Sieves for Selection of Genes According to Desired Behaviors - Systematic Design for Emergence in Cellular Nonlinear Networks - page 108

of the CAs associated with remaining genes reveals long transient behaviors
converging towards highly ordered (low periodic) global states. Therefore, they
were classified as “not-interesting” based on a subjective assumption of the “intelli-
gent observer” that interesting is associated with a certain complexity as encountered
in his life.
Still it would be interesting to refine the sieve for a better tuning to the expecta-
tions of the intelligent observer. It turns out that the interesting genes selected among
all, share such a common measurable feature, i.e. their corresponding variance
ranges as Var .

By redefining the sieves accordingly, a shorter list of genes becomes available
which contains individuals better tuned to the expected objective.
This continuously sieving tuning process is specific for any adaptation process
or for any interaction between two intelligent entities. What is really interesting
about it is the possibility to learn desired behaviors by iteratively tuning numerical
parameters of the sieves to be compared with a limited set of defined measures of
complexity. This is a confirmation of the usefulness of such measures as defined
and introduced in Chap. 4 and the iteratively tuning process of sieves allows for
the design of CA suited for various classes of signal processing applications, like
those presented in the next chapters.

0 0