Studies on Real-Valued Negative Selection Algorithms for Self-Nonself Discrimination

The artificial immune system (AIS) is an emerging research field of computational intelligence that is inspired by the principle of biological immune systems. With the adaptive learning ability and a self-organization and robustness nature, the immunology based AIS algorithms have successfully been applied to solve many engineering problems in recent years, such as computer network security analysis, fault detection, and data mining. The real-valued negative selection algorithm (RNSA) is a computational model of the self/non-self discrimination process performed by the T-cells in natural immune systems. In this research, three different real-valued negative selection algorithms (i.e., the detectors with fixed radius, the V-detector with variable radius, and the proliferating detectors) are studied and their applications in data classification and bioinformatics are investigated. A comprehensive study on various parameters that are related with the performance of RNSA, such as the dimensionality of input vectors, the estimation of detector coverage, and most importantly the selection of an appropriate distance metric, is conducted and the figure of merit (FOM) of each algorithm is evaluated using real-world datasets. As a comparison, a model based on artificial neural network is also included to further demonstrate the effectiveness and advantages of RNSA for specific applications.

Name thesis:degree_name

MS in Electrical Engineering

Discipline thesis:degree_discipline

Electrical Engineering

Year dc:date.available

2010

Author dc:creator

• Dixon, Shane E

Contributors dc:contributor

• Xiao-Hua (Helen) Yu

Identifier

10.15368/theses.2010.8

OAI identifier oai:identifier

oai:digitalcommons.calpoly.edu:theses-1247