一种求解多目标无约束0-1二次规划问题的文化基因算法

A memetic algorithm for multiobjective unconstrained binary quadratic programming

针对多目标无约束0-1二次规划问题,提出一种文化基因算法。该算法采用基于分解的多目标演化算法框架,能够获得分布均匀的非占优解;同时,采用一种简单、有效的禁忌搜索,能够利用更多问题相关的信息,获得质量更优的非占优解。该算法在优化的过程中能够动态地平衡多样性与收敛性。实验结果证明该算法能够很好地求解多目标无约束0-1二次规划问题,并且性能优于目前求解该问题较先进的算法。

This paper proposes a memetic algorithm （MA） for multiobjective unconstrained binary quadratic programming problem. In MA, multiobjective evolutionary algorithm based on decomposition （MOEA/D） framework is adopted to obtain well-distributed nondominated solutions. At the same time, More problem-specific knowledge can be extracted by using a simple and effective tabu search （TS）, and high-quality solutions can be generated. Therefore, MA can balance the diversity and convergence well during the whole optimization process. Experimental results show that MA outperforms the previous state-of-the-art algorithm for mUBQP cases.