Sieves for Selection of Genes According to Desired Behaviors - Systematic Design for Emergence in Cellular Nonlinear Networks - page 94

the following statement A : “The cellular automata acts like a random number

generator”. How shall one define a proper sieve to select only cell genes providing
such a behavior?
In the next we will exemplify this process, giving a methodology that can be
generally applied for other definitions of some desired behaviors.

Fig. 6.1. Locating interesting behaviors using two “sieves”. The result is a pool of genes for
which the CA is expected to have a desired behavior. Defining sieves is equivalent to defining
two functions:

P

x for the upper sieve and

P

x for the lower sieve

2

First note that defining a sieve requires some previous expert knowledge of the
relationship between behaviors and various complexity measures. Such knowledge
may be improved by additional experiments and transferred to the sieve definition
using the same methodology as the one used to design a fuzzy expert system. Since

the number of emergence measures is limited and since within each such measure

there are distributions suggesting several categories, it is quite easy to analyze a
few representative categories for each complexity measure noting down the behavior
observed visually in the cellular automata. It is assumed (and actually confirmed
by experiments) that slight changes in the value of the emergence measure does
not affect substantially the overall behavior unless a bifurcation point is reached.
Such bifurcation points were revealed for most of the complexity measures intro-
duced in the previous chapters and practical methods to locate them were also
given in Chap. 4.