Information integration in elementary cellular automata
Authors:
Kátia K. Cassiano,
Valmir C. Barbosa
Abstract:
We study the emergence of information integration in cellular automata (CA) with respect to states in the long run. Information integration is in this case quantified by applying the information-theoretic measure known as total correlation to the long-run distribution of CA states. Total correlation is the amount by which the total uncertainty associated with cell states surpasses the uncertainty…
▽ More
We study the emergence of information integration in cellular automata (CA) with respect to states in the long run. Information integration is in this case quantified by applying the information-theoretic measure known as total correlation to the long-run distribution of CA states. Total correlation is the amount by which the total uncertainty associated with cell states surpasses the uncertainty of the CA state taken as a whole. It is an emergent property, in the sense that it can only be ascribed to how the cells interact with one another, and has been linked to the rise of consciousness in the brain. We investigate total correlation in the evolution of elementary CA for all update rules that are unique with respect to negation or reflection. For each rule we consider the usual, deterministic CA behavior, assuming that the initial state is chosen uniformly at random, and also the probabilistic variant in which every cell, at all time steps and independently of all others, disobeys the rule's prescription with a fixed probability. We have found rules that generate as much total correlation as possible, or nearly so, particularly in Wolfram classes 2 and 3. We conjecture that some of these rules can be used as CA models of information integration.
△ Less
Submitted 28 October, 2014;
originally announced October 2014.
Error-prone cellular automata as metaphors of immunity as computation
Authors:
Kátia K. Cassiano,
Valmir C. Barbosa
Abstract:
Modeling the immune system so that its essential functionalities stand out without the need for every molecular or cellular interaction to be taken into account has been challenging for many decades. Two competing approaches have been the clonal selection theory and the idiotypic-network theory, each stemming from a relatively separate set of principles. One recent perspective holding the promise…
▽ More
Modeling the immune system so that its essential functionalities stand out without the need for every molecular or cellular interaction to be taken into account has been challenging for many decades. Two competing approaches have been the clonal selection theory and the idiotypic-network theory, each stemming from a relatively separate set of principles. One recent perspective holding the promise of unification is that of immunity as computation, that is, of immunity as the process of computing the state of the body so that protection can be effected, as well as boosted through learning. Here we investigate the use of cellular automata (CA) as the core abstraction supporting this new perspective. Our choice of CA for this role is based on the potential variety of basins in a CA attractor field. Associating each basin with a consistent set of body states, and moreover providing for the noisy evolution of the CA in time so that jumping between basins is possible, have provided the necessary backdrop. Given a CA rule to be followed by all cells synchronously, our model is based on a probability with which each cell, at each time step, independently updates its own state differently than the rule mandates. Setting up and solving the corresponding Markov chain for its stationary probabilities have revealed that already in the context of elementary CA there exist rules that, while allowing transitions between basins, display remarkable resiliency in terms of basin occupation. For these rules, the long-run probability that the CA is found in a given basin is practically the same as in the deterministic case when the initial CA state is chosen uniformly at random. We argue that, consequently, our single-parameter CA model may be a suitable abstraction of immunity as computation.
△ Less
Submitted 14 January, 2014;
originally announced January 2014.
