ICCS ABSTRACTS: Alphabetically by first author 

 Iqbal Adjali - BTexact Technologies

Modelling Third Generation Mobile Spectrum Auctions. Iqbal Adjali, David Collings and Paul Marrow

 Licences for Third Generation Mobile (3GM) spectrum have been auctioned in the UK and other countries
around the world. An understanding of the dynamics of the auction design used, and the likely strategies
which bidders might have to adopt, was of vital importance to telecom operators who wished to win a
licence in these spectrum auctions. There is also increasing interest in auctions in general, born out
of the burgeoning Internet-based auction business and the potential use of auctions to sell telecoms
bandwidth on the wholesale and even retail markets. Research into auction theory will therefore be of
increasing importance to providers of telecommunication services. The UK 3GM Auction used a structure
based on simultaneous compulsory bidding in a form of multi-unit ascending auction. The structure is
based upon the type of auction used successfully by the US Federal Communications Commission in selling
radio spectrum in the United States, but introduces a number of new features which are thought to
improve the effectiveness of the auction mechanism. A team from BT with expertise in game theory and
complex system modelling applied two techniques to aid the understanding of the UK 3GM Auction. First,
they constructed a simulation tool which allowed the modelling of entire auctions, and the analysis of
the different outcomes produced by an ensemble of auctions. In this paper results from the simulation
tool are presented and compared with the actual auction. The second analysis technique used was the
construction of a mock auction tool which could be used to facilitate practise auctions, in order to
learn bidding strategies interactively. This tool was used in practise for the auction in training the
teams involved in bidding. Here results of the mock auctions are discussed and their effectiveness in
preparing for the real auction is analysed. In the final section of the paper the likely role of
auctions in telecommunications in a wider context are discussed. 


Christof Aegerter - Vrihe Universiteit Amsterdam

Two-Dimensional Rough Surfaces: Experiments on Superconductors and Rice-Piles.

 The study of the kinetic roughening of interfaces has long been studied in e.g. imbibition systems.
These studies were however up to now restricted to a single spatial dimension. We present two new
experimental systems, where the roughening process can be studied in two spatial dimensions. In the
first case, the surface of a 2d pile of rice is studied and in the second case the magnetic fluxscape in
a type-II superconductor is investigated. Furthermore, we discuss some questions arising in the analysis
of the structure of such surfaces. 

 Nelli Ajabyan - Institute
Hydroecology&Ichthyology, NAS Armenia

Global Stability and Oscillations in the Mathematical Models of Population Interactions With Spatial
Heterogenity

 This paper presents the investigation of oscillations and nonlinear dynamics of models of ecological
systems with spatial heterogeneity. The dynamics of coupled oscillators is proved to be relevant in the
study of pattern generation of a biological system. Patterns of Hopf bifurcation, started with Turing
model, have been an active subject of research in recent years [1]. The processes of pattern generation
in biological systems exhibit universal geometric properties in one hand and a high sensitivity with
respect of external parameters on the other hand. The contrast between general geometric properties of
generated patterns and their strong dependence on the parameters of a particular process make them a
very attractive subject of theoretical and experimental research. The recognition that the coupling
network acts as a filter suggests that it is possible to alter bifurcation scenarios by using different
filters in the coupling. It is well acknowledged that the presence of symmetry in a dynamical system
will change the generic behavior of that system. This provided credence to the idea of using approximate
maps and the concepts of symmetry breaking bifurcation of chaotic attractors in physical experiments
[2]. In this paper some global bifurcation phenomena associated with networks of identical oscillators
will be reviewed. As an application global bifurcation of phase locked oscillators are applied to the
study of migratory affects in predator-prey system. The model of n systems connected with migration
flows was studied by Svirezhev Yu M.[3]The study of nonlinear phenomena has applied motivation in
providing the concept for some generations of ecological models. To obtain the models that based on more
realistic hypotheses about natural system behavior it is possible to lean up on well-known mathematical
concepts. A trophic chain with nonlinear functions describing prey-predator interaction is one of the
examples of such generalization. The application of the ecological stability concept allows gaining
model explanation of some observations and effects that had not interpretation in the frame of existing
mathematical theory. Literature cited1. Ashwin P., King C.P., W. Swift. Three identical oscillators with
symmetric coupling. Nonlinearity, 3,1990, p581-601. Priented in UK.2. J.C. Alexander G Auchmuty Global
Bifurcations of Phase-Locked oscillators . Archive for rational meachanics and Analysis Springer_Verl.
Berlin, v. 93, N 3, 1988, p 253-2703. Svirezhev Yu. M. Nonlinear waves dissipative structures and
catastrophes in ecology. Moscow, 1987. 

 Shah Alam - Shahjalal University of
Science and Technology

Algebra of Mixded Number

 Mixed number is the sum of a scalar and and vector like quaternion but the algebra of mixed number is
different from that of quaternion. In this paper the details algebra (math tools) of mixed number is
explained. 

 Eric Allison - Pratt Institute

Self Organization in Cities: Case Studies in Neighborhood Historic Preservation

 While the theory of cities as self organizing entities is documented and acknowledged in both academic
and popular writing (Jacobs, 1993; Johnson 2001; Portugali, 2000; Bar-Yam, 1977, nd), there are
relatively few case studies documenting actual examples of emergent self organization in cities. This
paper describes the self organization which took place in New York City to effect the designation of
multiple neighborhoods as historic districts in the last third of the twentieth century. This unexpected
emergent behavior is continuing and spreading via formal and informal networks to other cities. In 1965,
the Mayor of New York signed into law a bill creating a Landmarks Preservation Commission charged with
designating and protecting historic structures in New York City. The newly-created Landmark Preservation
Commission could designate both individual buildings and historic districts. The New York City law was
the first of its kind in the United States. The Landmarks Preservation Commission, after public
hearings, could designate a building or district administratively. The designations took place
immediately. They were then submitted to the local legislature to preserve due process. The legislative
body originally the Board of Estimate, now the City Council were required to give a yes or no vote
within sixty days. The intent was that the Commission, staffed by experts, would survey the city, select
appropriate buildings, hold hearings, and designate those found worthy, via a top-down hierarchical
process. At the hearing for the bill creating the Commission, the future Executive Director of the
Commission who had helped write the bill testified that he anticipated eventually designating
approximately 1,000 individual landmarks and two or three historic districts. These initial conditions
led to a very different result than anticipated. Thirty-seven years later, New York City has some 1,300
individual landmarks not far off the original estimate but it also has over seventy historic districts,
encompassing more than 22,000 buildings. Among other factors, this unanticipated result emerged from the
responses of individual agents (city dwellers) to changing environment conditions, including:* changes
in the funding and personnel time and emotional energy available to support the goals;* changes in the
goals, themselves, as a result of the need to solve a host of complex social problems; * increased
competition for desirable, affordable places to live; and* immersion in the phase change occurring
throughout human civilization from hierarchical to hybrid and networked control structures (Bar-Yam,
nd). Space permitting, I will develop four case studies. Three cases will describe different instances
of self organizing neighborhoods and citizens (the Fort Greene Historic District in Brooklyn and the
Tribeca Historic District and the Ladies Mile Historic District both in Manhattan). The fourth will
document how New York City's Historic Districts Council created to garner support in the city's historic
districts for increasing the budget of the Landmarks Preservation Commission re-organized itself as a
network node among the various independent neighborhood groups. 

 Madhur Anand
- Laurentian University

The Evolution of Complexity in Natural and Reconstructed Ecological Assemblages. Madhur Anand, Ke-Ming
Ma, Brian Tucker and Rachelle Desrochers

 We study the complexity of ecological assemblages that have been severely damaged by man-made
perturbation (mainly air pollution) and ask the question: how is complexity affected and how well can we
alter recovery trajectories by intervention? We use data from both field and laboratory experiments in
Sudbury, Ontario that documented plant and microbial assemblage dynamics after varying degrees of
perturbation and rehabilitation in 2001. Assemblage-level patterns are modelled using two approaches.
The first attempts to summarize taxa-interactions either using information-theoretical measures or a
reduced multivariate space. The second attempts to fit a stationary Markov model to assemblage dynamics.
Assemblage-level dynamics was effectively summarized using Principal Components Analysis (> 80%
variation captured in first two axes); however spatiotemporal structure was not very well captured. The
Markov models provide an excellent fit to observed dynamics (p<0.001). The results show that in all
treatments the microbial assemblages reach equilibrium quickly but have different recovery pathways
under different treatments and that initial state matters. Information theoretical measures reveal that
the assemblage dynamics cannot be attributed to any single or dominant taxon but is rather an emergent
property of the system. 

 James J. Anderson - Center for Computational
Biology and Bioinformatics NIGMS

NIGMS Funding Opportunities for Complex Biomedical Systems Research and Training

 The National Institute of General Medical Sciences (NIGMS: a component of the U.S. National Institutes
of Health) announces a new Center for Bioinformatics and Computational Biology. This Center will promote
advances in cross-disciplinary research, education, and training involving quantitative approaches to
complex biological and biomedical problems, and allied bioinformatics. NIGMS has issued a number of
program announcements (PA) and requests for applications (RFA) that will provide support for these
areas. The suite of initiatives have as their foci: 1. The understanding of system principles and
dynamics in processes involving large numbers of interacting components, at all levels of
biologicalorganization within the scope of the NIGMS mission; 2. The development of analytical
methodologies to discover the genetic architecture of complex genetic traits; 3. The study of the
evolutionary dynamics of pathogens and their hosts with their environments; 4. The development of
enabling technologies useful for the study of metabolic processes and metabolic engineering; 5. The
development of basic mathematical concepts and algorithms that have the potential for significantly
advancing the state of the art inbiomedical research. The initiatives comprise mechanisms to fund
research projects (traditional research project grants (R01) and program project grants (P01)), to fund
establishment of integrative research efforts ("glue grants," R24), to fund extensive programs of
research related activities (Center grants(P50)), to provide support for short courses and workshops
(R25 education grants) for both biologists and non-biologists, and to provide trainingat both pre- and
postdoctoral level (T32 and T33 Training Grants). Detailed information on these programs will be
available, and can also be found at the following URL:http://www.nih.gov/nigms/funding. 

 Pierpaolo Andriani - University of Durham Business School

 The work presented in this paper aims at providing an empirical validation to some aspects of
Kauffman's laws of complex systems (Investigations, 2000). "...On a coevolutionary timescale, coevolving
autonomous agents as a community attain a self-organised critical state by tuning landscape structure
and couplings between landscapes, yielding a global power law distribution of extinction and speciation
events and a power law distribution of species lifetimes" and "...As an average trend, biospheres and
the universe create novelty and diversity as fast as they can manage to do so without destroying the
accumulated propagating organisation that is the basis and nexus from which further novelty is
discovered and incorporated into the propagating organisation" The work examines Italian industrial
clusters (ICs). ICs are geographic concentrations of interconnected companies and institutions in a
particular field, encompassing an array of linked industries and other entities important to
competition. Individual firms tend to be flexibly specialised in a particular production phase, through
relationships of both competition and co-operation. The work takes as unit of analysis self-contained
socio-economic areas known as travel-to-work areas (self-contained areas of home-work commuting),
classified in a range spanning from ICs to low industrial activity areas. A statistical analysis based
on complexity theory frameworks shows: 1. ICs are closer to self organised critical (SOC) systems than
other types of geographic agglomerates. SOC is explored in terms of nodal properties (rank-size rule) 2.
ICs seem to explore the space of diversity within the envelope of their extended value chain maintaining
the 'propagating organisation' at a higher rate than non ICs. 3. The product of internal diversity
(industrial species) times internal connectivity is higher for a community of autonomous agents than for
an aggregate. This work defines diversity as composed of three elements: variety (roughly number of
categories necessary to classify species), disparity (distance between categories) and balance
(apportionment of species per category) and applies distance metrics ideas to measure diversity. The
propagating organisation is measured by a proxy, that is the internal connectivity of the system. This
work arrives at the following conclusions: 1. The frameworks of complexity theory used in this work,
namely self-organised criticality and exploration of diversity at the subcritical-supracritical
boundary, are useful to interpret socio-economic structural and dynamical properties of geographic
agglomerations of firms. 2. SOC and diversity can be used as tools to discriminate between (industrial)
aggregate and system of autonomous agents, thereby suggesting a phase transition between the two. 3. The
structure of interdependence among agents (organisations) in a local agglomerate is related to the
closeness to a power law (SOC behaviour) of the local aggregate's structural and behavioural properties,
thereby confirming Kauffman's intuition. 4. The preliminary exploration of the structural properties of
Italian industrial agglomerations confirms some aspects of Kauffman's fourth laws, specifically the fact
that communities of autonomous agents explore the phase space of diversity (adjacent possible) at a rate
superior to that of aggregates of autonomous agents. 

 Takeshi Arai - Tokyo
University of Science

A CA Based Two-Stage Model of Land Use Dynamics in Urban Fringe Area. 

 Arturo
H. Ariño - University of Navarra

Optimal Sampling For Complexity In Soil Ecosystems

 Soil ecosystems are inherently complex: space, time and biological diversity interact giving way to
emergence of dynamic, complex features such as distribution patterns, abundance profiles, nutrient
paths, etc. Specifically, sampling for soil diversity is fraught with problems that arise form the very
different spatial scales that involve biological populations' aggregates and subpopulations. Typical
sampling techniques tend either to be ineffective for complexity assessment (i.e. too small to capture a
representative subset of most populations and their distributions) or overshot their target with very
large samples that can be cost-ineffective. Optimized sampling techniques that use the species-area
curves may be inadequate for the purpose of measuring diversity, as they typically focus on the species
accumulation rather than on the measurement of structure. Also, species-area curves are sensitive to the
accumulation mechanism: the order in which subsamples accumulate matters. We propose an algorithmic
method that tries to capture enough data for a cost-effective diversity (complexity) assessment while
statistically ensuring consistency. Tests have been done with actual, species-level soil mesofauna fauna
data. A C program implements the algorithm. 

 Rajkumar Arumugam -
Department of Electrical & Computer Engineering and Computer Science University of Cincinnati

Intelligent Broadcast in Random Large-Scale Sensor Networks Rajkumar Arumugam, Vinod Subramanian, Ali A.
Minai

 With advances in miniaturization, wireless communication, and the theory of self-organizing systems, it
has become possible to consider scenarios where a very large number of networkable sensors are deployed
randomly over an extended environment and organize themselves into a network. Such networks --- which we
term large-scale sensor networks (LSSN's) --- can be useful in many situations, including military
surveillance, environmental monitoring, disaster relief, etc. The idea is that, by deploying a LSSN, an
extended environment can be rendered observable for an external user (e.g., amonitoring station) or for
users within the system (e.g., persons walking around with palm-sized devices). Unlike custom-designed
networks, these randomly deployed networks need no pre-design and configure themselves through a process
of self-organization. The sensor nodes themselves are typically anonymous, and information is addressed
by location or attribute rather than by node ID. This approach provides several advantages, including:
1) Scalability; 2) Robustness; 3) Flexibility; 4) Expandability; and5) Versatility. Indeed, this
abstraction is implicit in such ideas as smart paint, smart dust, and smart matter. The purpose of our
research is to explore how a system comprising a very large number of randomly distributed nodes can
organize itself to communicate information between designated geographical locations. To keep the system
realistic, we assume that each node has only limited reliability, energy resources, wireless
communication capabilities, and computational capacity. Thus, direct long-range communication between
nodes is not possible, and most messaging involves a large number of ``hops''between neighboring nodes.
In particular, we are interested in obtaining reliable communication at the system level from simple,
unreliable nodes. Wireless networks that operate without fixed infrastructure are called ad-hoc
networks, and are a very active focus of research by the wireless community. However, most of the work
focuses on networks with tens or hundreds of nodes, where most message paths are only a few hops long.
All data messages in such a system are unicast, i.e., they are between specific pairs of nodes. There
are two major formulations for this. In some message routing algorithms, a path discovery process is
used to first find a route between the source and destination nodes (or locations), and the message is
then sent along this path. This is clearly a top-down approach with limited scalability. Other routing
protocols use next-hop routing, where each node, knowing the destination of an incoming message, only
determines the next node to forward the message to. These protocols scale much better, but at the cost
of maintaining and updating extensive amounts of information about network topology. This can be
expensive in terms of energy, and can often lead to problems if the individual nodes are unreliable,
causing broken links and lost messages. From a complex systems viewpoint, the problem with unicast-based
next-hop methods is that they do not exploit the inherent parallellism of the system to achieve
robustness. This is the issue we consider in our research. Rather than using directed unicast between
nodes, we study the possibilities of broadcast. In the simplest case, this corresponds to flooding,
where every message received by a non-destination node is ``flooded'' to all the node's neighbors. While
this is a simple apprach, it is extremely wasteful of bandwidth and creates a lot of collisions --- the
simultaneous use of the wireless channel by multiple messages, all of which are lost as a consequence.
To overcome the problems of flooding while retaining its inherent parallellism, we explore the method of
intelligent broadcast. In this approach, each node receiving a message decides whether to re-broadcast
it to all its neighbors or to ignore it. Note that the decision does not involve selecting which
neighbor the message is forwarded to, but only whether to forward the message. The latter is a much
simpler decision, and can be made on the basis of the information carried by the message incombination
with that available within the potential forwarding node. This approach leads to a self-organized
communication process where local decisions by the nodes produce global availability of information. In
the paper, we present a well-developed paradigm for random LSSN's, including a model for the nodes and
viable broadcast-based protocols for channel access and network organization. We evaluate the
performance of the network in the case of simple flooding, and then study the effect of a simple
decision heuristic that allows nodes to limit messagere-broadcast based on how many hops the message has
already travelled. We show that this heuristic leads to a dramatic improvement in performance, making
the broadcast-based system a viable --- and more robust --- alternative to more complicated systems
under some conditions. We also characterize how network parameters such as size, node density, messaging
rate and node reliability affect the performance of the heuristic. 

 Edward A Bach
- Boston University

SIMP/STEP: A Platform for Fine-Grained Lattice Computing.

 SIMP/STEP, is a platform for fine-grained lattice computing such as that of cellular automata (CA),
lattice gases/partitioned CA (LG/PCA), and pixel-level image processing (IP) operators. The SIMP
programming environment targets the needs of complexity experimenters, physical modelers, and IP
programmers who want to quickly write efficient and readable massively-parallel programs without
worrying about the underlying implementation. The STEP abstract machine interface is a set of
fine-grained lattice computing primitives into which SIMP programs are compiled. Through some specific
examples, we demonstrate how to program CA and LG/PCA in SIMP. We also describe a few complex-system
modeling approaches such as using LG/PCA to make an invertible global dynamics out of an invertible
local dynamics and using a fine-grained, discrete, microscopic dynamics to synthesize a system that
obeys some macroscopic continuous differential equation. Finally, we highlight some implementation and
performance aspects of STEP. In particular, we discuss PC-STEP, a software STEP kernel and a STEP
hardware accelerator design styled after the CAM-8 of Toffoli and Margolus. 


Alan Baker - Xavier University

Philosophy and Complexity

 Complexity theory has largely ignored – and been ignored by mainstream philosophy. This is
unfortunate and also surprising, for both fields are by nature wide-ranging in their score and
interdisciplinary in their impact. My aim in this paper is to sketch some possible paths for fruitful
interaction between these two fields. I shall argue that each has conceptual tools which are of
potential benefit to the other. For the purposes of this paper I shall concentrate primarily on the
philosophy of science, since it is here that the most immediate connections with complexity theory are
to be found. Contemporary analytic philosophy has largely abandoned to the natural sciences the detailed
work of explaining and predicting physical phenomena (a task which used to fall under the heading of
natural philosophy). This is especially true for physical phenomena that are complex. Identifying,
experimenting upon, analyzing, and modeling such phenomena is most effectively accomplished by those
with expertise in the broad range of mathematical, scientific, and social scientific fields which fall
under the broad heading of complexity theory or complexity science. Philosophy, by contrast, is a
meta-discipline, focusing not so much on the nature of the world itself as on our various ways of
engaging with the world. Philosophical expertise may play a valuable role, especially with respect to
the following two topics; (i) Complexity theory, considered as an object of study in its own right. Is
complexity theory methodologically distinct from traditional science? Is it revolutionary in the sense
articulated by Thomas Kuhn? (ii) Complexity, considered as an abstract concept. Is it a unitary concept
or a cluster of related but distinct concepts? Can complexity be adequately defined? How does it relate
to other notions such as disorder, randomness, etc.? My particular focus in what follows concerns the
methodology of theory choice in science. Philosophers of science have long been interested in the
criteria by which scientists evaluate and choose between competing theories. There are many factors
influencing these decisions which have little if anything to do with the content of the theory itself,
factors such as the reputation of the theory's authors, the place and manner of its publication, and the
likelihood of future funding. Other factors are internal to the theory itself. One important such factor
is simplicity. The preference for simple theories is often referred to as Occam's Razor. An area of
ongoing debate among philosophers of science concerns how to define simplicity. There has been a
parallel debate going on among complexity theorists concerning how to define a satisfactory notion of
complexity. There has been little, if any, feedback between these two debates, yet each has the
potential to inform the other. Below I list some of the main points I shall discuss in the body of the
paper; (a) Philosophers tend to analyze simplicity as a property of theories or hypotheses, whereas
complexity theorists typically analyze complexity as a property of phenomena. This links back to my
earlier point that philosophy is by nature a meta-discipline. (b) Philosophers distinguish between
syntactic simplicity (or elegance), which is a measure of a theory's conciseness, and ontological
simplicity (or parsimony), which is a measure of how many things, or kinds of things, a theory claims
exist. Parallel concepts of syntactic and ontological complexity may be defined. (c) Philosophers have
focused on defining comparative but non-quantitative notions of simplicity. By contrast, complexity
theorists have generally aimed to define quantitative measures of complexity. (d) What, if any, rational
justification can be provided for Occam's Razor? Why should we value simple theories, or expect them to
be more likely to be true, especially in a world apparently full of chaotic and complex phenomena? Ought
Occam's Razor to be a methodological principle of complexity theory? One of the goals of complexity
theory has been to identify general features common to complex phenomena in different contexts such as
the human brain, patterns of earthquakes, or the stock market and to identify patterns and regularities
they have in common. This is one sort of higher-level simplicity. 

 Ariel
Balter - Indiana University

Levy Flights in Climate Data

 Inovations in high frequency (> 1/day) climate variables appear to be Levy Stable random variables.
Many climate variables show strongly Gaussian behavior (such as directional wind) and many (such as
temperature) do not. However, the inovations (i.e. differences) are strongly non-Gaussian. Instead, they
can be fit extremely well byLevy Stable distributions. We have examined this phenomenon for a large
number of climate variables representing a wide range of years, geographies, climatologies and sampling
frequencies. The effect appears highly universal. We believe this effect emerges from nonlinearities
which can be modeled as multiplicative noise. By establishing the ubiquitous manifestation of this
effect we hope to provide a tool for climate modelling. Most stochastic climate models use white
Gaussian noise for inovations. Switching to Levy Stable noise will undoubtably improve the usefulness of
these models. 

 Ernest Barany - New Mexico Tech

Dynamics of Ethernet Protocol

 The critical output bandwidth needed to ensure that a CSMA/CD based LAN does not accumulate packets
unboundedly in its queues is computed. This is an emergent property of the network that follows from
decentralized microscopic laws. The result follows from the stationary distribution property of ergodic
Markov chains 

 Joana Barros - Centre for Advanced Spatial Analysis -
University College London

City of Slums: Self-Organisation Across Scales. Joana Barros and Fabiano Sobreira

 Third World cities are known for their inherent chaotic and discontinuous spatial patterns and rapid
and unorganised development process. Due to the very same characteristics, in the present paper, these
cities are seen as excellent objects for the study of complex systems. We argue that the morphological
structure of these cities can be analysed by the interplay of two different urban processes across
scales: the local process of formation of inner-city squatter settlements and the global process of
"peripherization" (typical growth process of Third World cities). The basic aim of this paper is to
analyse the interrelationship between these two processes. This issue is explored through
'City-of-slums', an agent-based model that focuses on the process of consolidation of inner-city
squatter settlements within a peripherization process. The paper presents briefly two previous studies
on these topics where the dynamics of these two urban processes are examined as two isolated complex
systems through heuristic agent-based models and their morphologies are discussed. We then combine
aspects of these two dynamics to compose City-of-slums, in an attempt to discuss the role of
self-organisation in the spatial dynamics of Third World cities. It is suggested that the resulting
urban morphology, although related to distinct scales, present similar degree of fragmentation (fractal
pattern). Preliminary observations also suggest that these complex processes are involved in a spatial
logic in which resistance is the cause, consolidation is the process and fragmentation is the result. 

 Jacob Beal - MIT AI Lab

Themes: Emergence, Self-Organization; System Categories: Psychological, Engineered

In a distributed model of intelligence, peer components need to communicate with one another. I present
a system which enables two agents connected by a thick twisted bundle of wires to bootstrap a simple
communication system from observations of a shared environment. The agents learn a large vocabulary of
symbols, as well as inflections on those symbols which allow thematic role-frames to be transmitted.
Language acquisition time is rapid and linear in the number of symbols and inflections. The final
communication system is robust and performance degrades gradually in the face of problems. 

 Pierre Sener - Iridia

The Connections Between the Frustrated Chaos and the Intermittency Chaos in Small Hopfield Network.
Hugues Bersini and Pierre Sener

 Frustrated chaos is a dynamical regime which appears in a network when the global structure is such
that local connectivity patterns responsible for stable oscillatory behaviours are intertwined, leading
to mutually competing attractors and unpredictable itinerancy among brief appearance of these
attractors. In this paper, through a detailed study of the bifurcation diagram given for some connection
weights, we will show that this frustrated chaos belongs to the family of intermittency type of chaos.
The transition to chaos is a critical one, and all along the bifurcation diagram, in any chaotic window,
the duration of the intermittent cycles, between two chaotic bursts, grows as an invert ratio of the
connection weight. We will more specifically show that anywhere in the bifurcation diagram, a chaotic
window always lies between two oscillatory regimes, and that the resulting chaos is a merging of, among
others, the cycles at both ends. Since in our study, the bifurcation diagram concerns the same
connection weights responsible for the learning mechanism of the Hopfield network, we will discuss the
relations existing between bifurcation, learning and control of chaos. We will show that, in some cases,
the addition of a slower Hebbian learning mechanism onto the Hopfield networks makes the resulting
global dynamics to drive the network into a stable oscillatory regime, through a succession of
intermittent and quasiperiodic regimes. 

 Howard A. Blair - Syracuse University

Unifying Discrete and Continuous Dynamical Systems

 We analyze the structure of dynamical systems (DS) in such a way as to reveal a structural parameter
that differentiates among the kinds of DSs that frequently arise. This structural parameter is present
in each of four components of nearly all DSs, including those that arise in the context of quantum
computing. The four components are (1) the computation space, (2) the temporal structure, (3) the local
state space, and (4) the differential state space. Each of these four components can be varied
independently, and can be chosen to have either a discrete structure in the sense of data-structures, or
a continuous structure in a strong topological sense. There are a variety of product operations that
enable the crafting of hybrid structures. The desired structure for each of the DS's components is
obtained in a principled fashion by specifying the structural parameter: a field of filter families that
satisfy certain closure properties to yield a so-called convergence space. The field of filter families
is to a convergence space as a topology is to a topological space. The notion of a convergence space is
stronger than that of a topological space. The virtue of the convergence space notion is that directed
graphs are convergence spaces, and continuity specializes in the case of directed graphs to graph
homomorphisms.The convergence space components of the DS can be built out of limits of directed graphs.
In particular this provides a natural way to combine discrete data-structures with topological spaces.
The trajectories of familiar DSs are the continuous solutions of constraints on the convergence space
components of those systems. An ordinary differential equation in several variables, or in countably
infinitely many variables, as in a Fermi-Pasta-Ulam continuous-valued cellular automaton, is a DS that
fits cleanly into our analysis, once it is realized that the phase space is not the computation space,
rather the phase space is the (implied global) state space. The computation space is the underlying
cellular automaton structure. One novel aspect is that various approximating solutions, as would be
given by Euler and Runge-Kutta methods, are obtained by altering the field of filter families on the
temporal structure. The utility of the convergence space analysis provides for efficiently describing
combinations of, for example, cellular automata in which the state of a cell C evolves in discrete time
steps while cells in the neighborhood of C evolves differentially in continuous time in a manner
dependent in part on the evolution of the state of C. In the end, the distinction between, say, an ODE
and a Turing machine, comes down to the field of filter families on the four components of the systems.
REFERENCES 1. Fermi, E., J. Pasta and S. Ulam, Studies of Nonlinear Problems, S. Ulam, Sets, Numbers and
Universes, MIT Press: Cambridge, 1974, pp. 491-501 2. Heckmann, R. A Non-topological View of dcpo's as
Convergence Spaces. First Irish Conference on the Mathematical Foundations of Computer Science and
Information Technology (MFCSIT 2000) July, 2000. 

 Janine Bolliger - Swiss
Federal Research Institute

A Case Study for Self-Organized Criticality in Landscape Ecology Janine Bolliger and Julien C. Sprott

 In ecology, the phenomenon of self-organized criticality may provide a powerful approach to complete
current theoretical frameworks (e.g., metapopulation theory) with a profound understanding of how
ecological feedback, interaction, and historical coincidence act together so that biotic units co-occur
at their present locations. This study investigates the self-organized critical state and the complexity
of the historical landscape of southern Wisconsin (60,000 km2). The landscape was classified into 27
discrete forest types using statistical cluster analysis. The data for classification was derived from
the United States General Land Office Surveys that were conducted during the 19th century prior to
Euro-American settlement. We applied a two-dimensional cellular automaton model with a single adjustable
parameter. The model evolves by replacing a cell that dies out at random times by a cell chosen randomly
within a circular radius r (neighborhood), where r takes values between 1 (local) and 10 units
(regional). Cluster probability measures the degree of organization. Good agreement is found when
comparing the simulated to the observed landscape using fractal dimension (spatial), fluctuations in
cluster probability (temporal), and algorithmic complexity (interaction characteristics). All results
are robust to a variety of perturbations. In our example, the self-organized state is scale-invariant in
both time and space and depends on the neighborhood size chosen for the model runs. Small neighborhoods
(r = 1 or 4 cells), representing low connectivity across landscapes, over-organize. Adjacent cells are
likely to exhibit similar properties and are thus likelier to organize, however, too small fractions of
the overall forest-type diversity are accounted for. Large neighborhoods (r = 10 or 314 cells)
representing high connection among the forest types do not organize, indicating that the likelihood of
cells interacting with cells exhibiting similar properties in large neighborhoods are rare. Intermediate
neighborhood sizes (r = 3 or 28 cells) representing intermediate levels of connectivity in forests,
where many local, but some longer distance interactions occur, give rise to self-organization to the
level of the observed forest landscape. Such ‘small-world’ phenomena studied by Watts and
Strogatz (1998) have been observed for many systems that typically range somewhere between regular and
random. We view the self-organized critical state as a measure of connectivity of the forest landscape,
since the functional significance of scale invariance is, among others, a description of how system
elements interact across the system. With this example of self-organized criticality, we show that
simple models may suffice to replicate the forest landscapes originating from complex spatial and
temporal interactions. 

 Eric Bonabeau - Icosystem Corporation

Co-Evolving Business Models

 Trying to predict the future structure of an industry is a key ingredient when it comes to defining a
company's strategy. Scenario planning is a popular method to aggregate the knowledge of industry
specialists into possible scenarios using special brainstorming sessions. Another route consists of
putting that knowledge into a model of the industry and let the industrys players co-evolve their
business strategies or business models. Existing industry knowledge is first transformed into relevant
strategic building blocks (value proposition components, operational components, revenue components)
that serve as the basic units for evolutionary recombination. The initial state of the simulation
reflects the current industry structure with respect to both business models and market shares. The
weakest players disappear and are replaced by new players that borrow building blocks from the strongest
players. After a number of generations, the industry may or may not converge toward a stationary
structure. This approach has been applied the evolution of the Internet Service Provider (ISP) industry.
In this context, a number of runs showed the emergence and later disappearance of the free ISP business
model, reflecting the actual dynamics of the ISP industry in Europe. In order to obtain reliable results
regarding the industry's stationary structure, one thousand co-evolutionary simulations were run for 500
generations. For each simulation, the endpoint population was analysed to try to answer the following
questions: are there stationary industry states, stationary business models, what are the winning
business models, what are the most likely endpoints for the industry? Surprisingly, in most runs the ISP
industry converged to a monochromatic industry structure, that is, one where all business models are
similar. Furthermore, 60% of the runs converged toward the same color, that is, the same business model,
suggesting that the industry has a very robust evolutionary attractor. Many more statistical
measurements can be extracted out of the simulations, allowing us to understand how and why the industry
converged toward a particular structure. This outcome can then be used to define an ISP's strategy. 

 Jose M. Borreguero - Center For Polymer Studies, Boston University

Fluctuation Analysis in the Transition State Ensemble of the SH3 Domain

 We perform a detailed analysis of the thermodynamics and folding kinetics of the SH3 domain fold with
discrete molecular dynamic simulations. We propose a protein model that reproduces the cooperative
folding transition experimentally observed in globular proteins. We use our model to study the
transition state ensemble (TSE) of SH3 fold proteins --- specifically, we study a set of unstable
conformations that fold to the protein native state with a probability close to 1/2. We analyze the
participation of each secondary structure element in the formation of the TSE and we find good agreement
with xperimental results of Src SH3 domain and alpha-Spectrin SH3 domain proteins. We also identify the
folding nucleus of the SH3 --- a set of specific amino acid contacts that determine whether a
conformation belonging to the TSE will fold. We predict that a set of contacts between the secondary
structure elements RT--loop and distal hairpinare the critical folding nucleus of the SH3 fold, and we
propose a hypothesis that explains this result. 

 Paul Box - Utah State University

A Computational Environment for Studying Human-Environment Interaction in Bering Sea Fisheries

 The southern Bering Sea is considered to be one of the most important fisheries in the world. Salmon
populations and salmon catches have been fluctuating in recent years, precipitating policy initiatives
that often do not consider long-term dynamics of populations, indigenous knowledge of use of the
population, nor importance of theresource to viability of local communities. There is archaeological
evidence that suggests that there have been massive fluctuations in salmon populations over the last
5000 years, and catch data show great changes in abundance over the last 150 years. The relationship
between salmon populations and the human communities that depend on them seems to be a complex adaptive
system, with many non-linear interactions between the human and salmon populations, and the environment
in which they live (including lake and river dynamics, climate, predator populations, etc.). An
agent-based computational framework is described here that specifies properties of salmon andhuman
populations, together with coupled limnological, food-web, and climate models that affect salmon
populations. 

 Raymond Trevor Bradley

 Using the concepts of energy and information, and the principles of energy conservation, holographic
organization, self-organization, and cooperation, I have collaborated with neuropsychologist, Dr. Karl
Pribram, to build a theory of communication that accounts for the endogenous processes by which stable
organization emerges in bounded social systems ("Communication and Stability in Social Collectives," R.
T. Bradley and K. H. Pribram, J. of Social and Evolutionary Systems, Vol. 21 (1): 29-81, 1998). In other
joint work, Dr. Pribram and I have mapped empirical commonalities in the neural organization in brains,
the organization of interaction in psychosocial development, and the organization of collaborative
action in social collectives (K. H. Pribram and R. T. Bradley, "The Brain the Me and the I," in
Self-Awareness: It Naure and Development, M. Ferrari and R. Sternberg (eds), Chapter Ten, pp. 273-307,
New York, The Guilford Press, 1998. 

 Alfred Brandstein - US Marine Corps

The Role of Analysis in the Brave New World

 Recent advances in science, especially computer science, have made dramatic changes in our approach to
the role of analysis. Our faith in the capabilities of modeling to support decision making has been
severely shaken. In the complex world of non-linearities and co-evolution, we are struggling to find the
proper role for modeling and analysis. The approach we are taking to define this role is discussed. 

 Michael Bretz - Dept. of Physics, Univ. of Michigan

Emergent Probability Lonergan's Genetic Model of Knowledge Growth, Development and Decline.

 A little known but intriguing heuristic model of knowledge growth and structural change was conceived
many decades ago[1]. In his treatise, Lonergan successfully disentangled the dynamic elements
surrounding the scientific intellectual process and modeled how explanatory knowledge is generated. He
extended this model to knowledge growth itself and to the dynamics underlying all development - be it
chemical, evolutionary, historical, environmental, economical, psychological, organizational or ethical.
Lonergan characterized generic growth as the successive appearance of conditioned Recurrent Schemes(RS),
each of which come into existence probabilistically once all required prior conditions (selected earlier
schemes) are in place. When formed, a new dynamic recurrent scheme becomes locked into long term
stability (some examples of RS networks are resource cycles, motor skills and habits). He envisioned the
overall concrete growth process of recurrent schemes to be highly dynamic, convoluted, non-linear and
genetic in form, so appropriately designated it "Emergent Probability" (EP). Although developed
qualitatively, EP constitutes a complex dynamic system that is ripe for computer exploration. In this
talk I will present first results from a MATLAB toy model for state space growth and evolution of EP as
simulated by a scale-free, directed growing network (nodes as RS?s, links as conditions). The appearance
of RS clusters and their interplay with each other, competition for scare resources, and dependence of
clusters on the underlying ecological situation will be emphasized. Aspects of EP appear to have been
reinvented as key elements in present day hypercycle, neuronal group and bioinformatics models, making
EP a potential vantage point for unification between, and fertilization among, the disparate
calculational approaches and interdisciplinary fields (as mentioned above). Lonergan?s stated global EP
features extend beyond the properties usually attributed to complex genetic systems, so central
questions must be addressed in the further study of Emergent Probability. 1.Insight, A Study of Human
Understanding by Bernard J. F. Lonergan (Longman, Greens & Co., London, 1957; Collected Works(3), edit.
F.E.Crowe and R.M. Doran, Toronto Press, 1992) 

 Colby Brown - Sociology,
Wesleyan University

A Graph-Dynamical Model of Transportation Development

 Charles Horton Cooley described conversation and transportation as subsets of the larger social
phenomenon of general communication. (Cooley, 1894) Cooley's classification system seems particularly
useful when Shannon's information theoretic-concept of uncertainty and the institutional economic
concept of transactions costs are compared. (Shannon, 1949, and perhaps Pitelis, 1993) Transportation
and social networks also share the potential application of methodologies based in graph theory. Noting
these precedents, we suggest that transportation development can be analyzed as an iterative graph
dynamical process in which changes in a society's structural networks effect changes in that society's
transportation infrastructure, effecting changes in social structure, etc. Within this model, the
American urban expressway development process of the mid-20th century presents an interesting puzzle. If
the social networks most relevant to interstate highway development were locally limited to the urban
areas in which development occurred, then the well-documented destructive effects of development appear
paradoxical. Theories of power or asymmetrical information might explain this state of affairs, yet, it
is more likely that the pattern of development of urban expressways depended heavily upon linkages to
dispersed (state- and national-level) institutions and actors. In that case the behavior of either
"articulation points" or "articulation groups," (i.e. groups of people who, if removed, would leave a
network cut into two or more disconnected cliques), would have been crucial to the exact pattern of
development experienced in a given locality. We examine mid-century changes in the structure of urban
social networks in America, socio-economic phenomena governing transportation development through that
time period, and construct a multi-layer graph dynamical model of transportation development based on
historical and statistical evidence. In the process, we consider the emergence of what Manuel Castells
(1983) has called "information age" urban social networks; a self-organized critical phase transition
from locally-based communications to more dispersed, though not necessarily less communal, relations. 

 Mark Burgin - UCLA

Levels of System Function Description: From Algorithm to Program to Technology

One of the most important problems of system theory is to give an efficient model for an adequate
description of system functioning. Dynamic representations become more and more important. There are
three main types of such descriptions: a contracted description as transitions of system states, e.g., a
trajectory in a state space, an extended description as the structure of operations performed by the
system, e.g., an algorithm, and a full description as a process. An extended representation of system
functioning combines advantages of both contracted and full representations. Like a contracted
representation, an extended representation is enough tractable and compact, while like a full
representation an extended representation gives a sufficiently detailed description of the process. As a
result, an extended representation eliminates scarcity of information in a contracted representation and
abundance of information in a full representation. In each type of descriptions of system functioning,
there are several levels. An extended description has three levels: algorithms/procedures, programs, and
technologies. An algorithm/procedure gives purely structural description of a process. A program gives
an internal with respect to the system under consideration description of a process. A technology gives
an external with respect to the system under consideration description of a process. Such general
understanding shows that programs are related not only to computers, while technology is used not only
in industry. It is possible to consider programs and technologies for an arbitrary system. Companies
have programs for their development. Physicists develop technologies for physical experiments.
Politicians and political scientists discuss political technologies and so on. Within this approach any
program is an algorithm or a procedure, while any technology is a program. However, it is necessary to
remark that all these concepts are relative and depend on the system under consideration: what is an
algorithm for one system may be a program or even a technology for another system. There is a developed
theory of algorithms. There are some components of the theory of programs, for example, theory of
programming languages. The situation is much worse with technology. There is no even a generally
accepted definition of technology. In spite of being one of the central phenomena of the modern
civilization, the concept 'technology' is a difficult to define. That is why the main emphasis of this
work is on technology and its mathematical theory. There are many mathematical models for different
technological processes. However, being useful, such models reflect only some parts of the whole
process. Complete multifaceted models of technological processes are developed in the mathematical
theory of technology. The development of technological knowledge and advancement of mathematics made it
possible to elaborate the mathematical theory of technology (Burgin, 1997). Its starting point is an
exact, however, informal definition of technology as a specific system of knowledge that describes how
different systems are produced, utilized or function. Basing on this definition, two classes of
technologies (the general and specific technologies) are introduced to reflect the situation existing in
industry and engineering. This provides for the construction of a general mathematical model of a
specific technology as well as for the development of a relevant mathematical apparatus and exact
methods for an investigation and design of various technologies (in industry, management, information
processing and so on). In the context of systems, which are created and used by people, the basic
concept of the theory, a technological operator reflects operations perfgd by people and machines, as
well as natural processes that are included in a technological process. For example, biotechnologies
utilize biological processes, while chemical technologies are based on chemical processes. In this
context, a general technology is a system of knowledge about a class of specific technologies, e.g.,
biotechnologies or information technology. The mathematical theory of technology utilizes new
mathematical disciplines such as theory of named sets, fuzzy set theory, and theory of structured
multidimensional models of systems and processes as well as traditional fields such as algebra, theory
of probabilities, and theory of algorithms. In the mathematical theory of technology, such problems as
stability, reliability, equivalence, constructibility, and realizability of technologies are studied.
The aim is the development of efficient methods and algorithms for design and improvement of different
technologies. 

 Galina Bushueva- Institute of Eye Diseases and Tissue
Therapy, Odessa, Ukraine

Psychology-Physiological Approach for the Analysis of the states

Pathology. Galina Bushueva, Nadia Kabachi and Arnold Kiv

 All pathologies are complex disturbances of physiological processes and deep changes in their
psychological state. In our presentation we would like to put forward a conception of complex
diagnostics of different diseases based on analysis of patients' physiological and psychological
characteristics. In [1] it was found correlations between the states of people cardio-vegetative system
and their creative thinking characteristics. These correlations are caused in particular by disturbances
of relaxation time of spasmodic state (RTSS) of organism. The results obtained in [1] opened a
possibility to study relaxation processes in the vegetative nervous system using a special device
(computer pupillograph), which gives quantitative characteristics of accommodation - convergence
pupillary (ACP) system of eye. In this work we studied patients with symptoms of disturbances in
vegetative nervous system. We investigated RTSS by measurement ACP system of eye at different stages
from beginning until the end of illness. At the same time we performed computer testing of creative
thinking. The method of testing programs development is described for example in [2].By computer
pupillograph we obtained data about pupillary size, reaction to illumination, accommodation and other
parameters of eye. We measured visual acuity, reserves of accommodation and the pupillary sizes at
convergence. Measurements of the vegetative nervous system tonus were fulfilled by method described in
[3]. Parameters of creative thinking, which were measured, are: I (Intuition), L (Logic), VTS (Volume of
thinking space). It was found in this study that in the process of treatment of patients an improvement
of creative thinking parameters mostly forestalls an appearance of usual signs of recovery, which are
used in medical practices. So it may be concluded that a complex psychology - physiological diagnostics
allows to define more precisely the end of recovery and the necessary time of treatment of patients for
an illness. We continued our research by trying to exploit the data obtained previously (results) to
predict the patients psychological behavior after their cure. For that, we are using generally the Data
Mining techniques [4] and in particular the Bayesian networks (probabilistic networks) which combine
probability theory with a graphical representation of domain models. The first results obtained we
seemed promising. The results can be used as an element in decisions of organizational and economic
problems in public health. REFERENCES 1. N. Bushueva, A. Kiv, V. Orischenko, E. Ushan, G. Finn and Sue
Holmes. Psychological limitation of creative activity. Computer Modeling&New Technologies 3 (1999)
135-137 2. A. E. Kiv, V.G.Orischenko et al. Computer Modeling of Learning Organization. In Advances in
Agile Manufacturing. Eds.P.T.Kidd, W.Karwowski. Amsterdam (1994) 553-556 3. A.M. Kluev. The State of
Vegetative Nervous System of People with Accommodation Spasm. Journ. of Ophtalm. 6 (1976) 443-45 4. N.
Kabachi, S. Levionnois, A. Happe, F. Le Duff, M. Bremond. Data Mining Techniques for Patient Care
Pathway. In Third International Conference on Data Mining Methods and Databases for Engineering, Finance
and Other Fields, 25 - 27 September 2002 Bologna, Italy (subdued). 

 Val Bykoski
- Boston University

Complex Models and Model-Building Automation

 A model-building framework (MBF) is proposed to automate building complex models. The framework is
designed to combine dynamically model components, analyze dependencies, construct a model code, build
it, connect the model to the input and output data, run the model, evaluate it, and, if necessary,
modify it. A construction environment and its control and configuration logic, a basic component of the
proposed framework, isdescribed. A generic model-building logic is presented and discussed as well as
building techniques and tools. Complex models such as business/enterprise models, geographical sites
such as cities, traffic systems (including air traffic), bioinformatics gene expression models, and
similar ones contain hundreds parameters, adaptive components, interdependencies. Initially, the
framework is a highly generic medium/space whichis being shaped and customized by the data into a
specific medium/space reflecting a structure and dynamics built into the data. The goal of a
model-building activity is to generalize the original data into equivalent ³model form² and to use the
model then to generate (³predict²) new data for a new context. This in fact is the goal of any
scientific framework, a ³theory². The architecture of a model used to be fixed, that is, equations,
functions involved (such as propagation functions, nonlinearity elements), dependencies, network
architecture, kernels, etc. Also, building a model used to be separated from the model running,
evaluation, and model modifying phase, a sort of offline data/model exploration pipeline. The MBF makes
those elements and phases highly dynamic. Construction tools² are necessary since complex models cannot
be hand-built. Since they cannot be solved ³analytically², in terms of symbols, the data-driven
technology to build a model is the only choice, indeed. The first step is a generic model prototype with
highly flexible/adaptive structural and functional components. It could be, for example, a cellular
neural networks (NN). Data as well as certain fundamental criteria/constraints are used to incrementally
build and customize the prototype making it a data container, a generating database which can absorb new
data and also to generate new data for new contexts. The data used to drive the model-building process
have to have a format ofinput-output (or if/then) pairs. So, the model-building process is ³supervised²
or driven by the output data. As an example, the NN models structure, their building rules,
interdependencies are discussed as well as integration of the (training) data into the framework. What
makes the MBF approach different is that the model is being built ³on-the-fly² based on a
meta-description, and therefore various models can be built ³to order² as a highly dedicated tool. The
MBF may as well generate the testing, visualization, and other appropriate tools for that model. A
prototype of a MBF is designed and developed, and versions of NN models have been generated
automatically using a generic core. Results will be presented and discussed. 

 Elena Bystrova - PhD student; Saint-Petersburg State University, Faculty of
Biology and Soil Sciences, Department of Biophysics

Fungal Colony Patterning as an Example of Biological Self-Organization. Elena Bystrova, Eugenia
Bogomolova, Anton Bulianitsa, Ludmila Panina, Vladimir Kurochkin

 One of the interesting examples of biological self-organization is the formation of different
spatiotemporal patterns in colonies of microorganisms such as bacteria, myxomycete Dyctiostelium
discoideum and fungi. We consider four main types of stationary dissipative structures, which can arise
in colonies of mycelial fungus Ulocladium chartarum - periodic rings (zones), sparse and dense "lawn"
(continuous mycelial growth), ramified (fractal-like) structures. We investigate conditions required in
order for patterns to appear and explain how changes in fungal environment influence the morphology of
the colony, or, in other words, how the system interacts with its environment. The experimental data
obtained enable us to construct a morphological diagram which demonstrates the morphological change due
to environmental conditions (in our case, we varied two parameters - nutrient concentration and agar
medium thickness). It has been suggested to consider fungal colony development as being determined by
two simultaneous processes - consumption of substrate (activator) and production of diffusible
metabolites (growth inhibitors). On the basis of proposed mechanism we developed a mathematical model
for description of observed non-linear phenomena in colonies of mycelial fungi. In computational
experiments we have revealed ranges of parameters in which the formation of zones and continuous
mycelial growth occur. In particular, metabolite diffusion coefficient was shown to be one of the main
parameters defining the colony morphology. By means of the model and experimental results we have
estimated the value of the given coefficient. 

 Raffaele Calabretta -
Institute of Cognitive Sciences and Technologies National Research Council, Rome, Italy

An Evo-Devo Approach to Modularity of Mind

 A module can be defined as a specialized, encapsulated organ that has evolved to handle specific
information types of enormous relevance to the species. Given their extremely general nature modular
systems can be found in different aspects and levels of reality, from genomes to nervous systems, from
cognitive and linguistic systems to social systems, and they are an object of study for a variety of
scientific disciplines, from genetics and developmental biology to evolutionary biology, from the
cognitive sciences to the neuro-sciences. Modules can emerge during the life of an individual organism
as part of the biological development of the individual. Biological development is controlled by both
genetically inherited information and the environment in which the individual develops. The two sources
of information do not simply add linearly but there is a strong nonlinear interaction between them.
Genetic expression is environmentally constrained in that the external environment (even in utero) and
the already produced phenotype at any stage of development have a crucial role in determining the
further developmental stages of the phenotype. And, conversely, learning, experience, and the
environment external to the organism's body can influence the phenotype only in ways that are, more
broadly or more stringently in different cases, genetically fixed. In any case it is clear that the
modular structure of the brain is strongly influenced by the species-specific information contained in
the DNA of all individuals of the same species. This poses two important research questions: How
information relating to the modular structure of the brain is encoded in the genotype? How is this
genetically encoded information acquired during an evolutionary process taking place in a succession of
generations of individuals? With respect to the first question we have to consider that the genotype
itself is a modular system. The DNA molecule can be analyzed as made up of modules at different levels:
bases, aminoacids, genes. If we consider genes as the genetic modules, we know that genes map in complex
ways into phenotypical traits. The mapping rarely is one-to-one but more frequently is many-to-many.
Many genes enter into the determination of a single phenotypical trait and at the same time a single
gene tends to influence many different phenotypical traits. Hence, we cannot in general assume that one
gene entirely controls the development of a single neural module but, more probably, many genes control
the development of one and the same module and one and the same gene controls the development of many
different modules. The second research question concerns the evolutionary origin of the genetic
modularity of the genotype which underlies the development of neural modules. For example, one can
hypothesize that at first a segment of the genotype controls the development of some particular neural
module; then, this segment of the genotype is duplicated as a result of some mutation; and finally, the
duplicated genetic module changes its function as a result of some further mutation and it becomes
responsible for a new neural module with a different function. For the purpose of answering the two
questions described above, it is obviously necessary to consider in the chosen model of study both the
process of organism development and organism phylogenetic history. However, it is surprising to notice
that the developmental and evolutionary integrated study of organism modularity is a quite recent
conquest even in biology. Traditionally, the process of organism development and that of phylogenetic
evolution were considered as being very different and thus have been studied separately in two different
research fields (i.e., developmental biology and evolutionary biology). In order to achieve a better
comprehension of reality, the importance of considering them at the same time it has instead been
recently pointed out (consider, for example, the recent symposium on Modularity in development and
evolution, Delmenhorst, Germany, May 11-14, 2000). The combined approach - now known as an "evo-devo"
approach - to studying the meaning of biological is very innovative and is therefore not yet widespread
in biology research, but it is showing itself to be plausible from a biological point of view and
therefore very useful from an heuristic point of view. Unfortunately also in the cognitive sciences,
development and evolution have not been considered as being two very important processes to be studied
together in order to get a better understanding of the modular nature of mind. On the contrary, they are
often considered as being two opposed processes that are incompatible for explaining the acquisition of
brain competence. On the one side, developmental psychology, while admitting the modular nature of mind,
claims that this is most of all the product of a process of development. On the other side, nativists
and evolutionary psychologists maintain the view that humans born already furnished with hardwired
cognitive modules. Without wanting to take a part in this controversy, which remains nevertheless one of
central interest in the cognitive sciences (i.e., the nature-nurture debate), we suggest a way to
reconcile different position in a new way of studying the evolution of modularity of mind, i.e., an
evo-devo approach. REFERENCES Calabretta, R., Di Ferdinando, A., Wagner, G. P. and Parisi, D. (In
press). What does it take to evolve behaviorally complex organisms? BioSystemsCalabretta, R. & Parisi,
D. (In press). Evolutionary Connectionism and Mind/Brain Modularity. In W. Callabaut & D. Rasskin-Gutman
(Eds.), Modularity. Understanding the development and evolution of complex natural systems. The MIT
Press, Cambridge, MA. [draft: doc, pdf; updated June 4, 2001] Calabretta, R., Nolfi, S., Parisi, D. and
Wagner, G. P. (2000). An artificial life model for investigating the evolution of modularity. In Y.
Bar-Yam, (ed.), Unifying Themes in Complex Systems. Perseus Books, Cambridge, MA.Di Ferdinando, A.,
Calabretta, R., and Parisi, D. (2001). Evolving modular architectures for neural networks. In R. French
& J. SougnÈ (Eds.), Proceedings of the Sixth Neural Computation and Psychology Workshop Evolution,
Learning, and Development, Springer Verlag, London. [pdf, ps, abstract]Fodor, J. (1983). The modularity
of mind. The MIT Press, Cambridge, MA. Karmiloff-Smith, A. (1999). Modularity of mind. In R. A. Wilson
and F. C. Keil (Eds.), The MIT encyclopedia of the cognitive sciences. The MIT Press, Cambridge, MA.
[html]Keil, F. C. (2000). Nativism. In R. A. Wilson and F. C. Keil (Eds.), The MIT encyclopedia of the
cognitive sciences, The MIT Press, Cambridge, MA [html]Simon, H. A. (2000). Can there be a science of
complex systems? In Y. Bar-Yam, (ed.), Unifying Themes in Complex Systems. Perseus Books, Cambridge,
MA.Wagner, G. P., Mezey, J. & Calabretta, R. (In press). Natural Selection and the origin of modules. In
W. Callabaut & D. Rasskin-Gutman (Eds.), Modularity. Understanding the development and evolution of
complex natural systems. The MIT Press, Cambridge, MA [draft: doc, pdf; updated March 13, 2001]Wallace,
A. (2002). The emerging conceptual framework of evolutionary developmental biology. Nature 415, 757-764.


 Mathieu Capcarrere - Logic Systems Laboratory. Swiss Federal Institute
of Technology, Lausanne

Emergent Computation in CA: A Matter of Visual Efficiency

 Cellular Automata as a computational tool have been the subject of interest from the computing
community for many years now. More precisely, the development of the Artificial Life field led many to
wonder on how to do computation with such tools. Artificial Evolution which gave good results on
specific tasks, like density or synchronization was often given as an answer. However, it appeared that
the limitations of such an approach were severe and really the question of WHAT meant computation with
cellular automata became pregnant. The answer to this question is far from obvious. Mitchell,
Crutchfield, Hanson et al proposed an analysis of "particles" as a partial answer. Wuensche more
recently developed the Z parameter as a paraphernalia to treating this question. Before this question
appeared in its full-blown form in the A-life/Computer scientist community,there were already
propositions going this way with Wolfram's class III and, related, Langton's computing at the edge of
chaos. In this presentation/paper, I will argue that computation of CAs is a matter of visual
efficiency. Basing our argument on past, recent and also previously unpublished results (ours and
others) mainly but not only on the density and the synchronization task, I will propose a definition of
what is computation by means of CAs. This will be the occasion to (re)define emergent behavior, in a
limited scope, but also to envisage differently the whole question of what may be sought in computing
research in CAs. The practical consequences of this approach will alter the HOW question answer, and
most notably how to evolve computing CAs. Though the talk will be centered around the density task, will
encompass a much bigger chunk of the CA field. However, the claim is NOT a redefinition of the whole
computation by means of CAs, but rather a tentative definition in the limited scope of emergent
computation. 

 Giuseppe Castagnoli - Elsag, IT Division

The Quantum Computer: A Complex System Irreducible to Classical Models

 All devices created by man to support his everyday life, from the stone age to modern engineering, are
functionally representable in classical physics. The quantum computer might make the first exception.
Its operations essentially call for a quantum mechanical representation (Mahler 2001). They fully draw
on a richness of quantum physics that vanishes in classical physics. Being a complex and purposeful
(problem-solving) system strictly based on non-classical laws, the quantum computer as a model might
enrich the vision of Complex Systems. We provide here a pedagogical presentation. By omitting
technicalities, the special way of working of the quantum computer can be explained in a conceptually
complete fashion to an interdisciplinary audience. We start by explaining the special quantum effects
used in to-day quantum computation: quantum mode superposition (through the metaphor of the
parallel-possible universes), parallel non interfering and interfering histories, the exponential
explosion of the number of modes/histories of a compound object with the number of its parts, quantum
togetherness (entanglement), quantum measurement as filtering. Pictorial aids illustrating possible
histories and their quantum transformations substitute mathematical formulation. Then we show in
conceptual detail how the above special effects yield the quantum speed-up of Shor's (1994) and Grover's
(1997) algorithms. Speed-up can mean solving in less then a second problems whose solutions by to-day
classical computers would in principle require billions of billions of years. Long-standing
computational notions become deeply altered in quantum computing, starting from the very notion of
algorithm as a procedure for dynamically constructing the solution of a problem. In quantum computation,
the implicit definition of the solution, inherent in the statement of the problem, directly determines
the solution in a non procedural way, through an extra-dynamical quantum transition. Such a transition
is jointly influenced by the initial and the final selection (Castagnoli and Finkelstein, 2001). This is
unlike classical computation which is the dynamical development of the initial selection (i.e.
condition) alone. We discuss how the current approach to quantum computation might hold a classical
vestige, being still algorithmic in character, although enriched with some special quantum effects. We
introduce at a conceptual level a non-algorithmic, non-dynamical approach to quantum computation (Jones,
2000, Castagnoli and Finkelstein, 2002). This can be based on the same quantum relaxation processes and
symmetries (due to identical particle indistinguishability) that govern the formation of atomic and
molecular structure. This latter approach naturally raises the question whether biological processes and
molecular evolution draw on the richness of the quantum level. REFERENCES: G. Castagnoli and D.R.
Finkelstein (2002) Quantum-statistical computation arXiv:quant-ph /0111120 v4 30 Jan 2002. G. Castagnoli
and D.R. Finkelstein (2001) Theory of the quantum speed-up Proc. R. Soc. Lond. A 457, 1799.L. Grover ,
(1996) In Proc. 28th A. ACM Symp. On Theory of Computing, P. 212. Philadelphia, PA: ACM Press. J. A.
Jones, V. Vedral, A. Ekert, G. Castagnoli (2000), Nature 403, 869.G. Mahler (2001) Science 292, 57.P.
Shore (1994) In Proc. 35th A. Symp. of the Foundation of Computer Science, Los Alamitos, CA, P. 124. Los
Alamitos, CA: IEEE Computer Society Press. 

 Alok Chaturvedi - Purdue
University

Synthetic Environments for Analysis and Simulations. Alok Chaturvedi and Shailendra Raj Mehta

 SEAS is a distributed, interactive, real-time synthetic economy populated with human and artificial
agents. It allows realistic representations of markets and economies at any level of detail. The
artificial agents represent decision-makers who engage in relatively non-strategic decision making, such
as consumers in large markets. Human agents, on the other hand, represent decision-makers such as firms
and governments that engage in strategic interaction, and are provided with extensive decision support
systems to make effective decisions. This combination of human and artificial agents, unique to SEAS,
allows for the creation of environments that combine complexity and realism. In addition, we
systematically collect data on costs, demands, market shares, demographics and technological trends and
calibrate SEAS to replicate the economic or management situation under consideration. On account of
these capabilities, SEAS has exciting uses in business "war gaming" and training. It is structured
around the interplay of human decisions and game events that requires active involvement of
participants. It helps participants come to a more complete understanding of the sources and motivations
underlying the decisions by placing them in the shoes of executives running the firms at different
points in time. Games dealing with current or future situations help explore the potential implications
of various courses of action, and raise important questions for further investigation. 

 H. F. Chau - University of Hong Kong

How To Avoid Fooling Around In Minority Game H. F. Chau and F. K. Chow

 Minority game (MG) [1][2] is a simple model of heterogeneous players who think inductively. It plays a
dominant role in the study of the global collective behavior in free market economy in econophysics
since it is a powerful tool to study the detailed pattern of fluctuations. In MG, there are three
important parameters: the number of players N, the number of each player’s strategies S and the
length of histories M. Maximal cooperation of players is observed in MG whenever 2^M » NS. However, is
it possible to keep optimal cooperation amongst the players for any fixed values of N, S and M? We
report a simple and elegant way to alter the complexity of each strategy in MG with fixed N, S and M so
that the system can always be locked in a global cooperative phase. Indeed, our investigation concludes
that player cooperation is the result of a suitable sampling in the available strategy space. [1] D.
Challet and Y.-C. Zhang, Physica A 246, 407 (1997). [2] Y.-C. Zhang, Europhys. News 29, 51 (1998). 

 Adrian Li Mow Ching - University College London

User and Service Interaction Dynamics. Adrian Li Mow Ching, Venus Shum and Lionel Sacks

 Active networks enable a faster deployment of services in a telecommunications network. To develop
autonomous management systems for such a network requires understanding the emergent behaviour arising
from the interactions between the underlying users and the network services. In this paper we use
complex systems modelling to perform service engineering analysis on a generalised telecommunications
service network abstracted from an ANDROID network. This work aims to develop ideas and issues regarding
service engineering for future networks. In particular, we analyse the interactions between users and
the network and study the affects of load balancing on the service platform. As a result we see a
self-limiting characteristic that prevents the load on the communications network increasing beyond its
capacity. However, the resulting effects on the users are an increase in the volatility of the response
time for each service request, which is highly undesirable. We find that load balancing has significant
improvements on resource utilisation and load management. 

 D. Chistilin -
Institute World Economy and International relation

Development and Self-Organization of Complex System. Case of World Economy

 This work is an intermediate result of a research of the process of development and self-organization
of world economy; this research was begun in 1998 and it uses a complex systematic approach (the theory
of complex systems). Some elements of the work were presented at scientific forums of NECS ( ICCS-98,
ICCS-2000). The objective of the work is to reveal and define phenomena which are characteristic to the
behavior of complex systems in the process of their development on the basis of factual material of the
world economy development during the period of 1825-2000. The following methods are used in the
research: complex-systematic analysis, historical and interdisciplinary methods and method of simulation
in verbal-logical form. Social system - "the world economy" - is considered as a complex system
consisting of two global subsystems: economic and political. Common agents for both subsystems are
national economies, which interact in economic and political spheres and form connections and structure
of social system of world economy. The principal functional purpose of social system "the world economy"
is an achievement of self-regulation of relations between the agents in economic system through
political system which results in the state of dynamic equilibrium, i.e. economic growth of world
economy. The functioning of world economy means an implementation of economic and political relations in
the process of international exchange of resources on the basis of international division of labour with
the aim of the most effective distribution of valuable resources for production of valuable benefits for
consumption. Development of social system means a process of increasing its stability under the
influence of outside environment (maintenance of stability in the given limits - homeostasis) by means
of accumulation of structural information changing the quantity of organization (effectiveness) of the
system and making its structure more complicated. Increasing of stability is expressed in accumulation
of economic effectiveness and formation of more complicated structure of society. Development of world
economy means a co-evolution of economic and political subsystem development resulted in gradual
complication of social order, which increases the system stability under environment influence, pressure
of population growth and resources limits. Development means a change of equilibrium states with
different macro-economic characteristics. Each state is expressed in structural and quantitative
characteristics. For world economy we will consider as the system of international monetary relations
(IMR) as a structural characteristic. The growth rate of the gross national product of the
countries-members of international economic relations we will consider as a quantitative characteristic.
On the basis of historical material we distinguish three structural characteristics of IMR type and,
correspondingly, three periods of time and three states of world economy (gold standard, system of
Á›åòîí-Bóäñ, Jamaika system). On the basis of statistical data we calculate the quantitative
characteristics for each state. Then we bring all data together into the table forms. On the basis of
distinguished states we build verbal-logical model of the world economy (figure). As an specimen we take
the ‹.Àâäååâ's model, which is based on the idea that phenomenon of development can be considered as a
struggle of two opposite trends - organization and disorganization. The process of development, which is
begun from the maximum of disorganization, can be described as a process of accumulation of structural
information, which is calculated as the difference between the real and maximum values of entropy. The
model allows to make the following conclusions: 1. Each consequent distinguished state of world economy
has more complicated organization of political system and international monetary relations. And this
fact demonstrates the tendency of complication of world community structure. 2. Each consequent state is
more effective from economic point of view and has the higher rate of economic growth. This allows world
economy to develop steadily in conditions of grown population of the planet and limited resources. A
tendency of growth of economic effectiveness of the whole system is observed in the long-lasting
interval of time. 3. The process of formation of consequent structures of both political and economic
organization of the world economy took place in condition of strong non-equilibrium environment resulted
in numerous military and civil conflicts and economic crises. 4. All stated above allows to conclude
that: system of the world economy possesses the property of the complex systems - self-organization. the
Îêñàíãå›à-Ï›èãîæèíà's principle of minimum of energy dissipation is realized in the process of
development. Each consequent organization of world economy produces less entropy than the previous one.
The category of energy in physical system corresponds to category of resources in social system. Thus,
principle of minimum of dissipation of limited resources acts in social systems. The model reflects
realization of this phenomenon in the process of development and self-organization of the world economy.
Direction of development of the system "world economy" is defined by the law of conservation of
accumulated effectiveness and this allows to say that the model has predicted potential for realization
of prognosis of future organization of the world economy. 

 Claudia
Ciorascu - University, Iasi, Romania

The Accuracy of Auto-Adaptive Models for Estimating Romanian Firm's Cost of Equity. Claudia Ciorascu &
Irina Manolescu

 In the paper the mutations of the Romanian firms capital structures and the relations with the cost of
equity capital are analyzed. The validity of leverage effect of capital structure to financial return is
also tested. We consider in this analysis the public data of more than one hundred Romanian firms listed
at Bucharest Stock Market and on RASDAQ, between 1997 and 2000. The initial hypotheses in this research
are: the relation of leverage effect is not validated for Romanian firms; using auto-adaptive models for
the estimation of firm’s cost of equity the results are better than the classical techniques (as
CAPM or auto-regressive models). Because of their adaptation to the specific of the input data, the
self-adapting models can be successfully used in real problems with large data sets. We are studying the
quality of this approach in the case of estimating firm’s cost of equity. This also defines the
pre-requisites for considering new instances of the problem: share valuation, investment
decision-making, etc. The presence of the determinant factors of capital structure (bankruptcy and
monitoring costs, motivation of the managers, institutional restrictions, transaction costs, taxes)
represents an argument for the validation of the first hypothesis, but their incidence is exceeded by
the low liquidity of the Romanian capital market. All this elements have opposed effects on the cost of
capital and the resulted conclusion is the impossibility to obtain an optimal capital structure,
especially on Romanian capital market. The existence on Romanian stock market of the listed firms with a
debt ratio (debt / equity capital) higher than 500%, but sometimes reaching incredible values like
10000%, raises serious questions over the admittance criterions on stock market quotation. These highly
indebt firms have descending trends in activity and profits and the questions about financing policy can
be raised both firm and financial institutes level. 

 Michael Connell -
Harvard

Neuroscience and Education--Bridging the Gap

 Many researchers, educators, and laypersons are excited about the prospect that neuroscience can
fruitfully inform educational research and practice, ultimately leading to significant improvements in
curriculum design, pedagogical techniques, and overall efficacy of educational institutions. Despite the
enormous amount of hype surrounding these issues, however, to date there have been few tangible,
rigorous results demonstrating how findings from neuroscience can actually be brought to bear on
educationally relevant problems such as mechanisms of knowledge transfer. In this presentation, I argue
that computational neuroscience offers tools that can provide a theoretical bridge for applying
neuroscience findings to educational issues. The approach I describe involves identifying key
constraints at the neurological level (i.e. that structural changes to synapses are implicated in
long-term storage of information encoded in the nervous system, whereas dynamic functional activation
patterns seem to be the basis for thought and action), incorporating these assumptions into a
computational neuroscience model (a connectionist or other kind of artificial neural network), and then
tracing the effect of these (biologically plausible) low-level constraints on higher-order patterns (at
the level of cognitive processes), attempting to filter out model characteristics and behaviors that are
mere artifacts of the model and its more ad hoc assumptions. This complex systems approach provides a
straightforward way to link two levels of analysis (neural structure and cognitive function) that is
complementary to the more common reductionist approach of building detailed models of specific phenomena
(such as language acquisition or concept formation), and may be particularly attractive at the present
time for people interested in exploring qualitative properties of higher order cognitive functions
involving neural structures that are far removed from the sensory periphery. I describe how
computational neuroscience offers a non-intuitive (and biologically informed) paradigm for understanding
human knowledge organization and conceptual structure (in terms of a semi-parametric representational
substrate), and I describe how it can shed new light on some educationally relevant issues. In
particular, I discuss how this model provides insight into the nature of representation in the mammalian
nervous system, I argue that this analysis suggests a set of meaningful conceptual categories that can
inform meta-theoretic reasoning about classes of psychological theories, and I describe the mechanism of
stimulus generalization (near transfer) that is revealed by this model. 


Emilia Conte - Polytechnic of Bari

Managing Urban Traffic Dynamics by a Multi-Agent DSS

 Complexity of urban environments is severely challenging science and technology, making Information
Technology (IT) tools strongly significant for enabling innovative and more appropriate decisions.
Concerning urban environments, the paper deals with the important issue of the traffic dynamics
proposing a tool for managing and controlling traffic air pollution, since its effects on human health
are recognized as widely relevant. The design of the proposed tool is based on the belief that IT
systems can support decision making processes reducing routine tasks thus enlarging substantive and
problem oriented decision activities, through improving man-machine interaction. Basing on a local case
study, the paper reports about a two-years research work of the authors, studying traffic problems in a
middle-sized city of Southern Italy, starting from daily monitored data about air quality conditions.
The research was aimed at developing the architecture of a Decision Support System (DSS) assisting
decision makers of municipal offices to implement strategic actions for controlling traffic air
pollution. The DSS is designed as a multi-agent system in order to replicate the single and autonomous
tasks of the decision makers, at the same time saving and strengthening the interaction among those
tasks and fuelling dialogue among different data sources and receivers. The paper describes the
development of the DSS architecture which was carried out starting from the real structure of both the
decision making process and its main activities and focusing on the use and the production of knowledge
within the process itself. In the DSS design, a special attention was given to three main tasks of the
decision process: validation of data from pollution measurement stations, assessment of measurement
stations functioning, and production of short term traffic actions. A neural network approach and an
expert system environment represent the main technological basis for the implementation of the DSS.
Further research perspectives are finally investigated in the wider dimension of medium/long term
traffic control, where strategies can benefit by the multi-agent interactive approach, which can lead to
gradually shift to different, higher, levels of organization in problem management. Some considerations
are made with regards to knowledge representation tasks within DSSs devoted to complex problem
management, to the use of machine learning as a form of knowledge representation, and to organizational
learning within the decision making environments facing air quality problems. 


Ron Cottam - Brussels Free University

Self-Organization and Complexity in Large Networked Information-processing Systems. Ron Cottam, Willy
Ranson & Roger Vounckx

 Classical analysis of large networked information-processing systems from a ³quasi-external² point of
view begins to create problems as the range of hierarchical structural scales is extended. Most
particularly, the viability of deterministic distributed control becomes questionable in extended-scale
temporally-dynamic (i.e. interesting!) networks. The ³traditional² split between ³body² and ³mind²
appears to be most particularlyrelated to our mental incapacity to relate to large systems whose
character is primarily distributed but whose characteristics collapse to those of a synchronous
deterministic network when reduced to a unified perspective. The major problem in forming such a
representation is the necessarily irrational coupling across multiple scales of a large disparate
complex organization such as the brain and our consequent inability toformulate correctly a causal tree
for the system. We investigate the implications of these difficulties beyond simply the establishment of
an upper systemic scaling limit, and relate them to a recently recorded Windows Local Area Network
browser election breakdown. 

 D. E. Creanga - Univ. Al.I. Cuza

Computational Analysis in Temporal Series Describing Kidney Excretory Function. D.E. Creanga,
E.Lozneanu, J.C.Sprott

Semi-quantitative analysis was carried out on the excretory function of human kidney. Health and
pathological kidney are investigated by means of nuclear medicine using radio-pharmaceutical technique
based on radio-isotopeTc99m and a gamma-camera device assisted by a specialized computer. The amount of
Tc99m physiological solution in every kidney is given by a temporal series computationally which we
processed using the analysis strategy based on linear and non-linear tests. Fast Fourier transform,
auto-correlation function, the portrait in the phase space and the corresponding fractal dimension
present similarities as well as differences when the health kidney is compared to the pathological one.
The histogram of probability distribution is presenting a repetitive character for the normal kidney
while for the non-functional kidney the strong asymmetry of the probability distribution is the dominant
feature of the histogram shape. Fast Fourier transform does not present significant differences nor in
the lin-log representation neither in the log-log representation but the wavelet transform seems to be
marked by some qualitative differences. The portrait in the state space reconstructed using the first
derivative of the raw data graph shows a higher dispersion of the points corresponding to the ill kidney
than for the health one (while the reconstruction using the delay coordinates appears in the same form
for both cases). The most important difference between the two cases is offered by the correlation
dimension which is significantly higher for the health kidney in comparison to the ill one. 

 Atin Das

Nonlinear Data Analysis of Experimental [EEG] data and Comparison with Theoretical [ANN] Data

 In this paper, nonlinear dynamical tools like largest Lyapunov exponents (LE), fractal dimension,
correlation dimension, pointwise correlation dimension will be employed to analyze electroencephalogram
[EEG] data obtained from healthy young subjects with eyes open and eyes closed condition with the view
to compare brain complexity under this two condition. Results of similar calculations from some earlier
works will be produced for comparison with present results. Also, a brief report on difference of
opinion among coworkers regarding such tools will be reported; particularly applicability of LE will be
reviewed. The issue of nonlinearity will be addressed by using surrogate data technique. We have
extracted another data set which represented chaotic state of the system considered in our earlier work
of mathematical modeling of artificial neural network. We further attempt to compare results to find
nature of chaos arising from such theoretical models.

Neural Net Model for Featured Word Extraction A Das, M.Marko, A. Probst

 Existing search engines have many drawbacks while situations demand more refined operations. We have
shown with examples that two of the mostpopular conventional search engines return out-of-context
results. Our group is actively engaged in developing new algorithms for this purposewhich are different
in understanding the topology of searching. There are two main independent approaches to achieve this
tasking. The first one, using the concepts of semantics, has been implemented partially.(For more
details see another paper presented by at the conference: Transforming the World Wide Web into a
Complexity-Based Semantic NetworkM.Marko, A.Probst, A.Das.) The second approach is reported in this
paper. It is a theoretical model based on using Neural Network (NN) learning features. Instead of
usingkeywords or reading mechanically words from the first few lines from papers/articles, the present
model gives emphasis on extracting 'featuredwords' from an article. We call those words as 'featured
words' that occur most frequently. Obviously we have to exclude English words like "of,the, are, so,
therefore" etc. from the list of featured word. To form such a word list, an article is read first. To
read a full paper as input tothe model would be a heavy load of computation, so a choice of the first
few hundred words can serve the purpose -also because beyond this limit,generally technical or
scientific notations appear which are not relevant for the present purpose. These words are raw data and
will be used asthe input to the model. The NN model will chose from N such words (raw data) to find M
featured words (refined data). This output is then fed tothe search engine to produce more accurate
words. Working of the proposed NN model is based on Principal Component Analysis (PCA). Another
important feature of the proposed model is an association of words so that when related words combine to
form a meaningful word which maynot be in the user-supplied list of search terms- is also included in
the featured word and hence in the search result. We also propose to giveweights to the exact place of
occurrence of a word. For example those in the headline or as "key word" are more important than those
in the bodyof the text. Finally, a scheme is proposed to train such a network to accomplish the entire
scheme outlined above. This will be implemented by managing theweight matrix of the proposed model. 

 Christopher J. Davis - Carnegie Mellon Univ

Biology, Brains and Catalysis

 In this talk, I examine the hypothesis that living processes, from the smallest to the largest
(including the brain), are catalytic processes, and hence, that life is a fractal catalytic process.
Although a physical thermodynamic process, I suggest that catalysis is not confined to enzymatic
biochemistry; rather, enzyme catalysis is a microscopic example of a general process. It is suggested
that rather than characterizing biological processes in terms of 'functions,' they are better
characterized in terms of their property of persistence. The persistence of a living process, which may
be chemical, neural, perceptual or behavioral, is a direct consequence of the catalytic process that
makes 'explicit' the orders and relationships that are 'implicit' in the environment. Because catalysts,
and therefore processes of catalysis, emerge unchanged from the reactions that they mediate, there is a
relation between catalysis and persistence; this relation is most evident in living processes. Moreover,
the persistence of a catalyst may be a consequence of the way in which it mediates the
chemical/thermodynamic tendencies in its environment. This process may involve the order (or structure,
possibly related to the entropy) of the system. In the domain of enzyme catalysis, several researchers
have theorized that the principle agent of catalysis is a type of wave called a soliton (e.g., Caspi &
Ben-Jacob, 2002; Sataric et al., 1991). Solitons are waves that can travel large distances without
significant loss of energy or structure. It is argued that this property is consequent on the
relationship between the soliton and the structure (or boundary conditions) of the medium. In the case
of enzyme catalysis, the soliton is expressed as a vibrational mode involving the molecules that
comprise the enzyme. It will be argued that the persistence of the soliton and the persistence of the
enzyme are necessarily related. A soliton may provide a path between the 'before' and 'after' energy
states via points of ambiguity or'fixed points' (i.e. points that do not change), which may be related
to the 'order' (or structure) in the environment. I will discuss how the soliton wave has been
implicated in several levels of biological process including enzymes, DNA, microtubules, heart function,
muscle function and nerve impulses, evidence that lends support for the fractal catalytic theory. The
fractal catalytic hypothesis is consistent with (but more specific than) the 'order from order'
principle of Schrodinger (1979), who suggested that living systems maintain their order by utilizing the
order in their environments. At the same time, the present hypothesis contrasts with the 'emergentism'
of chaos theory, which fails to address the relationship between far-from-equilibrium dynamic systems,
as may be observed at the level of the brain, and the other systems (both internal to the organism and
external) with which it must interface. Finally, the hypothesis of an abstract but unifying process,
such as catalysis, that characterizes all of life challenges the implicit assumption of theoretical
biology that the strategies and processes that characterize a species are completely contingent upon the
chance evolution of a process that reproduces and mutates. 

 Matthew T.
Dearing - Cornell University

Digitally Mapping Cultured Neuron Networks Matthew T. Dearing , Harold G. Craighead

 An understanding of structural and functional characteristics in a complex network requires a detailed
map of the network's components and interconnections. Data sets representing the Internet,
World-Wide-Web, scientific collaboration networks, and biological processes have been used to analyze
these systems' network characteristics. However, there has been a lack of sufficient architectural data
for another interesting complex network system: interconnected neurons. We present a method of automated
digital image analysis to extract critical network properties from a cultured neuron network. Our data
collection software will provide needed information for mapping cultured neuron systems, which will
later be systematically compared to the functional characteristics of neuron devices with similar
network structures. 

 Ronald DeGray - Saint Joseph College

Developing a Web-based Interactive Syllabus for an Undergraduate Course in Systems Thinking and
Complexity

 We will demonstrate a web-based interactive syllabus for a one semester course in systems thinking and
complexity that is suitable for acapstone course in an undergraduate curriculum in information
technology. We also report our pedagogical and epistemological experiences in developing the course
content. We wanted to provide students with a cognitiveframework beyond the technical aspects that they
would acquire from a text based Information Technology curriculum. We also wanted to takeadvantage of
technology and the rich source of materials available on the Internet. This will be a joint presentation
by Saint Joseph College professors: Dr. Ronald DeGray, Associate Professor of Mathematics and Dr.
ShyamalaRaman, Associate Professor of Economics and Director of International Programs 

 Eugenio Degroote - Universidad Plitecnica de Madrid

Flame Spreading Over Liquid Fuels: A General Model

Understanding flame spreading over liquid fuels is a matter of both fundamental interest and crucial
practical importance, mainly for its relevance to safety issues, as it is the base for the knowledge and
ultimate control of fire propagation/suppression mechanisms.In this work, a complete overview of the
problem is shown. A complete set of experiments have been carried out; they show that, depending on the
initial surface fuel temperature, at least five different spreading regimes have been found. The
appearance of all these regimes is directly related with the existence of a preheated zone (in the
liquid phase) in front of the flame, that contributes in some way to its spreading. Our last
experimental results suggest that the different mechanisms involved in this process are also directly
related with the thermal transfer between the liquid phase (Liquid fuel) and the gas phase (fuel+air);
this interaction between both phases seems to be the main responsible of the existence of so many
differente spreading regimes. Finally, considering flame spreading as a dynamical system, the different
transition temperatures observed have been observed and characterised as well. A numerical model is
being proposed, that fits well with our experimental results. A dimensionless variable has been defined
too, that results to be very similar for al the fuels and exerimental geometries used in our
experiments, for one of the critical temperatures observed. 

 William A.
Dembski - Baylor University

Why Natural Selection Can't Design Anything

 In the early 1970s Leslie Orgel argued that the key problem facing origin-of-life researchers was to
explain the specified complexity inherent in the first living form. Thirty years later this remains the
key problem facing origin of life research. Nonetheless, the biological community is convinced that the
specified complexity of living forms is not a problem once replication is in place and the Darwinian
mechanism has become operative. In this paper I argue not only that we have yet to explain specified
complexity at the origin of life but also that the Darwinian mechanism fails to explain it for the
subsequent history of life. To see that the Darwinian mechanism is incapable of generating specified
complexity, I consider the mathematical underpinnings of that mechanism, evolutionary algorithms.
Roughly speaking, an evolutionary algorithm is any well-defined mathematical procedure that generates
contingency via some chance process and then sifts it via some law-like process. It is widely held that
evolutionary algorithms provide a computational justification for the Darwinian mechanism of natural
selection and random variation as the primary creative force in biology. Nonetheless, careful
examination of evolutionary algorithms and the informational constraints wherewith they are programmed
reveals that evolutionary algorithms, far from eliminating the specified complexity problem, merely push
it deeper. I employ Wolpert and Macready's No Free Lunch results to show that any output of specified
complexity from an evolutionary algorithm presupposes a prior input of specified complexity. And since
all biological design exhibits specified complexity, it follows that evolutionary algorithms (and the
Darwinian mechanism in particular) are incapable of resolving the problem of biological design. 

 Tessaleno C. Devezas - Los Alamos National Laboratory

Aggregate Output Model Describing the Limit-Cycle Behavior of a Logistic Growing Process in
Socioeconomic Systems

 A socioeconomic system is an evolving complex adaptive system with many kinds of participants, which
interact in intricate ways that continually reshape their collective future. During the ongoing
evolutionary process the system self-organizes and learns configuring and reconfiguring itself toward
greater efficiency among greater complexity. Each stage of the evolutionary process corresponds to a
given structure that encompasses previous self-organization, learning and current limitations. This is
to say that self-organization and learning are embodied in the system's structure and the learning rate
is an overall system's property. Such stages of the evolutionary path of a socioeconomic system are well
described by simple logistic curves that to some extent conceal the complexity of mechanisms involved In
this paper a cybernetic framework is proposed which, using a chaos based approach, may help to unveil
some hidden properties of the logistic learning collective dynamics. From the relationship between the
differential and the discrete logistic equations, it is demonstrated that the unfolding of a logistic
(learning) process is constrained by two control parameters: the aggregate learning rate  and a
generation-related characteristic time tG, whose product maintained in the interval tG<4 (deterministic
chaos) grants the enduring evolutionary process. Describing the socioeconomic system discretely as a
logistic growing number of interactorsadopting a new set of ideas (new learning) and using the logistic
function as the probabilistic distribution of individuals exchanging and processing information in a
finite niche of available information, it is demonstrated that the rate of information entropy change
(K-entropy) exhibits a four-phased limit-cycle behavior. Implications of this modeling on reducing
logical uncertainties in predicting the behavior of social systems are discussed. 

 Solomon Gilbert Diamond - Harvard University

Measuring Hypnosis: Relating Mental State to Systematic Physiological Changes

 A fundamental problem in hypnosis research is to quantitatively assess the hypnotic depth of subjects
because the subjective experience of patients during hypnosis cannot be measured directly. Prior
evidence exists that systematic physiological changes during hypnosis may be reflected in heart rate
variability (HRV). A novel method is presented for estimating HRV parameters that change dynamically on
the time scale of seconds. The estimated parameters are combined into a single normalized HRV dynamic
parameter (nHRVdp). This parameter was found to increase systematically across 10 subjects during the
hypnotic state when compared with a commensurate control condition (p<0.000001). Significant
correlations were found between the number of subjective hypnotic phenomena experienced by subjects and
mean nHRVdp (p=0.043). Dynamic self-rating of hypnotic depth during hypnosis was also found to correlate
significantly with nHRVdp (p=0.0497). These results suggest that an ECG monitor together with the
proposed algorithm can objectively measure hypnotic depth. This "hypnometer" could have broad
applications in clinical hypnosis and in research to better understand the physiology of the hypnotic
state. 

 Commander John Q. Dickmann - US Navy

Complex Systems Research and Information Age Warfare

 Defense community innovators have proposed concepts for using cutting-edge technologies to solve long
standing military challenges, including destruction of time-critical targets, theater-wide surveillance,
power projection and access to littorals. These concepts assume great benefits from networking that will
enable military advantage by use of distributed systems. However, the advantages of networking as well
as the implications of engineering distributed systems have not been fully articulated. This paper
defines and describes how distributed, networked forces provide advantage; translates the advantage to
engineering aspects of distributed system characteristics, functionality and design goals; and
introduces a method of developing the engineering competencies required to design effective distributed,
networked military forces. 

 Fred M. Discenzo - Rockwell Automation

Managed Complexity in An Agent-based Vent Fan Control System Based on Dynamic Re-configuration

 New developments in advanced control techniques are occurring in parallel with advances in sensors,
algorithms, and architectures Dubbelsthat support next-generation condition-based maintenance systems.
The emergence of Multi-agent Systems in the Distributed Artificial Intelligence arena has shifted
control system research into a very challenging and complex domain. A multi-agent system approach enable
us to encapsulate the fundamental behavior of intelligent devices as autonomous components that exhibit
primitive attitudes to act on behalf of equipment or complex processes. Based on this approach, we have
implemented an initial set of systems that validate this methodology to manage the inherent complexity
of highly distributed systems to provide a consistent and aggregated system behavior. There are three
essential aspects of this new paradigm that suggest the significant potential of intelligent,
self-organizing distributed architectures; these are: · The complexity of solutions of highly monolithic
control systems is now understood as an aggregated family of well-delimited primitives. This aspect
exposes the new systems as hybrid solutions that area made of well-connected hardware and software
de-coupled units · The primitive components represent an emergent infinitesimal behavior in which each
primitive is only programmed with complete rules to satisfy local pursuits. Primitives are programmed
with incomplete (or none) rules to describe the relationship with other primitives. Each primitive uses
an agent language and is capable ofcommunicating peer primitives (e.g. Agents, Holons, or Cellular
Automata). · The system is then initially exposed to chaotic interactions among the primitives. The
chaotic interactions are defined as a highly intense capability matching task (association). It is also
postulated that the overall behavior will progressively converge to temporal clustering of primitives.
This is what we refer to as managed complexity. The concepts above and new engineering developments have
helped achieve new and important capabilities for integrating CBM technologies including diagnostics and
prognostics with predictive and compensating control techniques. Integrated prognostics and control
systems provide unique opportunities for optimizing system operation such as maximizing revenue
generated for capital assets, maximizing component lifetime, or minimizing total life-cycle costs. We
have defined and implemented an integrated diagnostic / prognostic / controller system in the context of
an agent-based representation. This integrated agent-based system has been implemented in a variable
frequency drive (VFD) and in a programmable logic controller (PLC). The VFD agent has been demonstrated
operating an axial vent fan system. During operation the system dynamically adjusts system operation
based on the collaboration with other agents operating in related parts of a chiller system. This paper
will describe the foundation technologies that are essential to realizing an adaptive, re-configurable
automation system. These technologies include diagnostic and prognostic algorithms, advanced control
techniques, agent-based framework for real-time automation systems, and system integration / modeling
techniques. We will describe a demonstration system we have developed for HVAC applications that utilize
these technologies. We will also cite the important, unprecedented capabilities this new system can
provide along with open issues and research challenges associated with the wide-scale deployment of
robust integrated prognostics and control systems in an agent framework. 

 Raj
P. Divakaran - Utah State University

Potential of Ant Colony Optimization to Satellite Image Classification

 Studies on the behavior of social insects such as ants have shown that they exhibit a collective
intelligence which cannot be explained based on the actions of the individuals that compose the colony.
In an ant colony, global or emergent patterns emerge due to the local interactions of the ants. This
ability to generate global patterns by a swarm of ants has served as the basis for the development of a
set of algorithms that emphasize distributedness, robustness and flexibility - Ant Colony Optimization
(ACO) Algorithms. Ant Colony Optimization techniques have been used to solve a number of network
problems such as the Travelling Salesmans Problem and the Job Scheduling Problem. Vitorino Ramos and
Filipe Almeida have shown that ACO algorithms could also be developed to solve the problem of image
segmentation. This development is of immense use to researchers in the field of image classification and
pattern recognition who have of late concentrated on developing robust, autonomous and flexible
algorithms that can delineate the classes with minimum or zero inputs from the user. The potential of
ACO algorithms to image classification has not been fully investigated so far. The purpose of this work
is therefore to determine if ACO algorithms can be successfully applied to the problem of image
classification of satellite imagery. 

 Brock Dubbels - The Norwegian
University of Science and Technology

Building a Network to the People

In consideration of the goals of the New England Complex Systems Institute to educate the population of
complex systems in awareness, perspective, and methods, it is important to consider ways to create a
learning web for the distribution of NECSI's most valuable resource: knowledge. Let it be made clear
form the beginning, beyond the knowledge and methods, the most valuable resource that this community has
is the individuals who are interested in Complex Systems work. These individuals could, through their
training at NECSI, become mentors and trainers to local community schools for a limited time. This might
enable NECSI to apply for educational dollars from local and federal sources and provide much needed
support for young people, broadening awareness of complex systems for young people, their teachers, and
a deeper learning for the students affiliated with NECSI in their training. The issue becomes one of
vision and method. In order to make this work, one might suggest that we use any methods possible, but
in consideration of the importance of practicing what we preach, it might be best to offer education
about complex systems through a complex systems approach. There is a wide variety of method in teaching
and learning, and many of the methods compliment if not endorse a complex systems approach to the world.
The goal of this paper is to put forward suggestions for teaching methodologies that are representative
of the complex systems approach to problem solving, i.e., it might be easy to have one person stand in
front of a group and lecture, but does this represent a movement away from centralized control or the
processes of group interaction. In the body of this paper, a review of approaches to curriculum and
learning promotion will be organized to present different modes of instruction from either a
traditional, or complex systems approach; recommendations will be made to offer an approach that not
only serves to educate the populace about complex systems, but also does so from a complex systems
approach in curriculum in form and substance. A review of the ideas of the K-16 education initiative and
the session on education at the ICCS3 conference will be included with current ideas in education and
learning. recommendations will be made for enacting this curriculum with members of NECSI, and future
members. 

 Juan Carlos Chimal Eguia - ESCOM-IPN

Some Further Analogies Between the Bak-Sneppen Model for Biologcal Evolution and the Spring-block
Earthquake Model

 Along recent years a lot of attention has been devoted to the so-called self-organized critical
systems: Which are open,extended systems, that organized themselves into steady metastable states,
without any temporal or spacial predominant scale (except those impose by the finite size of the
system). The SOC concept has been used to describe statistical properties of several physical systems by
means of numerical models based on cellular automata. In particular, Bak and Sneppen proposed a
SOC-model for biological evolution at the level of entire species or faunas, that exhibits punctuated
equilibrium behavior. On the other hand, Olami., Feder and Christensen, have suggested that the
two-dimensional spring-block earthquake model can explain some properties of real seismicity. In the
present work we show that there exists several interesting analogies between these SOC-models. Both of
them exhibit punctuated equilibrium in the long term, which leads us to suggest an equivalent
characterization of seismic and "evolutionary" provinces through the long term slopes of the
stair-shaped graphs of cumulative activity in the courseof time. 

 Ibrahim
Erdem - Yildiz Technical University

Customer Relationship Management in Banking Sector and A Model Design for Banking Performance
Enhancement

 Huge growth of Customer Relationship Management (CRM) is predicted in the banking sector over the next
few years. Banks are aiming to increase customer profitability with any customer retention. This paper
deals with the role of Customer Relationship Management in banking sector and the need for Customer
Relationship Management to increase customer value by using some analitycal methods in CRM applications.
CRM is a sound business strategy to identify the bank’s most profitable customers and prospects,
and devotes time and attention to expanding account relationships with those customers through
individualized marketing, repricing, discretionary decision making, and customized service-all delivered
through the various sales channels that the bank uses. In banking sector, relationship management could
be defined as having and acting upon deeper knowledge about the customer such as how to find the
customer, get to know the customer, keep in touch with the customer, ensure that the customer gets what
she/he wishes from service provider, and understand when they are not satisfied and might leave the
service provider and act accordingly. This study will underline CRM objectives such as growth, retention
and cost reduction. Increasing customers’ product cross-holdings, maximizing the contribution from
each customer through time and increasing efficiency are couple of those objectives, which this study
will cover. Under this case study, a campaign management in a bank is conducted using data mining tasks
such as dependency analysis, cluster profile analysis, concept description, deviation detection, and
data visualization. Crucial business decisions with this campaign are made by extracting valid,
previously unknown and ultimately comprehensible and actionable knowledge from large databases. The
model developed here answers what the different customer segments are, who more likely to respond to a
given offer is, which customers are the bank likely to lose, who most likely to default on credit cards
is, what the risk associated with this loan applicant is. Finally, a cluster profile analysis is used
for revealing the distinct characteristics of each cluster, and for modeling product propensity, which
should be implemented in order to increase the sales. 

 Peyman Faratin - MIT

A Multiagent Simulation Model of the Emergence and Dynamics of Negotiation in Complex Contracting Games

Complex systems are often characterized by the emergence of global behaviors that are often difficult to
anticipate simply by inspection of the system's components. In this paper we present some interesting
emergent properties that arise in the context of negotiation over contracts with multiple interdependent
issues. In particular, we show that a concessionary strategy is actually superior if adopted by both
parties, but the possibility of allowing agents to adopt a non-conessinary strategy introduces a
prisoner's dilemma. These results, derived from multi-agent system simulations, are potentially relevant
to human contexts such as collaborative design. 

 Timothy Field - Victoria
University of Wellington

Complex Adaptive Behaviour in Physical Simulation

 Continuing advances in computational power allows for the study of artificial life at a greater degree
of realism than ever before. We utilise this power to simulate physically realistic environments and
investigate the adaptation of virtual agents which optimize their behaviour in these environments with
respect to predefined goals. In particular, the performance and interaction of online reinforcement
learning and offline population-based learning algorithms in adapting both agent morphology and neural
network control is examined in simulations of rigid-body and fluid dynamics. We demonstrate examples in
which the tight coupling between an agent and the non-linear dynamical system it is situated in gives
rise to the emergence of non-trivial behaviour. 

 Brigitte Fleeman -
University of Texas at Austin

Sensemaking of a Change Intervention With Insights From Complexity Science

 Using Weick's (1995) framework of sensemaking and the insights gained from research on complex-adaptive
systems, I am interested in empirically investigating the question of whether and how differences among
agents affect the sensemaking processes of small groups as complex-adaptive systems. In general,
diversity is a requisite and hallmark of complex adaptive systems. In the org