Digital Signal Processing Reference
In-Depth Information
Historical Notes
The history of evolutionary computation began with a number of different,
albeit conceptually interrelated, attempts to incorporate the idea of evolution
into a number of problem-solving techniques. The origins of the field can
be considered to be in the 1950s and 1960s [22]. Three branches that grad-
ually affirmed themselves can be highlighted: evolutionary programming
(EP) [109], evolution strategies (ESs) [247,266,267], and GAs [145].
The proposal of GAs is indelibly associated with the name of John Hol-
land, whose 1975 book,
Adaptation in Natural and Artificial Systems
[145], is a
classic of the field. In the next decades, the study of GAs attracted exponen-
tially growing attention, which led to the establishment of a vast community
of researchers in the field.
AISs [83,87] are a more recent field of research, although it can be safely
considered to be a well-established one (an example of this assertion is
that the important
International Conference on Artificial Immune Systems
has
reached, in 2009, its eighth edition). AISs compose a vast repertoire of com-
putational tools that are being broadly applied in many domains, such as
pattern recognition, autonomous navigation and control, data analysis (clus-
tering and data mining), and optimization [87]. Since our interest lies in
immune-inspired optimization tools, we would like to highlight the work
of de Castro and von Zuben [88], and also the work of de Castro and Timmis
[86], who developed an artificial immune network for optimization, known
as opt-aiNet.
Kennedy and Eberhart's work can be considered to be the origin of the
approach known as PSO [167], although not of the field of swarm intel-
ligence, which other branches like ant-colony approach [102, 103]. Swarm
intelligence can be related to important notions like the emergence of collec-
tive behavior, which accounts for what can be considered a new computation
paradigm. PSO is, nowadays, a well-established optimization tool, with a
number of variants and versions [73,273].
8.1 Why Bio-Inspired Computing?
The origins of adaptive filter theory are intimately related to certain hypothe-
ses such as the use of linear feedforward structures and of a mean-squared
error criterion, which give shape to the Wiener theory. In the context of the
Wiener framework, we are faced with an optimization scenario that suits
very well the classical tools based on the derivatives of the MSE cost function,
which form the basis of the LMS and RLS algorithms, as well of many other
techniques that were not discussed in this work.
