This paper presents an artificial immune system (AIS) which produces artificial immune networks that are meaningful, of a bounded size and dynamic over a very large number of data presentations. This behaviour had proved elusive up to this time but has now permitted the application of the AIS to situations requiring continuous learning. It also removes the need to decide when to stop training an AIS. The new version of the algorithm is described, and results are presented for analysis of static and dynamic versions of a trivial two-dimensional data set and Fisher’s Iris data. It is argued that the changes made from previous versions of the “resource limited” algorithm are in keeping with the goals of remaining true to the immune system analogy and making the system as simple as possible.
|Number of pages||10|
|Publication status||Published - 2002|
- network model
- artificial immune systems