Information Technology Reference
In-Depth Information
multi-objective problems and uni or multi-global problems. The proposed
omni-
optimizer
algorithm, hereafter denoted DT omni-optimizer, is mainly based on
a
ranking
procedure that uses a modified
constrained dominance
principle and
adapts itself to solve different kinds of problems. Further explanation of this
ranking procedure will be given in Section 4.
In the field of evolutionary computation, a relatively novel computational
paradigm, namely Artificial Immune System (AIS), was originated from at-
tempts to model and apply immunological principles to problem solving in a wide
range of areas such as optimization, data mining, computer security and robotics
[4]. Three advantages of advanced AISs over other population-based strategies
are: (
i
) they are inherently able to maintain population diversity (modules with
some resemblance with niching and fitness sharing are intrinsic parts of the algo-
rithm); (
ii
) the size of the population at each generation is automatically defined
according to the demands of the application; and (
iii
) local optimal solutions
are simultaneously preserved once located.
Based on the successful application of AISs to several kinds of function opti-
mization problems ([3], [5] and [7]), this work presents a novel proposal called
omni-aiNet
, which unites the flexibility given by the principles of the DT omni-
optimizer [8] with the intrinsic advantages of AISs over other population-based
strategies. The results obtained with this basic version of omni-aiNet indicated
that the algorithm is very effective to deal with demanding scenarios, although
some improvements are still required.
This paper is organized as follows. Section 2 presents a brief introduction
to the concepts of AISs and the main immunological theories that inspired the
proposed algorithm. Section 3 introduces some formalism of function optimiza-
tion and depicts the notation that will be used throughout the paper. Section
4 presents and details the proposed algorithm, and Section 5 outlines a brief
conceptual comparison between omni-aiNet and the DT omni-optimizer [8]. The
description of the experiments and the presentation of the obtained results are
fulfilled in Section 6. Finally, Section 7 draws some concluding remarks.
2
Artificial Immune Systems
The natural immune system can be considered one of the most important com-
ponents of superior living organisms. The permanent cycle of recognition and
combat against pathogens (infectious foreign elements) has the goal of keep-
ing the organism healthy. The molecular patterns expressed in those invading
pathogens or
antigens
are responsible for triggering the immune response when
properly recognized by the immune cells.
Some of the cells with major roles in the immune response are the
lymphocytes
,
which can be divided into two types: B lymphocytes (B cells) and T lymphocytes
(T cells). The present description will focus only on the B cells. When an antigen
is detected, the B cells that best recognize the antigen (best
a
nity
) will prolif-
erate by cloning. Some of the clones will differentiate into
plasma cells
(the main
antibody secretors
) while the others will differentiate into
memory cells
.These
