Handle invalid inputs for mutual information computation where mathematically undefined values (e.g. LOG(0)) are returned.

PiperOrigin-RevId: 454202540

Author: tf-transform-team

tensorflow_transform/info_theory.py

@@ -84,7 +84,7 @@ def calculate_partial_mutual_information(n_ij, x_i, y_j, n):
   Returns:
     Mutual information for the cell x=i, y=j.
   """
-  if n_ij == 0:
+  if n_ij == 0 or x_i == 0 or y_j == 0:
     return 0
   return n_ij * ((log2(n_ij) + log2(n)) -
                  (log2(x_i) + log2(y_j)))

tensorflow_transform/info_theory_test.py

@@ -137,6 +137,27 @@ def test_calculate_partial_expected_mutual_information(
           col_count=8,
           total_count=16,
           expected_mi=0),
+      dict(
+          testcase_name='invalid_input_zero_cell_count',
+          cell_count=4,
+          row_count=0,
+          col_count=8,
+          total_count=8,
+          expected_mi=0),
+      dict(
+          testcase_name='invalid_input_zero_row_count',
+          cell_count=4,
+          row_count=0,
+          col_count=8,
+          total_count=8,
+          expected_mi=0),
+      dict(
+          testcase_name='invalid_input_zero_col_count',
+          cell_count=4,
+          row_count=8,
+          col_count=0,
+          total_count=8,
+          expected_mi=0),
   )
   def test_mutual_information(self, cell_count, row_count, col_count,
                               total_count, expected_mi):