Madison Chaos and Complex Systems Seminar

Fall 1995 Seminars

Dates, speakers, titles and abstracts will be listed as they become available. All seminars will meet in 5134 Chamberlin, at 12:05 Wednesdays, except as noted. At the moment this very bare, to say the least!

Short List

September 20. William Brock, UW Economics. ``Stock Price Modelling in Complex Environments.''

No abstract yet.

September 27.Richard Belew, Departments of Computer Science and Engineering, University of California at San Diego. ``Adaptive Individuals in Evolving Populations: Models and Algorithms.''

Abstract: Even the simplest creature is marvelous to observe as it transforms itself to better match the environment in which it finds itself. How is such adaptation accomplished? How much of this capability should be attributed to the particular individual we happen to be observing, how much to its species, and how much to the inclusive evolutionary processes wedding all life to this planet? How did the elaborate individual learning process we find in complex organisms evolve? Once in place, how does an increased individual capacity for adaptation alter the selective pressures causing the species to adapt to its niche?

 This talk will give an overview of a book (of the same title) edited by R. K. Belew and Melanie Mitchell, to be published soon by Addison-Wesley. The book grew out of a workshop at the Santa Fe Institute which brought together a group of about 20 scientists from biology, psychology, and computer science, all studying interactions between the evolution of populations and individuals' adaptations in those populations, and all of whom made some use of computational tools in their work. A good example is the ``Baldwin effect,'' a phenomena identified by the psychologist J. Mark Baldwin almost exactly a century ago, that arises repeatedly in many modern computer simulations. This talk will touch on the rich historical heritatge of such phenomena, ranging from biologists like Lamarck and Waddington to psychologists like Piaget and Skinner, as well as the new insights offered by computer simulations and new algorithms inspired by the same insights.

October 4. Larry Getzler, UW Economics. ``Bifurcations in Industrial Location.''

Abstract: If initial conditions were slightly different, might have the North American industrial belt been located elsewhere, or have been several small geographically disperse industrial centers instead? To study this and other issues I have created a dynamic plant location model with multiple firms and products, as well as spatially diverse inputs and consumers. In the model, geographic patterns of industrialization depend on specific locations of demand, strength of demand at different locations, transport costs for inputs, transport costs for outputs, input prices, input locations, and the locations of competitors. Firms act as price discriminating monopolists when there are no competitors close enough to be able to undercut their price.

I implement this model by performing a series of computer simulations. For some large ranges of initial conditions essentially the same industrial patterns occur. However, there are frontiers (bifurcations) where a small difference in initial conditions leads to drastically different results. A monotonic change in a single condition, the marginal cost of transporting inputs, can cause firms to make large spatial jumps back and forth between coordinates in their location choice. The multiple causes of these bifurcations are discussed.

October 11. Tim Allen, UW Botany. ``What Is Chaos Good For? Not As Much As You Think.''

No abstract yet.

October 18. Joel Robbin, UW Math. ``Lyapunov Exponents.''

No abstract yet.

October 25. Una-May O'Reilly, AI Lab, MIT (from Jan. 1996). ``A Comparative Approach to Understanding Genetic Programming.''

Abstract: Genetic Programming (GP) [Koza, 1993] is a Genetic Algorithm specialized to perform program induction (i.e. the automatic generation of a computer program from a set of input-output pairs). As a GA it is inspired by the adaptive process of evolution and has computational functionality crudely equivalent to ``survival of the fittest'', reproduction and ``genetic crossover''.

I shall introduce GP and compare it to two well understood adaptive search algorithms: Iterated Hill Climbing and Simulated Annealing. All three algorithms are used to solve the same suite of program induction problems, posed in exactly the same style. The comparison quantitatively evaluates the optimization power of the evolution-based GP algorithm.

I also uncouple the ``genetic crossover'' operator of GP and use it as the search operator in Hill Climbing and Simulated Annealing. There are two reasons for this: to understand whether the resulting crossover induced ``fitness landscapes'' [Sewall Wright] are amenable to efficient search and to evaluate the role of recombination when the crossover operator is used in GP.

November 1. No seminar.

However, there is a talk of possible interest: Ivar Ekeland, Professor of Mathematics at the University of Paris-Dauphine, and author of one of the best popular books on chaos and nonlinear dynamics (Mathematica and the Unexpected) will be giving a Hilldale Lecture on ``Variaitional Principles and Symplectic Geometry: From Galileo's Pendulum to Modern Symplectic Geometry''. (Place: room B130, Van Vleck Hall; Time: Wed., Nov. 1, 4:00 pm.)

November 8. Ron Burnette, UW Pharmacology. ``Analysis of the Influence of Aging on Human Hemodynamics and Its Impact on Drug Action.''

Abstract: Biological systems are extremely complex and one could argue that human beings represent the most complex of the biological systems. Assuming this to be true, one wonders how it is posssible for clinicians to have any hope of treating patients in an optimal fashion? What makes this question of even more concern is that a clinician functions in an extremely data poor environment.

The purpose of this seminar is to try to partially answer this question, in a restricted sense, by using the techniques of principal component analysis, constrained Monte Carlo simulation, sensitivity analysis and graph theory. Every attempt will be made convey this material from a qualitative conceptual viewpoint along with providing needed biological background information.

November 15. Paul Plummer, UW Geography. ``Nonlinear Dynamics in Spatially Interdependent Markets.''

Abstract: Typically, attempts to explain spatial variations in prices, profits, and outputs have concentrated on developing models of spatial price equilibria. However, such equilibria are only of interest if it can be shown that they are stable in the sense that firms pursuing a disequilibrium adjustment strategies tend to drive the market towards an equilibrium configuration of prices, profits and outputs in a sufficiently short time period. In this presentation, I examine the disequilibrium dynamics of a model of intra-urban spatial competition in which oligopolistically competitive firms sell a relatively homogenous commodity directly to consumers. I focus on three questions: the conditions under which the model converges to a spatial price equilibrium; the nature of the space-time dynamics of both price and profit differentials during the period of disequilibrium as compared to spatial price equilibrium; the diffusion of price changes throughout the configuration of the urban market.

November 22


November 29. Scott Page, Dept. of Economics, California Institute of Technology. `Problem Solving by Teams of Heterogeneous Agents.'' (Co-author, Lu Hong, Dept. of Economics, Syracuse University.)

Abstract: We construct a model of problem solving by teams of heterogeneous agents with limited ability which elucidates differences between problem solving firms and manufacturing firms. The heterogeneity refers to differences in how individual problem solvers perceive problems and in how they attempt to solve them. These differences enable teams of problem solvers to outperform individuals. In applying this model of heterogeneous problem solvers to production theory, we arrive at some uncomfortable conclusions: among them that arbitrary returns to additional problem solvers are possible and that a team of people of ``equal ability'' applied to a single problem might exhibit increasing returns or decreasing returns depending upon the order they are hired. We can formulate assumptions which generate decreasing returns, but they rely on such pessimistic assessments of the abilities of problem solvers so as to be unrealistic.

December 6


December 13. George Sugihara, Scripps Institution of Oceanography, UC San Diego. ``Extracting Nonlinearity from Natural Time Series.'' --- CANCELLED

Abstract: Although identifying chaos in real data sets is a controversial subject, there is much to be gained by focusing on the nonlinearity that may be found in certain time series data. Here I review some of the issues concerning the detection of nonlinearities and possible chaos in nature, particularly with regard to stochastic chaos. I will also discuss several examples, where characterizing and exploiting nonlinearity can provide some fundamental insights about nature. Among these, is a demonstration using atmospheric data that shows how one can extract the functional form of the dominant nonlinear signal. This signal, in turn, is used to gain insight into the underlying mechanisms involved and can lead to better forecasts.

CANCELLED! Dr. Sugihara has unfortunately come down with the flu.

December 20. Organizational meeting.

Usual time and place; to decide on time and place for next semester's seminars, possible speakers, and other administrivia.
Up to the Chaos and Complex Systems Seminar page.
(Tue Dec 12 15:30:26 CST 1995)