An Information-theoretic Classification of Complex Systems

Using elementary cellular automata as an example, a novel, information-based classification of complex systems is proposed that circumvents the problems associated with isolating the complexity generated as a product of an initial state from that which is intrinsic to a dynamical rule. Transfer entropy variations processed by the system for different initial states split the 256 elementary rules into three information classes. These classes form a hierarchy such that coarse-graining transitions permitted among automata rules predominately occur within each information-based class, or much more rarely down the hierarchy.