A Boundary-aware Negative Selection Algorithm

Negative selection algorithms generate their detector sets based on the points of self data. In the approach described in this paper, the continuous self region is defined by the collection of self data. This has important differences from the negative selection algorithms that simply take each self point and its vicinity as the self region: when the training self points are used together as a whole, more information is provided than used as individual points; the boundary between self and nonself regions are detected in the algorithm. It also demonstrated that a negative selection algorithm as a unique strategy can obtain certain results that straightforward positive selection cannot. Experiments are carried out using both synthetic data and real world applications. The former was designed to highlight the difference from conventional point-wise interpretation of self data in negative selection algorithms.