Reevaluating Immune-Inspired Hypermutations Using the Fixed Budget Perspective

摘要: Different studies have theoretically analyzed the performance of artificial immune systems in context optimization. It has been noted that, comparison with evolutionary algorithms and local search, hypermutations tend to be inferior on typical example functions. These used expected optimization time as criterion cannot explain why are popular spite these proven drawbacks. Recently, a different perspective for theoretical analysis introduced, concentrating within fixed frame instead needed Using this we reevaluate somatic contiguous inverse fitness-proportional random search one well-known function which is known efficient much more than respect time. We prove depending choice initial point, can by far outperform given frame. This insight helps success seemingly inefficient mutation operators practice. Moreover, demonstrate how benefit from obtained insights designing hybrid heuristics.