Abstract

We exploit the similarity between irregular Cellular Automata (CA) and Geometric Proportional Analogies (GPA), as both involve manipulations of geometric objects (points, lines and polygons). We describe how each GPA effectively defines a CA-like transition rule and we adapt an algorithm (called Structure Matching) used for solving GPAs to solving CAs. Irregular CAs improve on regular CAs by allowing an irregular tessellation of the plane, while further extensions support transition rules that lie beyond the scope of traditional CA. We describe three facets of the resulting model; layered inferences, incremental structures and the merge operation. Examples describe how structure matching (Mullally et al, 2005) is used to update and enhance a topographic land-cover map.