Evaluating Stability and Comparing Output of Feature Selectors that Optimize Feature Subset Cardinality