Adaptation Noise And Self Organizing Systems

We consider the population dynamics of a set of species whose network of catalytic interactions is described by a directed graph. The relationship between the attractors of this dynamics and the underlying graph theoretic structures like cycles and autocatalytic sets is discussed. It is shown that when the population dynamics is suitably coupled to a slow dynamics of the graph itself, the network evolves towards increasing complexity driven by autocatalytic sets. Some quantitative measures of network complexity are described.

In genetic systems there is a non-trivial interface between the sequence of symbols which constitutes the chromosome, or ``genotype'', and the products which this sequence encodes --- the ``phenotype''. This interface can be thought of as a ``computer''. In this case the chromosome is viewed as an algorithm and the phenotype as the result of the computation. In general only a small fraction of all possible sequences of symbols makes any sense for a given computer. The difficulty of finding meaningful algorithms by random mutation is known as the brittleness problem. In this paper we show that mutation and crossover favour the emergence of an algorithmic language which facilitates the production of meaningful sequences following random mutations of the genotype. We base our conclusions on an analysis of the population dynamics of a variant of Kitano's neurogenetic model wherein the chromosome encodes the rules for cellular division and the phenotype is a 16-cell organism interpreted as a connectivity matrix for a feedforward neural network. We show that an algorithmic language emerges, describe this language in extenso, and show how it helps to solve the brittleness problem.

A binary game is introduced and analysed. N players have to choose one of the two sides independently and those on the minority side win. Players uses a finite set of ad hoc strategies to make their decision, based on the past record. The analysing power is limited and can adapt when necessary. Interesting cooperation and competition pattern of the society seem to arise and to be responsive to the payoff function.

A dynamic model for cell differentiation is studied, where cells with internal chemical reaction dynamics interact with each other and replicate. It leads to spontaneous differentiation of cells and determination, as is discussed in the isologous diversification. Following features of the differentiation are obtained: (1)Hierarchical differentiation from a ``stem'' cell to other cell types, with the emergence of the interaction-dependent rules for differentiation; (2)Global stability of an ensemble of cells consisting of several cell types, that is sustained by the emergent, autonomous control on the rate of differentiation; (3)Existence of several cell colonies with different cell-type distributions. The results provide a novel viewpoint on the origin of complex cell society, while relevance to some biological problems, especially to the hemopoietic system, is also discussed.

We study a single-lane traffic model that is based on human driving behavior. The outflow from a traffic jam self-organizes to a critical state of maximum throughput. Small perturbations of the outflow far downstream create emergent traffic jams with a power law distribution P(t)∼ t −3/2 of lifetimes, t . On varying the vehicle density in a closed system, this critical state separates lamellar and jammed regimes, and exhibits 1/f noise in the power spectrum. Using random walk arguments, in conjunction with a cascade equation, we develop a phenomenological theory that predicts the critical exponents for this transition and explains the self-organizing behavior. These predictions are consistent with all of our numerical results.

We study the spatial pattern formation and emerging long range correlations in a model of three species coevolving in space and time according to stochastic contact rules. Analytical results for the pair correlation functions, based on a truncation approximation and supported by computer simulations, reveal emergent strategies of survival for minority agents based on selection of patterns. Minority agents exhibit defensive clustering and cooperative behavior close to phase transitions.

We investigate the dynamics of a 2-level atom (or spin-1/2) coupled to a mass-less bosonic field at positive temperature. We prove that, at small coupling, the combined quantum system approaches thermal equilibrium. Moreover we establish that this approach is exponentially fast in time. We first reduce the question to a spectral problem for the Liouvillean, a self-adjoint operator naturally associated with the system. To compute this operator, we invoke Tomita-Takesaki theory. Once this is done we use complex deformation techniques to study its spectrum. The corresponding zero temperature model is also reviewed and compared.

Macroevolution is considered as a problem of stochastic dynamics in a system with many competing agents. Evolutionary events (speciations and extinctions) are triggered by fitness records found by random exploration of the agents' fitness landscapes. As a consequence, the average fitness in the system increases logarithmically with time, while the rate of extinction steadily decreases. This dynamics is studied by numerical simulations and, in a simpler mean field version, analytically. We also study the effect of externally added `mass' extinctions. The predictions for various quantities of paleontological interest (life-time distributions, distribution of event sizes and behavior of the rate of extinction) are robust and in good agreement with available data. Brief version of parts of this work have been published as Letters. (PRL 75, 2055, (1995) and PRL, 79, 1413, (1997))

We study populations of agents evolving in fitness landscapes constructed according to the rules of a modified NK model with a tunable amount of neutral paths. In the `punctuated equilibrium' regime evolutionary events are identified as jumps in the mean population fitness, and the statistics of these jumps in an ensemble of independently evolving populations is analyzed. We show that, for a wide range of landscapes parameters, the number of events in time t is Poisson distributed, with the time parameter replaced by the logarithm of time. This simple log-Poisson statistics likewise describes the number n of records in any sequence of t independently generated random numbers. The implications of this behavior for evolution dynamics are discussed.

In this paper we study the minority game in the presence of evolution. In particular, we examine the behavior in games in which the dimension of the strategy space, m, is the same for all agents and fixed for all time. We find that for all values of m, not too large, evolution results in a substantial improvement in overall system performance. We also show that after evolution, results obey a scaling relation among games played with different values of m and different numbers of agents, analogous to that found in the non-evolutionary, adaptive games. Best system performance still occurs, for a given number of agents, at m_c, the same value of the dimension of the strategy space as in the non-evolutionary case, but system performance is now nearly an order of magnitude better than the non- evolutionary result. For m<m_c, the system evolves to states in which average agent wealth is better than in the random choice game,despite (and in some sense because of) the persistence of maladaptive behavior by some agents. As m gets large, overall systems performance approaches that of the random choice game.