Self/Non-self Discrimination Based on Fractional Distance

Self/Non-self discrimination is the basic mission of the artificial immune systems (AIS). In traditional AIS, detectors discriminate non-self elements based on the Minkowski distance between detectors and antigens. However, in high dimensional feature space the distances between elements converge to a similar value, so the data discrimination will become much difficult. In this article, the relationship between data dimension and data relative contrast is analyzed and the fractional distance is utilized to improve the detection performance.