Showing papers by "Santa Fe Institute published in 1994"

Fast folding and comparison of RNA secondary structures

Abstract: Computer codes for computation and comparison of RNA secondary structures, the Vienna RNA package, are presented, that are based on dynamic programming algorithms and aim at predictions of structures with minimum free energies as well as at computations of the equilibrium partition functions and base pairing probabilities. An efficient heuristic for the inverse folding problem of RNA is introduced. In addition we present compact and efficient programs for the comparison of RNA secondary structures based on tree editing and alignment. All computer codes are written in ANSI C. They include implementations of modified algorithms on parallel computers with distributed memory. Performance analysis carried out on an Intel Hypercube shows that parallel computing becomes gradually more and more efficient the longer the sequences are.

Genetic Programming II: Automatic Discovery of Reusable Programs.

Abstract: Reading Genetic Programming IE Automatic Discovery ofReusable Programs (GPII) in its entirety is not a task for the weak-willed because the book without appendices is about 650 pages. An entire previous book by the same author [1] is devoted to describing Genetic Programming (GP), while this book is a sequel extolling an extension called Automatically Defined Functions (ADFs). The author, John R. Koza, argues that ADFs can be used in conjunction with GP to improve its efficacy on large problems. "An automatically defined function (ADF) is a function (i.e., subroutine, procedure, module) that is dynamically evolved during a run of genetic programming and which may be called by a calling program (e.g., a main program) that is simultaneously being evolved" (p. 1). Dr. Koza recommends adding the ADF technique to the "GP toolkit." The book presents evidence that it is possible to interpret GP with ADFs as performing either a top-down process of problem decomposition or a bottom-up process of representational change to exploit identified regularities. This is stated as Main Point 1. Main Point 2 states that ADFs work by exploiting inherent regularities, symmetries, patterns, modularities, and homogeneities within a problem, though perhaps in ways that are very different from the style of programmers. Main Points 3 to 7 are appropriately qualified statements to the effect that, with a variety of problems, ADFs pay off be-

From Sequences to Shapes and Back: A Case Study in RNA Secondary Structures

Abstract: RNA folding is viewed here as a map assigning secondary structures to sequences. At fixed chain length the number of sequences far exceeds the number of structures. Frequencies of structures are highly non-uniform and follow a generalized form of Zipf's law: we find relatively few common and many rare ones. By using an algorithm for inverse folding, we show that sequences sharing the same structure are distributed randomly over sequence space. All common structures can be accessed from an arbitrary sequence by a number of mutations much smaller than the chain length. The sequence space is percolated by extensive neutral networks connecting nearest neighbours folding into identical structures. Implications for evolutionary adaptation and for applied molecular evolution are evident: finding a particular structure by mutation and selection is much simpler than expected and, even if catalytic activity should turn out to be sparse in the space of RNA structures, it can hardly be missed by evolutionary processes.

Generating surrogate data for time series with several simultaneously measured variables.

Abstract: We propose an extension to multivariate time series of the phase-randomized Fourier-transform algorithm for generating surrogate data. Such surrogate data sets must mimic not only the autocorrelations of each of the variables in the original data set, they must mimic the cross correlations between all the variables as well. The method is applied both to a simulated example (the three components of the Lorentz equations) and to data from a multichannel electroencephalogram.

The Quark and the Jaguar: Adventures in the Simple and the Complex

Abstract: This book is about how the wonderful diversity of the universe can arise out of a set of fairly simple basic laws. It is written by an expert in both the fundamental laws and the complex structures that they can produce.' Stephen Hawking's acclaim of Murray Gell-Mann's literary debut is typical of the reception the book received on first publication in 1994. From one of the twentieth century's greatest scientists comes this unique, highly personal vision of the connections between the basic laws of physics and the complexity and diversity of the natural world. THE QUARK AND THE JAGUAR - the simple and the complex - is an irresistibly engaging and rewarding introduction to the life's work of physicist, polymath and Nobel Laureate Murray Gell-Mann.

Complexity: Metaphors, Models, and Reality

Abstract: Fundamental Concepts Examples of Complex Adaptive Systems Nonadaptive Systems, Scaling, Self-Similarity, and Measures of Complexity General Discussion Afterwords.

Genetic algorithms and artificial life

Abstract: Genetic algorithms are computational models of evolution that play a central role in many artificial-life models. We review the history and current scope of research on genetic algorithms in artificial life, giving illustrative examples in which the genetic algorithm is used to study how learning and evolution interact, and to model ecosystems, immune system, cognitive systems, and social systems. We also outline a number of open questions and future directions for genetic algorithms in artificial-life research.

A Genetic Algorithm Discovers Particle-Based Computation in Cellular Automata

Abstract: How does evolution produce sophisticated emergent computation in systems composed of simple components limited to local interactions? To model such a process, we used a genetic algorithm (GA) to evolve cellular automata to perform a computational task requiring globally-coordinated information processing. On most runs a class of relatively unsophisticated strategies was evolved, but on a subset of runs a number of quite sophisticated strategies was discovered. We analyze the emergent logic underlying these strategies in terms of information processing performed by “particles” in space-time, and we describe in detail the generational progression of the GA evolution of these strategies. Our analysis is a preliminary step in understanding the general mechanisms by which sophisticated emergent computational capabilities can be automatically produced in decentralized multiprocessor systems.

What would be conserved if "the tape were played twice"?

Abstract: We develop an abstract chemistry, implemented in a lambda-calculus-based modeling platform, and argue that the following features are generic to this particular abstraction of chemistry; hence, they would be expected to reappear if "the tape were run twice": (i) hypercycles of self-reproducing objects arise; (ii) if self-replication is inhibited, self-maintaining organizations arise; and (iii) self-maintaining organizations, once established, can combine into higher-order self-maintaining organizations.

Program Search with a Hierarchical Variable Lenght Representation: Genetic Programming, Simulated Annealing and Hill Climbing

Abstract: This paper presents a comparison of Genetic Programming(GP) with Simulated Annealing (SA) and Stochastic Iterated Hill Climbing (SIHC) based on a suite of program discovery problems which have been previously tackled only with GP. All three search algorithms employ the hierarchical variable length representation for programs brought into recent prominence with the GP paradigm [8]. We feel it is not intuitively obvious that mutation-based adaptive search can handle program discovery yet, to date, for each GP problem we have tried, SA or SIHC also work.

Statistics of RNA melting kinetics

Abstract: We present and study the behavior of a simple kinetic model for the melting of RNA secondary structures, given that those structures are known. The model is then used as a map that. assigns structure dependent overall rate constants of melting (or refolding) to a sequence. This induces a “landscape” of reaction rates, or activation energies, over the space of sequences with fixed length. We study the distribution and the correlation structure of these activation energies.

A description of holland's royal road function

Abstract: This paper contains a description of John Holland's Royal Road function, which was presented at the Fifth International Conference on Genetic Algorithms in July 1993 and posted to the Internet Genetic Algorithms mailing list in August 1993.

Linear operators on correlated landscapes

Abstract: In this contribution we consider the effect of a class of «averaging operators» on isotropic fitness landscapes. Explicit expressions for the correlation function of the averaged landscapes are derived. A new class of tunably rugged landscapes, obtained by iterated smoothing of the random energy model, is established. The correlation structure of certain landscapes, among them the Sherrington-Kirkpatrick model, remains unchanged under the action of all averaging operators considered here

Testing For Nonlinearity Using Redundancies: Quantitative and Qualitative Aspects

Abstract: A method for testing nonlinearity in time series is described based on information-theoretic functionals -- redundancies, linear and nonlinear forms of which allow either qualitative, or, after incorporating the surrogate data technique, quantitative evaluation of dynamical properties of scrutinized data. An interplay of quantitative and qualitative testing on both the linear and nonlinear levels is analyzed and robustness of this combined approach against spurious nonlinearity detection is demonstrated. Evaluation of redundancies and redundancy-based statistics as functions of time lag and embedding dimension can further enhance insight into dynamics of a system under study.

Dynamics of Small Autocatalytic Reaction Network II: Replication, Mutation and Catalysis

Abstract: Mutation is introduced into autocatalytic reaction networks. Examples of low dimensional dynamical systems --- n = 2, 3 and 4 --- are discussed and complete qualitative analysis is presented. Error thresholds known from simple replication-mutation kinetics with frequency independent replication rates occur here as well. Instead of cooperative transitions or higher order phase transistions the thresholds appear here as supercritical or subcritical bifurcations being analogous to first order phase transitions.

Two tools to test time series data for evidence of chaos and/or nonlinearity.

Abstract: Two computer programs are described for evaluating the evidence for chaos and nonlinearity in time series data. “bx” is an efficient algorithm for computing the correlation integral (from which correlation dimension can be estimated); and “surrogat” is a Fourier-transform-based algorithm for generating surrogate data consistent with a null hypothesis that the data arise as a result of a linear stochastic process.

The influence of mutation on autocatalytic reaction networks

Abstract: A particular class of ordinary differential equations (ODEs) describing catalyzed, template-induced, and erroneous replication is investigated. The ODEs can be split into a replicator part accounting for the correct replication and a mutation term accounting for all miscopying processes. The set of all species is divided into the catalitically active “viable” species and an error tail subsuming all other species. Neglecting both the intermutation among the viable species and the reflux from the error tail allows for an extensive analysis of the autocatalytic network. If mutation rates are small enough, a perturbation approach is feasible showing that mutation in general simplifies the qualitative behavior of the dynamical system. Special cases, such as Schlogl's model, the uniform model, and the hypercycle, show that the viable species become unstable beyond a critical mutation rate: There is an analogue to the error threshold of the quasi-species model also in nonlinear autocatalytic reaction networks with mutation.

Lattice polymer automata

Abstract: We present a new style of molecular dynamics and self-assembly simulation, the Lattice Polymer Automaton [LPA]. In the LPA all interactions, including electromagnetic forces, are decomposed and communicated via propagating particles, “photons”. The monomer-monomer bond forces, the molecular excluded volume forces, the longer range intermolecular forces, and the polymer-solvent interactions may all be modeled with propagating particles. The LPA approach differs significantly from both of the standard approaches, Monte Carlo lattice methods and Molecular Dynamics simulations. On the one hand, the LPA provides more realism than Monte Carlo methods, because it produces a time series of configurations of a single molecule, rather than a set of causally unrelated samples from a distribution of configurations. The LPA can therefore be used directly to study dynamical properties; one can in fact watch polymers move in real time. On the other hand, the LPA is fully discrete, and therefore much simpler than traditional Molecular Dynamics models, which are continuous and operate on much shorter time scales. Due to this simplicity it is possible to simulate longer real time periods, which should enable the study of molecular self-organization on workstations; supercomputers are not needed.

Catching a baseball: a reinforcement learning perspective using a neural network

Abstract: Moments after a baseball batter has hit a fly ball, an outfielder has to decide whether to run forward or backward to catch the ball. Judging a fly ball is a difficult task, especially when the fielder is in the plane of the ball's trajectory. There exists several alternative hypotheses in the literature which identify different perceptual features available to the fielder that may provide useful cues as to the location of the ball's landing point. A recent study in experimental psychology suggests that to intercept the ball, the fielder has to run such that the double derivative of tanϕ with respect to time is close to zero (i.e. d2 (tanϕ) / dt2 ≈ 0), where ϕ is the elevation angle of the ball from the fielder's perspective (McLeod & Dlenes 1993). We investigate whether d2 (tanϕ)/dt2 information is a useful cue to learn this task in the Adaptive Heuristic Critic (AHC) reinforcement learning framework. Our results provide supporting evidence that d2 (tanϕ) / dt2 information furnishes strong initial cue in determining the landing point of the ball and plays a key role in the learning process. However our simulations show that during later stages of the ball's flight, yet another perceptual feature, the perpendicular velocity of the ball (vp) with respect to the fielder, provides stronger cues as to the location of the landing point. The trained network generalized to novel circumstances and also exhibited some of the behaviors recorded by experimental psychologists on human data. We believe that much can be gained by using reinforcement learning approaches to learn common physical tasks, and similarly motivated work could stimulate useful interdisciplinary research on the subject.