This work was supported by the National Science Foundation under the grant EET 87-21333.


The associative memory of a stack filter is defined to be the set of root signals of that filter. If the root sets of two stack filters both contain a desired set of patterns, but one filter’s root set is smaller than the other, then the filter with the smaller root set is said to be better for that set of patterns. Any filter which has the smallest number of roots containing the specified set of patterns is said to be a best filter. The configuration of the family of best filters is described via a graphical approach which specifies an upper and lower bound for the subset of possible best filters which are furthest from the sets of type-1 and type-2 stack filters. Knowledge of this configuration leads to an algorithm which can produce a near-best filter. This new method of constructing associative memories does not require the desired set of patterns to be independent and it can construct a much better filter than the methods in [I].

Date of this Version