[Simulchaord-discuss] More on Evo

[Paul Fernhout]

Evo looks like a possible platform -- GPLd and built on Swarm. Need to
find out more on it. (Not Python, but it may have other advantages...) I
will need to evaluate this package more as time permits.

Repeating a bit from before:

  Evo -- built on top of Swarm
  http://omicrongroup.org/evo/
  GPL
  "Evo is a software development framework that allows developers to
build complex alife simulations. Using Evo, researchers can easily build
systems of independent agents interacting with one another and with
their environment. Evo implements biological operators such as genetic
recombination and mutation to evolve the behavior of agents so that they
are more adapted to their environment. "

  And from a linked page on Evo:
  http://omicrongroup.org/evo/overview/html/node5.html
  "Although the examples discussed in this paper and in the Programmer's
Guide are biologically based, Evo was designed to be useful for
exploring any complex system of autonomous agents regardless of problem
domain. Social systems and economic simulations are also candidate
applications for Evo. Also, it is not necessary that there be a spatial
component to the simulation. For example, a simulation of trading
activity in a stock market would probably not have any notion of space.

  And the Evo overview:
  http://omicrongroup.org/evo/overview/html/node3.html
  "Individuals in an Evo simulation (called ``agents'') also have
programs as their genomes. These programs are not written in LISP,
however, but in a high-level programming language with a syntax similar
to C or Java. The programs of individuals in an Evo simulation are
concurrently executed over and over again by the Evo framework. Evo is
different from GP in that a single execution of an individual's program
is not sufficient to evaluate the fitness of the individual. In fact,
Evo doesn't evaluate the fitness at all; the mere survival of the
individual is the implicit measure of fitness. With Evo, we are not
trying to solve a particular problem. The problem is the environment,
and the solution is a strategy that survives well enough in the
environment such that the strategy becomes dominant in the population. 

The genetic recombination step is done differently in Evo than in GP. In
the GP algorithm, after we have selected a set of survivors to be
parents, they are paired up and their programs combined to produce
offspring programs. Then the fitness evaluation step is performed again
with this new population. In an Evo simulation, genetic recombination is
not explicitly handled in a centralized, lock-step fashion. Agents mate
with one another only if their programs contain instructions for mating
with other agents. Genetic recombination of individuals to produce
offspring is simply an effect of running the individuals' programs."

-Paul Fernhout
Kurtz-Fernhout Software 
=========================================================
Developers of custom software and educational simulations
Creators of the Garden with Insight(TM) garden simulator
http://www.kurtz-fernhout.com