On the other hand, any two input vectors that are dissimilar (i.e., far apart in input space) will select a highly disjoint subset of locations in the A* set. Most often, there will be no common memory locations in A*. Thus, the output response of a CMAC to dissimilar input vectors can be independent, making it easy for CMAC to distinguish between dissimilar input vectors. This means that CMAC can classify or recognize input
patterns. The sparse nature of the A vector makes it possible for CMAC to learn to classify a large number of patterns.