Estimating the detector coverage in a negative selection algorithm


This paper proposes a statistical mechanism to analyze the detector coverage in a negative selection algorithm, namely a quantitative measurement of a detector set's capability to detect nonself data. This novel method has the advantage of statistical confidence in the estimation of the actual coverage. Furthermore, unlike the existing analysis works of negative selection, it doesn't depend on specific detector representation and generation algorithm, Not only can it be implemented as a procedure independent from the steps to generate detectors, the experiments in this paper showed that it can also be tightly integrated into the detector generation algorithm to control the number of detectors. Copyright 2005 ACM.

Publication Title

GECCO 2005 - Genetic and Evolutionary Computation Conference