Convergence and dominance in the minimal clonal selection algorithm

The minimal clonal selection provides a reduced realisation of the principles of the clonal selection theory, embodying the information processing properties of a parallel mutation-based hill climber and the winner-take-all competitive learning policy. In the work proposing the approach, a set of performance expectations were proposed regarding convergence and resource utilisation of the approach. This work investigates these expectations of the minimal clonal selection algorithm in a equally minimal random antigenic determinant problem domain. The resource utilisation behaviour is demonstrated to occur as expected, although the number of determinants is show to linearly increase the number of exposures the system requires to acquire the full set of patterns from the domain. This variation on the expected consistent convergence behaviour is speculated to be caused by the interactive competition effects of the winner-take-all principle applied across the population of cells.