Presentation by Chris Salzberg & Antony Antony
"A Closer Look at the Evolutionary Dynamics of Self-Reproducing Cellular
Automata".

Abstract :
In this talk we present an overview of our current work on the study of
evolutionary processes emerging from an abstract logical form of
self-reproduction. The model we use, the ``Evoloop'' (Sayama, 1998),
evolves in a deterministic 9-state cellular automata (CA) space with a
Von Neumann neighbourhood. In open periodic domains, we show that
Sayama's model gravitates towards a limited subset of small-sized
species due to high-density homogeneous population distributions. To
counter this trend, we introduce a new and dynamic environment, one
which promotes free space through domain partitioning. We show new
large-scale (3K x 3K) results using this environment in which
populations demonstrate allopatric speciation leading to diversity over
very long time scales (1.4 M iterations) as well behaviour resembling
punctuated equilibrium. We then outline a new identification scheme for
better analysis and computational efficiency; this leads to the
discovery of thousands of different species which we classify using a
"genealogy graph" (a generalization on the familiar tree heirarchy). We
finish with a discussion of some new ideas emerging from this framework,
including the graph-space representation of an evolutionary bottleneck
and a concept which we call "evolutionary coupling".