Tunable Detectors for Artificial Immune Systems: From Model to Algorithm

Artificial immune systems (AIS) are often developed directly from textual descriptions of immune properties with little attempt to understand those properties free of an engineering domain or application. Here we present an example of how an AIS can be developed based on an in-depth investigation of an immune model. This model describes how the activation thresholds of immune cells can be tuned based on their recent history of excitation events. Using this model we detail how a population of randomly generated, degenerate and tunable detectors can produce a response pattern to a particular antigenic stimulus. Our understanding of this model leads to the identification of how these ideas can be incorporated into the engineering domain, highlighting the response pattern as a pre-processing of input data. Based on this we extract a framework for incorporating degenerate tunable detectors into an AIS. We then instantiate this framework by developing an AIS for pattern classification that shows promising results.