Digital Signal Processing Reference
In-Depth Information
8.5 Concluding Remarks
The natural tendency to explore alternative techniques and the requirements
of some concrete applications have led signal processing theory toward a
more widespread use of nonlinear structures and unsupervised adaptation
criteria. Extensions of this nature are likely to originate optimization tasks
of a nature considerably different from that associated with the more clas-
sical frameworks. Then, the search of alternative optimization methods to
improve the global search potential may become attractive.
In this chapter, we discussed some approaches belonging to the field of
bio-inspired computing. Our intention, it is important to remark, was not
to provide a complete view of the field, but to make use of three important
classes of methods: GAs, AISs, and PSO. The main goal is to analyze essen-
tial features of these tools and evaluate their potential of application in our
problems of interest.
We observed that a GA has the following features: there is local search,
there are elements of global search, and there is the influence of selection
in the survival of certain solutions in consonance with the characteristics of
the function to be optimized. It is worth noting that the method requires
information about the cost function itself, but usual requirements such as
continuity, differentiability, etc., do not play a key role here. Hence, it is par-
ticularly attractive when we deal with cost functions that are very difficult, or
even impossible, to manipulate. Certainly, in GA and other techniques, com-
plexity is a factor to be considered in many online applications. Nevertheless,
it does not discard the theoretical interest of the methods neither their appli-
cations in contexts where the real-time operation at high information rate is
not an issue.
Overall, we hope this chapter could serve to instigate a larger group of
the signal processing community to consider bio-inspired methods as viable
components of their toolkit.
