Latest Immune Models and Hybrid Approaches - Immunological Computation: Theory and Applications - page 165

T s cells suppress humoral-immune response by suppressing the T H . If T DH
cell is suppressed, then there is no cellular immune response. If T cy is sup-
pressed, then tumor and viral immunity is aff ected.

In innate immune response, some cytokines such as IL2, IL4, IL12, and IFN γ are
produced by macrophages; these cytokines can determine the generation of diff er-
ent types of T cells.
Although there is no computational model yet published on T-cell phylogen-
esis, this provides another area of further research in immunological computation.

6.5 Combining Negative Selection
and Classifi cation Techniques

In many anomaly detection applications, only positive (normal) samples are avail-
able at the training stage. However, most conventional classifi cation algorithms
need both self- and nonself samples.
To allow conventional algorithms to be used when only samples from one class
are provided, a hybrid algorithm was proposed (González et al., 2002), which is
able to create synthetic nonself samples from a set of self-samples. h e algorithm
generates a set of detectors (points) that cover the nonself space using NS, and then
these points are used as samples of the nonself class, allowing the use of conven-
tional classifi cation algorithms (Figure 6.6).

Normal
samples

New
samples

Negative
selection
algorithm

Negative
samples

Anomaly
detection
function

Classification
algorithm

Normal Abnormal

Training Detection

Figure 6.6 Combining NS-SOM in generating classifi er dataset.