Adaptation Noise And Self Organizing Systems

There are 512 two-locus, two-allele, two-phenotype, fully-penetrant disease models. Using the permutation between two alleles, between two loci, and between being affected and unaffected, one model can be considered to be equivalent to another model under the corresponding permutation. These permutations greatly reduce the number of two-locus models in the analysis of complex diseases. This paper determines the number of non-redundant two-locus models (which can be 102, 100, 96, 51, 50, or 48, depending on which permutations are used, and depending on whether zero-locus and single-locus models are excluded). Whenever possible, these non-redundant two-locus models are classified by their property. Besides the familiar features of multiplicative models (logical AND), heterogeneity models (logical OR), and threshold models, new classifications are added or expanded: modifying-effect models, logical XOR models, interference and negative interference models (neither dominant nor recessive), conditionally dominant/recessive models, missing lethal genotype models, and highly symmetric models. The following aspects of two-locus models are studied: the marginal penetrance tables at both loci, the expected joint identity-by-descent probabilities, and the correlation between marginal identity-by-descent probabilities at the two loci. These studies are useful for linkage analyses using single-locus models while the underlying disease model is two-locus, and for correlation analyses using the linkage signals at different locations obtained by a single-locus model.

Random boolean cellular automata are investigated, where each gate has two randomly chosen inputs and is randomly assigned a boolean function of its inputs. The effect of non-uniform distributions on the choice of the boolean functions is considered. The main results are that if the gates are more likely to be assigned constant functions than non-canalyzing functions, then with very high probability, the automaton will exhibit very stable behavior: most of the gates will stabilize, and the state cycles will be bounded in size.

Based on the law of mass action (and its microscopic foundation) and mass conservation, we present here a method to derive consistent dynamic models for the time evolution of systems with an arbitrary number of species. Equations are derived through a mechanistic description, ensuring that all parameters have ecological meaning. After discussing the biological mechanisms associated to the logistic and Lotka-Volterra equations, we show how to derive general models for trophic chains, including the effects of internal states at fast time scales. We show that conformity with the mass action law leads to different functional forms for the Lotka-Volterra and trophic chain models. We use mass conservation to recover the concept of carrying capacity for an arbitrary food chain.

I define a natural measure of the complexity of a parametric distribution relative to a given true distribution called the {\it razor} of a model family. The Minimum Description Length principle (MDL) and Bayesian inference are shown to give empirical approximations of the razor via an analysis that significantly extends existing results on the asymptotics of Bayesian model selection. I treat parametric families as manifolds embedded in the space of distributions and derive a canonical metric and a measure on the parameter manifold by appealing to the classical theory of hypothesis testing. I find that the Fisher information is the natural measure of distance, and give a novel justification for a choice of Jeffreys prior for Bayesian inference. The results of this paper suggest corrections to MDL that can be important for model selection with a small amount of data. These corrections are interpreted as natural measures of the simplicity of a model family. I show that in a certain sense the logarithm of the Bayesian posterior converges to the logarithm of the {\it razor} of a model family as defined here. Close connections with known results on density estimation and ``information geometry'' are discussed as they arise.

When a small number of individuals of organism of single species is confined in a closed space with limited amount of indispensable resources, their breading may start initially under suitable conditions, and after peaking, the population should go extinct as the resources are exhausted. Starting with the logistic equation and assuming that the carrying capacity of the environment is a function of the amount of resources, a mathematical model describing such pattern of population change is obtained. An application of this model to typical population records, that of deer herds by Scheffer (1951) and O'Roke and Hamerstrome (1948), yields estimations of the initial amount of indispensable food and its availability or nutritional efficiency which were previously unspecified.

We propose a network of oscillators to retrieve given patterns in which the oscillators keep a fixed phase relationship with one another. In this description, the phase and the amplitude of the oscillators can be regarded as the timing and the strength of the neuronal spikes, respectively. Using the amplitudes for encoding, we enable the network to realize not only oscillatory states but also non-firing states. In addition, it is shown that under suitable conditions the system has a Lyapunov function ensuring a stable retrieval process. Finally, the associative memory capability of the network is demonstrated numerically.

We study the nature and mechanisms of broken ergodicity (BE) in specific random walk models corresponding to diffusion on random potential surfaces, in both one and high dimension. Using both rigorous results and nonrigorous methods, we confirm several aspects of the standard BE picture and show that others apply in one dimension, but need to be modified in higher dimensions. These latter aspects include the notions that at fixed temperature confining barriers increase logarithmically with time, that ``components'' are necessarily bounded regions of state space which depend on the observational timescale, and that the system continually revisits previously traversed regions of state space. We examine our results in the context of several experiments, and discuss some implications of our results for the dynamics of disordered and/or complex systems.

A phenomenological model of human posture control is posited. The dynamics are modelled as an elastically pinned polymer under the influence of noise. The model accurately reproduces the two-point correlation functions of experimental posture data and makes predictions for the response function of the postural control system. The physiological and clinical significance of the model is discussed.

We propose a learning algorithm for solving the traveling salesman problem based on a simple strategy of trial and adaptation: i) A tour is selected by choosing cities probabilistically according to the ``synaptic'' strengths between cities. ii) The ``synaptic'' strengths of the links that form the tour are then enhanced (reduced) if the tour length is shorter (longer) than the average result of the previous trials. We perform extensive simulations of the random distance traveling-salesman problem. For sufficiently slow learning rates, near optimal tours can be obtained with the average optimal tour lengths close to the lower bounds for the shortest tour lengths.

In the spirit of the many recent simple models of evolution inspired by statistical physics, we put forward a simple model of the evolution of such models. Like its objects of study, it is (one supposes) in principle testable and capable of making predictions, and gives qualitative insights into a hitherto mysterious process.