An Adaptive Recombination-Based Extension Of The Imoaco(R) Algorithm

Commonly, Ant Colony Optimization algorithms have been applied to the solution of single- and multi-objective optimization problems (MOPs). However, in recent years, a number of approaches have been proposed to solve problems with continuous search spaces. One remarkable proposal is the indicator-based Multi-Objective Ant Colony Optimizer for continuous search spaces (iMOACO(R)) which is based on the ACO(R) algorithm and the R2 performance indicator, aiming to solve continuous many-objective optimization problems (i.e., MOPs having more than three objective functions). In previous work, we presented an extension of iMOACO(R), called iMOACO(R)-R, in which a recombination operator is employed for solution construction with a fixed, externally-specified probability. In the present work, we introduce a further adaptive variation, called iMOACO(R)-AR, in which the frequency of applying recombination is dynamically adapted based on the recent past performance of the recombination operator. Our proposal is compared to iMOACO(R) and iMOACO(R)-R using 64 standard problems from the multi-objective optimization literature with a number of objectives ranging from 3 to 10. Experimental results show that iMOACO(R)-AR outperforms iMOACO(R) and iMOACO(R)-R in most of the test problems.