On the Optimization of a Class of Blackbox Optimization Algorithms

Blackbox Optimization(BBO) algorithms are candidate methods when knowledge of the problem is too incomplete to allow development of an eecient heuristic algorithm. Many BBOs have suucient ex-ibility to allow them to adapt to the varying circumstances they encounter. These exibilities include user-determined choices among alternative parameters , operations, and logic structures, and also the algorithm-determined alternative paths chosen during the process of seeking a solution to a particular problem. This paper presents a unifying framework for using this exibility to tailor a BBO paradigm to the problem at hand, dynamically adjusting the algorithm during the search. It demonstrates this approach applied to a genetic algorithm. The experimental results show the eeectiveness and robustness of the proposed approach.