where is the weight assigned to the binary expanded feature for
the class of the training case In other words, is the sum of all the
value weights minus the value weights for the training and test case
if their values dier. This method, which we call expanded feature CDW (EFCDW)
has ner locality than the standard CDW since it varies with individual
feature values. It should perform best on tasks in which some but not all values
of a feature are important, and where the importance of each value varies from
class to class.