PROBABILISTIC BEHAVIOUR/STRUCTURE TRANSFORMATIONS USING TRANSITIVEMOORE MODELS

Abstract

The ATOM procedure for probabilistic behaviour/structure transformation developed by Gaines 1,2 operates by enumerating and evaluating all non-deterministic automata of a given size with respect to a given behaviour. It is shown that a re-ordering of the model space can usefully be effected by considering the class of transitive Moore models, leading to a much more efficient method of finding simple structure when it exists in the behaviour. Searching transitive Moore models is tantamount to searching Mealy models in order of number of transitions rather than state count, and results from a computer implementation are presented.