Madison Chaos and Complex Systems Seminar

Fall 1998 Seminars

All seminars are Tuesday at 12:05 pm in 4274 Chamberlin

Short List

September 8, 1998

Resolving Perceptual Ambiguity in the Necker Cube:  A Dynamical Systems Approach

Deborah J. Aks, University of Wisconsin - Whitewater

A dynamical systems approach is used to assess how the human visual system resolves perceptual ambiguity.  Three sets of subjects observed an ambiguous figure -- the Necker cube -- for 15, 30 and 60 minute periods during which they pressed a key each time they perceived a change in orientation of the cube (i.e., top-left vs. bottom-right). Manipulations of binocular disparity served as a parameter to control perceptual stability. Low depth conditions yielded more perceptual reversals than intermediate and high depth conditions.  Linear and non-linear time series analyses were performed on time intervals between reversals.  Data show 1/f (pink noise) was predominant in high depth conditions and white noise was predominant in low depth conditions.  These results suggest that depth information may guide our perceptual system into a self-organizing system allowing us to resolve ambiguous information. Such pink (1/f) noise is intermediate between brown: 1/f2 and white: 1/f0 noise, and possesses the important property of self-similarity at all scales.  I will discuss how this pattern of behavior is thought to be characteristic of efficent coding systems (i.e., Field, 1987) and adaptive complex systems with emergent, self-organizing properties (i.e., Gilden, Thornton, & Mallon, 1995)

September 15, 1998

Evaluating Fundamental Variables in Multivariate Studies of Nonlinear Complex Phenomena

Frederick David Abraham, Blueberry Brain Institute, Waterbury Center VT, & Visiting Professor of Psychology and Graduate Studies, Silliman University, Philippines

Traditional techniques of estimating the number and nature of basic processes involved in nonlinear research usually use step-wise geometric attractor construction and statistical evaluation of invariant properties of the attractor from a single time series. Stewart has generalized one of these techniques, false nearest neighbors, to the multivariate situation, which I argue, is preferable not only in efficiently identifying the degrees of freedom of a system, but in giving better hints as to what the basic processes might be that could provide more information for the modeling process. These will be compared with the usual linear methods of evaluating brain cooperativity, and also use Stewart's exploration of models, such as Lorenz's on weather. The presentation will be very elementary, presenting the main features of the process without mathematical details; audience participation is encouraged; informal and fun. (Ref: Abraham, "Nonlinear coherence in multivariate research: Invariants and the reconstruction of attractors", 1997, Nonlinear Dynamics, Psychology, and Life Sciences, 1, 7-33; short footnote:

September 22, 1998

Statistical Learning of Phonetic Patterns

Lori L. Holt, UW Department of Psychology

The most extraordinary feat universally accomplished by humans is the acquisition of speech. Remarkably, by their first birthday, human infants appear to have accomplished the rather daunting task of becoming initiated into a native language. Typically, this striking rapidity has been interpreted as evidence that language acquisition is guided by innate processes. I'll discuss studies driven by the alternative hypothesis that infants' emerging language-sensitive perception is a consequence of self-organized learning arising from experience with sound co-variation inherent to languages.

September 29, 1998

How Can Statistical Mechanics Contribute to the Study of Inequality?

Steve Durlauf, UW Department of Economics

This talk will review recent theoretical work in economics which has used methods from statistical mechanics to study how persistent inequality can arise across various socioeconomic groups.  Applications to a number of social pathologies such as crime and drug use will be considered. Implications of this perspective for antipoverty policies will also be examined.

October 6, 1998

High Throughput Computing With Distributively Owned Resources

Miron Livny, UW Computer Sciences Department

Recent dramatic decrease in the cost-performance ratio of computing resources placed powerful computing capabilities on almost every desk. These desktop resources are owned by individuals or small groups and are physically scattered throughout the globe. In the talk we will discuss the challenges we face in turning very large collections of loosely coupled and distributively owned computing resources into a High Throughput Computing environment that can effectively deliver large amounts of computing cycles over very long time periods. The talk is based on our decade long experience with the Condor system and our recent involvement in an effort to build a national computation grid.

October 13

Sleep: The Forest behind the Trees

Bill Obermeyer, Wisconsin Psychiatric Institute

The behavior and physiology of all birds and mammals is profoundly reorganized several times a day.  The very pervasiveness of sleep has made it, like the proverbial forest amid the trees, invisible to science - until recently.   Most models of sleep have arisen from models of cyclic behavior, whereas models in disciplines not focused on sleep largely ignore its influence.  I will present information on the nature and organization of sleep/wake.  I will try to show how what we know about sleep fits with current models of sleep, and how sleep itself changes the behavior of individuals, populations and ecological sytems.

October 20

DNA Computing on Surfaces

Anne Condon, UW Computer Sciences Department

In 1994, Len Adleman described how to solve a small instance of a famous combinatorial problem - the Traveling Salesman problem - in a novel way.  Adleman's method was to efficiently create a test tube of DNA strands, each representing a possible solution to the problem, and to extract the elusive true solution using tools from molecular biology.

Adleman's work raises many questions at the interface of chemistry, mathematics, and computer science.  How can digital information be efficiently stored in, and subsequently retrieved from, DNA molecules? How can logical operations be performed on information-carrying DNA strands?  What kinds of combinatorial problems can be solved using such tools?  How can the errors that are inherent in such DNA "computations" be controlled?  What might be useful applications of DNA computing?

In this talk, we describe work at U. Wisconsin-Madison that addresses these questions.  The premise of the Wisconsin DNA computing group is that surface-based chemistry will be a critical technology in realizing DNA computation.  With this approach, many DNA strands are immobilized on a planar surface.  Logical operations are performed on all of the strands in parallel, using chemical and enzymatic processes.  This surface-based approach allows a much greater degree of control in the chemical processes than that achievable via the test tube based methodology of Adleman.  The talk will emphasize the combinatorial problems that arise in building our prototype DNA computer.

The DNA computing project at UW-Madison involves Professors Max Lagally (Materials Science), Rob Corn and Lloyd Smith (Chemistry), and Anne Condon (Computer Sciences), along with several students.

October 27, 1998

The Internet as a Complex System

Tad Pinkerton, UW Department of Computer Sciences

The global data network known as the Internet is a very complex, if not chaotic, system.  There is no central management, parts of the infrastructure are provided by many different organizations, and its users determine how and for what it is used.  This talk will explore the dimensions of this network and the limited nature of the controls by which it can be managed.

November 3, 1998

How to Untangle a Pot of Boiling Spaghetti by Jiggling the Flame

Kevin Mirus, UW Department of Physics

The field of controlling chaotic systems began to heat up in 1990 when Ott, Grebogi, and Yorke prescribed a simple recipe for controlling the Henon Map by applying small-amplitude feedback.  This method was based on some general geometric properties of the unstable periodic orbits (UPOs) found in chaotic systems, and thus has had a broad range of applicability.  Using some of these same geometric considerations, a method for controlling chaotic systems without using feedback will be described.  This method involves a simple periodic perturbation applied to an accessible system parameter.  By evaluating the UPOs found in the chaotic system, the frequency of a controlling perturbation can be precisely predetermined.  The effectiveness of this method will be illustrated for a number of low and high-dimensional systems.

November 10, 1998

High-level Learning of Early Visual Tasks

Pawan Sinha, UW Department of Psychology

Does learning shape perception? Recent experimental results have provided strong evidence demonstrating the experience-dependent malleability of perceptual processes. Several lines of reasoning suggest that the locus of learning in the reported instances is situated very early along the processing pathway. The question of whether high-level, object specific learning can influence early perception is still largely unresolved. The answer is important for understanding the overall organization of information flow in the brain.

We examine this question in the context of a specific visual perceptual task -- the recovery of 3-D structures from single 2-D line-drawings. Our exploration has three conceptual parts.

1. Study of the limitations of a shape-recovery model devoid of high-level learning based influences.

2. Experiments to demonstrate the existence and importance of high-level learning on 3D form perception in humans.

3. A computational model for incorporating high-level learning in early perception.

The talk will describe our work so far in these directions.

November 17, 1998

The Development and Organization of Political Parties

Charles Franklin, UW Department of Political Science

Political parties are universal elements of democratic political systems. They provide the common organizational structure of a staggering variety of particular institutional forms. And parties organize politics for the mass public, helping citizens make sense of political conflict.  Despite this universal development, there are few good explanations for why parties form in the first place. Party organization necessarily requires individual politicians to constrain their options in order to support common positions. How politicians can cooperate in the face of incentives for defection is the key puzzle of this paper, and of a model of the development of parties in both legislative and mass public settings.

November 24, 1998

Lake Management:  Dynamics of Choice, Uncertainty and Water Quality

Steve Carpenter, UW Center for Limnology

Lake ecosystem services (such as clean water and fisheries) depend on dynamic processes that range from the very slow (soil pollutants) to intermediate speeds (fish habitat, recycling of pollutants from sediments) to fast (choices by farmers, property owners, and anglers). Models of social systems composed of heterogeneous agents interacting with lake ecosystems show very complex behavior, with multi-stable state outcomes for lake condition, expenditures on research and monitoring, and dominance of environmentalist or exploitationist viewpoints.  In some respects, dynamics of these models resemble the panarchy theory of C.S. Holling. I will show a few examples, and discuss future directions for the research.

December 1, 1998

Chaotic Nature of the Change of the Seasons

David D. Houghton, UW Department of Atmospheric and Oceanic Sciences

The irregular characteristics of atmospheric variability are particularly pronounced during the transition seasons of the annual cycle.  Some examples are presented for Madison's weather and for the wind-flow jet streams over large areas with a focus on the autumn season.  Overall the variability increases going from summer into autumn.  An analysis of abrupt nature of the increases in variability of the kinetic energy of the jet streams shows that the "onset time" of autumn conditions can be quite different from one year to the next.  The time series of the kinetic energy has characteristics which resemble those found by Lorenz in his study of chaos for a simple nonlinear equation system.  Examination of some of the related physical processes for variability, such as interactions between the extratropical storms and jet stream flow, highlight the challenge for ever being able to predict when autumn will "begin."

December 8, 1998

Lidar Derived Animations of Coherent Structure in Turbulent Atmospheric Boundary Layer Flows

Ed Eloranta, UW Department of Atmospheric and Oceanic Sciences

Turbulent flows are typically described in terms of statistical moments, length scales and spectral distributions without reference to coherent structure in the flow. Lidar, the optical equivalent of radar, provides vivid animations of the highly organized coherent structures present in convective flows. Lidar derived animations will be presented which include the wildly convective conditions which produce steam fog and steam devils over Lake Michigan in the winter.

December 15, 1998

Olfactory Cortex and Associative Memory: Analysis and Predictions based on a Neural Network Approach.

Kurt R. Illig, UW Department of Anatomy

Olfactory cortex has at least two features that parallel artificial neural networks: 1) it receives input that is widely distributed spatially, 2) it has a large number of connections with itself (associative connections). Currently, we are working to more carefully define the organization of the input and associative connections, and to establish the physiological response haracteristics of neurons within olfactory cortex. Results of these studies suggest that olfactory cortex may be a good biological correlate of artificial neural networks. It is my hope that presentation of these results in this context will lead to discussion of cortical information processing and associative memory, in both biological and artificial neural networks.