Parameter-free, adaptive clonal selection

The ´clonal selection´ approach to optimization - an abstraction of immune system adaptation, popularized mainly by de Castro and Von Zuben - may have reached something of an impasse. The method requires several parameters that must be tuned, and it normally uses a binary representation that can limit the accuracy of the results. This paper examines the uses and drawbacks of clonal selection, and suggests some ways forward, based on an analysis of the operators for choosing the amount of mutation and the number of clones. As a result, an effective, real-valued, parameter-free clonal selection algorithm is introduced, called adaptive clonal selection (ACS).