|
Krzysztof Trojanowski, Micha³ Jankowiak
Properties of the affinity measures in the immune network model wit binary representation of immune cells and molecules
977
Abstract
A bitstring model of an immune network model proposed by Farmer et al. in
1986 is studied in this text. As de Castro and Timmis wrote about the
paper in their book (2002): "Their works are of extreme importance, as
they constituted the first attempts to bring together mathematical
theoretical immunology and computational intelligence paradigms. As in
Jerne~Rs work their model consisted of a set of differential equations to
quantify the dynamics of the components of the immune network." One of
these equations is an affinity measure used to evaluate match between
binary strings. This is the subject of the study presented below. A set of
tests was performed to study the sensitivity of the measure to the
different types of binary patterns and to different numbers of distinct
bits between the strings. A new transformation operator for binary strings
is also proposed. Application of the operator modifies properties of the
affinity measure proposed by Farmer. This new measure proves to be more
sensitive to the number of pairwise different bits and the position of
these bits in the pair of strings than the previous one.
Keywords: immune network, artificial immune systems, affinity measure.
|
|
 |
 |
