We introduce a new model of evolution on a fitness landscape possessing a tunable degree of neutrality. The model allows us to study the general properties of molecular species undergoing neutral evolution. We find that a number of phenomena seen in RNA sequence-structure maps are present also in our general model. Examples are the occurrence of “common ” structures which occupy a fraction of the genotype space which tends to unity as the length of the genotype increases, and the formation of percolating neutral networks which cover the genotype space in such a way that a member of such a network can be found within a small radius of any point in the space. We also describe a number of new phenomena which appear to be general properties of neutrally evolving systems. In particular, we show that the maximum fitness attained during the adaptive walk of a population evolving on such a fitness landscape increases with increasing degree of neutrality, and is directly related to the fitness of the most fit percolating network. 1

Abstract not found

Local search methods constitute one of the most successful approaches to solving large-scale combinatorial optimization problems. A new result concerning the parallelization of such methods is presented. As parallelism is increased, optimization performance initially improves, but then abruptly degrades to no better than random search beyond a certain point. The existence of this transition is demonstrated for a family of generalized spin-glass models and the Traveling Salesman Problem. Finite-size scaling is used to characterize size-dependent effects near the transition and analytical insight is obtained through a mean field approximation.

I present a general theory of mean field approximation, which is based on information geometry and is applicable not only to Boltzmann machines but also to wider classes of statistical models. Using perturbation expansion of the Kullback divergence (or Plefka expansion in statistical physics), a formulation of mean field approximation of general orders is derived. It includes in a natural way the "naive" mean field approximation, and is consistent with the Thouless-Anderson-Palmer (TAP) approach and the linear response theorem in statistical physics.

The hierarchical Bayesian optimization algorithm (hBOA) can solve nearly decomposable and hierarchical problems of bounded difficulty in a robust and scalable manner by building and sampling probabilistic models of promising solutions. This paper analyzes probabilistic models in hBOA on two common test problems: concatenated traps and 2D Ising spin glasses with periodic boundary conditions. We argue that although Bayesian networks with local structures can encode complex probability distributions, analyzing these models in hBOA is relatively straightforward and the results of such analyses may provide practitioners with useful information about their problems. The results show that the probabilistic models in hBOA closely correspond to the structure of the underlying problem, the models do not change significantly in subsequent iterations of BOA, and creating adequate probabilistic models by hand is not straightforward even with complete knowledge of the optimization problem. Categories and Subject Descriptors

A number of authors have in recent years proposed that the processes of macroevolution may give rise to self-organized critical phenomena which could have a significant effect on the dynamics of ecosystems. In particular it has been suggested that mass extinction may arise through a purely biotic mechanism as the result of so-called coevolutionary avalanches. In this paper we first explore the empirical evidence which has been put forward in favor of this conclusion. The data center principally around the existence of power-law functional forms in the distribution of the sizes of extinction events and other quantities. We then propose a new mathematical model of mass extinction which does not rely on coevolutionary effects and in which extinction is caused entirely by the action of environmental stresses on species. In combination with a simple model of species adaptation we show that this process can account for all the observed data without the need to invoke coevolution and critical processes. The model also makes some independent predictions, such as the existence of “aftershock ” extinctions in the aftermath of large mass extinction events, which should in theory be testable against the fossil record. 1

We present a systematic approach to mean field theory (MFT) in a general probabilistic setting without assuming a particular model. The mean field equations derived here may serve as a local and thus very simple method for approximate inference in probabilistic models such as Boltzmann machines or Bayesian networks. "Model-independent" means that we do not assume a particular type of dependencies; in a Bayesian network, for example, we allow arbitrary tables to specify conditional dependencies. In general, there are multiple solutions to the mean field equations. We show that improved estimates can be obtained by forming a weighted mixture of the multiple mean field solutions. Simple approximate expressions for the mixture weights are given. The general formalism derived so far is evaluated for the special case of Bayesian networks. The benefits of taking into account multiple solutions are demonstrated by using MFT for inference in a small and in a very large Bayesian network. The results are compared to the exact results.

Economic Forum within the framework of the

(will be inserted by the editor)

Indexing, retrieval, and summarization in recordings of meetings have, to date, focused largely on the propositional content of what participants say. Although objectively relevant, such content may not be the sole or even the main aim of potential system users. Instead, users may be interested in information bearing on conversation flow. We explore the automatic detection of one example of such information, namely that of hotspots defined in terms of participant involvement. Our proposed system relies exclusively on low-level vocal activity features, and yields a classification accuracy of 84%, representing a 39 % reduction of error relative to a baseline which selects the majority class.