The entire score program should be checked to ensure that it outputs all expected scores, that is, those which should not be pruned due to the standard (e.g., Teyssier and Koller, UAI 2005) pruning rules.
Calculating more than a few scores by hand is not practical. These tests can be implemented by calculating all scores using a straightforward calculation and then applying the simple subset checks. The iris dataset is small enough for this strategy (i.e., calculating all scores without Suzuki/de Campos-style pruning).
The entire score program should be checked to ensure that it outputs all expected scores, that is, those which should not be pruned due to the standard (e.g., Teyssier and Koller, UAI 2005) pruning rules.
Calculating more than a few scores by hand is not practical. These tests can be implemented by calculating all scores using a straightforward calculation and then applying the simple subset checks. The iris dataset is small enough for this strategy (i.e., calculating all scores without Suzuki/de Campos-style pruning).