Information Technology Reference
In-Depth Information
The Idiotypic Network
with Binary Patterns Matching
Krzysztof Trojanowski
1
and Marcin Sasin
2
1
Institute of Computer Science, Polish Academy of Sciences,
Ordona 21, 01-237 Warsaw, Poland
2
Warsaw School of Information Technology,
Newelska 6, 01-447 Warsaw, Poland
Abstract.
A new specification of an immune network system is pro-
posed. The model works on a set of antibodies from the binary shape-
space and it is able to build a stable network and learn new patterns
as well. A set of rules based on diversity of the repertoire of patterns
which control relations of stimulation and suppression is proposed. The
model is described and the results of simple experiments with the imple-
mentation of the model without and with presentation of antigens are
presented.
1
Introduction
Ability to learn to distinguish and discriminate the self patterns from non-self
ones present in the immune systems is explained by the immune network theory
proposed by Jerne [6]. The hypothesis assumes that the system is composed of
a set of molecules, i.e. antibodies and antigens. There are also rules that con-
trol interactions between antibodies and between antibodies and antigens. The
interactions turn this set of molecules into a self regulated network which has
its own mechanisms responsible for insertion and elimination of molecules. The
rules are based on the molecules' traits which assembled all together can be
interpreted as points in the multidimensional space where each of dimensions
represents a single trait. The result of these interactions is a continuous proces
of modification i.e. growth or depreciation of the concentration of molecules in
the organism. Depreciation of the concentration of selected antibodies or anti-
gens brings them to elimination from the system. Eliminated molecules can be
replaced by new ones and this way the system is able to rebuild itself according
to its metadynamics. For the summary of the first models based on this theory
which were developed by Farmer et al. and Varela see [3].
In recent years a set of different practical specifications of the models was
already proposed. For extended discussion on properties of immune networks
based on two types of shape-space: the multidimensional real space and the
binary space see the Bersini's publication [2]. Our network manages binary pat-
terns and its activity is controlled by the rules that are easy to implement and of
low computational cost. However there is also a set of assumptions which differ-
entiates it from the networks presented by Bersini. In [4] Galeano et al. present
