         clustered_mask = (labels != -1)
         if np.sum(clustered_mask) > 1:
             X_clustered = X[clustered_mask]
             labels_clustered = labels[clustered_mask]
             try:
+                # Compute the score on the non-outlier points
                 return silhouette_score(X_clustered, labels_clustered, metric=metric)
             except ValueError:
                 return None