Characterization of one-dimensional cellular automata rules through topological network features

The paper investigates the relationship between the classification schemes, defined by Wolfram and Gilman, of onedimensional cellular automata through concepts coming from network theory. An automaton is represented with a network, generated from the elementary rule defining its behavior. Characteristic features of this graph are computed and machine learning classification models are built. Such models allow to classify automaton rules and to compare Wolfram's and Gilman's classes by comparing the classes predicted by these models.