Several previously introduced classiers share similarities with CDW and its
variants. For example VDM computes the feature
value distributions, but unlike CDW the weights it computes do not depend on
the class of the training instance. VDM computes dierent distances between
symbolic feature values, and also weights the features based on the feature value
of the test case. This can be viewed as weighting features locally based on both
the training and test cases, although the computed distance between any two
given feature values is the same across the entire data set. Of the methods
presented here, VDM is most similar to GMEF-CDW