A new model for improving symbolic product machine traversal

Abstract

The authors present algorithms for traversing product machines which improve on the results of H. Cho et al. (1991), with a speedup ranging from 3 up to 6. New features include a model that generalizes the product machine, resulting in simpler and more efficient representations and computations, as well as optimizations in symbolic image computation. In the latter case, the speedup ranged from 1.5 to 4.<>