A study on learning robustness using asynchronous 1D cellular automata rules

Numerous studies can be found in literature concerning the idea of learning cellular automata (CA) rules that perform a given task by means of machine learning methods. Among these methods, genetic algorithms (GAs) have often been used with excellent results. Nevertheless, few attention has been dedicated so far to the generality and robustness of the learned rules. In this paper, we show that when GAs are used to evolve asynchronous one-dimensional CA rules, they are able to find more general and robust solutions compared to the more usual case of evolving synchronous CA rules.     

Learning cellular automata rules for binary classification problem

This paper proposes a cellular automata-based solution of a binary classification problem. The proposed method is based on a two-dimensional, three-state cellular automaton (CA) with the von Neumann neighborhood. Since the number of possible CA rules (potential CA-based classifiers) is huge, searching efficient rules is conducted with use of a genetic algorithm (GA). Experiments show an excellent performance of discovered rules in solving the classification problem. The best found rules perform better than the heuristic CA rule designed by a human and also better than one of the most widely used statistical method: the k-nearest neighbors algorithm (k-NN). Experiments show that CAs rules can be successfully reused in the process of searching new rules.     

Alternation on cellular automata

In this paper we consider several notions of alternation in cellular automata: non-uniform, uniform and weak alternation. We study relations among these notions and with alternating Turing machines. It is proved that the languages accepted in polynomial time by alternating Turing machines are those accepted by alternating cellular automata in polynomial time for all the proposed alternating cellular automata. In particular, this is true for the weak model where the difference between existential and universal states is omitted for all the cells except the first one. It is proved that real time alternation in cellular automata is strictly more powerful than real time alternation in Turing machines, with only one read-write tape. Moreover, it is shown that in linear time uniform and weak models agree.     

On real-time cellular automata and trellis automata

Summary It is shown that f(n)-time one-way cellular automata are equivalent to f(n)-time trellis automata, the real-time one-way cellular automata languages are closed under reversal, the 2n-time one-way cellular automata are equivalent to real-time cellular automata and the latter are strictly more powerful than the real-time one-way cellular automata.This work has been done during the second author's visit at the University of Paris and during both authors' visit at the Institute für Informationsverarbeitung Graz, Austria     

Tree automata with one memory set constraints and cryptographic protocols

We introduce a class of tree automata that perform tests on a memory that is updated using function symbol application and projection. The language emptiness problem for this class of tree automata is shown to be in DEXPTIME.We also introduce a class of set constraints with equality tests and prove its decidability by completion techniques and a reduction to tree automata with one memory.Finally, we show how to apply these results to cryptographic protocols. We introduce a class of cryptographic protocols and show the decidability of secrecy for an arbitrary number of agents and an arbitrary number of (concurrent or successive) sessions, provided that only a bounded number of new data is generated. The hypothesis on the protocol (a restricted copying ability) is shown to be necessary: without this hypothesis, we prove that secrecy is undecidable, even for protocols without nonces.     

Computations on one-dimensional cellular automata

Cellular automata may be viewed as a modelization of synchronous parallel computation. Even in the one-dimensional case, they are known as capable of universal computations. The usual proof uses a simulation of a universal Turing machine. In this paper, we present how a one-dimensional cellular automata can simulate any recursive function in such a way that composition of computations occurs as soon as possible. In addition, this allows us to show that one-dimensional cellular automata may simulate asynchronous computations.This work was supported by the Programme de Recherches Coordonnées Mathématiques et Informatique and the Esprit Basic Research Action Algebraic and Syntactical Methods in Computer Science.     

A verifiable multi-secret sharing scheme based on cellular automata

In this paper, we propose a verifiable (t, n)-threshold multi-secret sharing scheme, based on one-dimensional cellular automata where the number of secrets is not restricted by n or t. We show that our scheme can be used to solve an open problem proposed recently in Alvarez et al. (2008) [G. Alvarez, L. Hernández Encinas, A. Martín del Rey, A multisecret sharing scheme for color images based on cellular automata, Information Sciences 178 (2008) 4382-4395].     

A cellular automata model for edge relaxation

An edge detection method based on a fuzzy cellular automata model which serves as the relaxation labeling process constraint is described. An initial estimate of edge locations is made and the remaining ambiguities are resolved by thinning and enhancing the edges through several iterations. An efficient fixed step algorithm is presented and its performance is evaluated for different noise level images. The method is useful for the detection of linear image features in three-dimensional robot vision systems.     

Sequential and cellular graph automata

Labeled graphs of bounded degree, with numbers assigned to the arcs at each node, are called d-graphs. Sequential and cellular d-graph automata are defined, and it is shown that they can simulate each other and are also equivalent to web-bounded automata.     

Non-deterministic cellular automata and languages

We summarize results on non-deterministic cellular language acceptors. The non-determinism is regarded as limited resource. For parallel devices, it is natural to bound the non-determinism in time and/or space. Depending on the length of the input, the number of allowed non-deterministic state transitions as well as the number of non-deterministic cells at all is limited. We centre our attention to real-time, linear-time, and unrestricted-time computations and discuss the computational power of these machines. Speed-up results and the possibility to reduce the non-determinism as well as closure properties of languages acceptable with a constant number of non-deterministic transitions are presented. By considering the relations with context-free languages, several relations between the devices in question are implied. We do not prove these results but we merely draw attention to the big picture and some of the main ideas involved, and open problems for further research.     

Fluctuation-driven computing on number-conserving cellular automata

A number-conserving cellular automaton (NCCA) is a cellular automaton in which the states of cells are denoted by integers, and the sum of all of the numbers in a configuration is conserved throughout its evolution. NCCAs have been widely used to model physical systems that are ruled by conservation laws of mass or energy. Imai et al. [13] showed that the local transition function of NCCA can be effectively translated into the sum of a binary flow function over pairs of neighboring cells. In this paper, we explore the computability of NCCAs in which the pairwise number flows are performed at fully asynchronous timings. Despite the randomness that is associated with asynchronous transitions, useful computation still can be accomplished efficiently in the cellular automata through the active exploitation of fluctuations [18]. Specifically, certain numbers may flow randomly fluctuating between forward and backward directions in the cellular space, as if they were subject to Brownian motion. Because random fluctuations promise a powerful resource for searching through a computational state space, the Brownian-like flow of the numbers allows for efficient embedding of logic circuits into our novel asynchronous NCCA.     

“Real-time” equivalence of cellular automata and linear-bounded automata

It is shown that a one-dimensional bounded cellular automaton (BCA) can simulate a linear-bounded automaton (LBA) in essentially real time, and that, conversely, an LBA can simulate a BCA, on an input of length n, in just n transitions of the LBA per BCA transition.     

A cellular automata model of an anticipatory system

An anticipatory system has been modeled using the dynamic characteristics of cellular automata. Rules governing the steps in an enzymatic conversion of substrates to products are operative in the system. A concentration of an intermediate product influences the creation of a supplemental enzyme that enhances the competence of an enzyme down stream. This anticipation of the future event creates a condition in which the concentration of a later substrate is suppressed, a property characteristic of the system. The model presents a useful opportunity to study a variety of aspects of this fascinating phenomena.     

A simple framework for real-time cryptographic protocol analysis with compositional proof rules

A real-time process algebra, enhanced with specific constructs for handling cryptographic primitives, is proposed to model cryptographic protocols in a simple way. We show that some security properties, such as authentication and secrecy, can be re-formulated in this timed setting. Moreover, we show that they can be seen as suitable instances of a general information flow-like scheme, called timed generalized non-deducibility on compositions (tGNDC), parametric w.r.t. the observational semantics of interest. We show that, when considering timed trace semantics, there exists a most powerful hostile environment (or enemy) that can try to compromise the protocol. Moreover, we present a couple of compositionality results for tGNDC, one of which is time dependent, and show their usefulness by means of a case study.     

Grids and universal computations on one-dimensional cellular automata

This article shows how universal computations can be achieved on one-dimensional cellular automata. We are interested in intrinsic universality: we want a CA in which any other CA can be represented and simulated with no intermediate coding relevant to another computation model. We first abstract the space-time diagram in favor of the dependency graph. Then we show how such a dependency graph (via treillis automata) can be realized by what is called a grid, leading to a simple uniform simulation. Finally, we exhibit a very simple universal brick that can be used in grids to obtain an intrinsic universal CA.     

Identification of probabilistic cellular automata

The identification of probabilistic cellular automata (PCA) is studied using a new two stage neighborhood detection algorithm. It is shown that a binary probabilistic cellular automaton (BPCA) can be described by an integer-parameterized polynomial corrupted by noise. Searching for the correct neighborhood of a BPCA is then equivalent to selecting the correct terms which constitute the polynomial model of the BPCA, from a large initial term set. It is proved that the contribution values for the correct terms can be calculated independently of the contribution values for the noise terms. This allows the neighborhood detection technique developed for deterministic rules in to be applied with a larger cutoff value to discard the majority of spurious terms and to produce an initial presearch for the BPCA neighborhood. A multiobjective genetic algorithm (GA) search with integer constraints is then evolved to refine the reduced neighborhood and to identify the polynomial rule which is equivalent to the probabilistic rule with the largest probability. A probability table representing the BPCA can then be determined based on the identified neighborhood and the deterministic rule. The new algorithm is tested over a large set of one-dimensional (1D), two-dimensional (2D), and three-dimensional (3D) BPCA rules. Simulation results demonstrate the efficiency of the new method.     

Computational classification of cellular automata

We discuss attempts at the classification of cellular automata, in particular with a view towards decidability. We will see that a large variety of properties relating to the short-term evolution of configurations are decidable in principle, but questions relating to the long-term evolution are typically undecidable. Even in the decidable case, computational hardness poses a major obstacle for the automatic analysis of cellular automata.     

Entropy of multidimensional cellular automata

Since the topological entropy of a vast class of two-dimensional cellular automata (CA) is infinite, of interest is the possibility to renormalize it so that to obtain a positive finite value. We find the asymptotics of the information function of a multidimensional CA and, accordingly, introduce the renormalized topological entropy as a coefficient of this asymptotics. We describe some properties of the introduced quantity, in particular, its positivity for CA of the type of “The Game of Life.” Also, we give an example of an explicit evaluation of this parameter for a particular cellular automaton.