Information Technology Reference
In-Depth Information
The second goal alludes to the problem of determining to which extent should the
reproducing activity of the cellular patterns share its resources with its self-
maintenance needs. Even the most complex proposed cellular reproducing patterns
cannot be compared with intricate nets of components as are autopoietic systems or
autocatalytic nets, but we think that, for the logic of reproduction, the replicating
entities must exhibit some degree of identity invariant that requires in its turn a logic
of self-production.
Finally, if we have to consider an evolutionary scenario it can be supposed that
there will be some random variation source that can be potentially harmful for the
normal operation of the reproducers. We also have to take into account that the
selective pressure on the patterns can produce “deceases” in the population. This
twofold reason determines also that these conditions under which the surviving
patterns will have to operate will be far from the blank quiescent space in which most
of the proposed reproducing models do operate.
The different models of cellular reproducers have been analyzed and classified into
three main groups [16], according to the role that a self-description of the pattern
plays in each of them. So, we can distinguish three different reproduction strategies:
•
Universal Construction Strategy (UCS)
•
Dynamic Description Strategy (DDS)
•
Self-Inspection Strategy (SIS)
UCS is best illustrated by von Neumann's automaton [2], in which reproduction is
achieved as a fixpoint in the more general feature of universal constructibility. First a
mechanism to decode a stored description of an arbitrary cellular pattern is defined.
Then a pattern able to construct itself by means of its own description is designed.
Langton's loops [4] represent the first and most known instance of DDS, in which the
need of decoding the blueprint of the reproducing pattern is removed. This goal is
achieved using an executable description, which only need to be transported to the
construction area. Finally, as opposed to the previous two, that we can call genetic
strategies because of their use of the genotype/phenotype duality, SIS was proposed in
our earlier work [12]. Morita and Imai's reversible cellular reproducers [13] have also
self-inspecting capabilities. These models don't need to store any kind of self-
description because they are able to produce it by self-inspection.
In [16] it is argued that SIS has an intrinsic complexity factor due to the
endosemiotic interpretation of all the information handled during the construction
process. Moreover, it is suggested that SIS is better adapted to reproduction, and
specifically to the emergence of reproducing entities, than genetic strategies.
In this paper we have tried to design an experimental framework in which we can
test this hypothesis. Obviously we have been forced to leave apart universal
construction-based reproduction: the technical complexity of the models proposed for
this paradigm barely allow their implementation, so it doesn't seem feasible to extend
their transition tables furthermore to achieve robustness of the patterns. A simple
estimation of the cost of such operation can be done extrapolating the data obtained
for the costs of our system (see section 3).
On the other hand, the models proposed for reproduction which use dynamic
description or self-inspection can be very suitable to be tested together, since they
have similar technical complexity (in number of states and transitions or size of the
patterns) and an important set of common features. Nonetheless, since robustness is a
key characteristic to be achieved, any comparison of these models would be biased if
