Optimization problems frequently appear in any scientific domain. Most of the times, the corresponding decision problem turns out to be NP-hard, and in these cases genetic algorithms are often used to obtain approximated solutions. However, the difficulty to approximate different NP-hard problems can vary a lot. In this paper, we analyze the usefulness of using genetic algorithms depending on the approximation class the problem belongs to. In particular, we use the standard approximability hierarchy, showing that genetic algorithms are especially useful for the most pessimistic classes of the hierarchy

翻译：优化问题频繁出现在任何科学领域。大多数情况下，对应的决策问题被证明是NP难的，而在这些情况下，遗传算法常被用于获得近似解。然而，不同NP难问题的近似难度可能存在很大差异。本文根据问题所属的近似类别，分析了使用遗传算法的有效性。特别地，我们利用标准的近似性层级，表明遗传算法对于层级中最悲观的那些类别尤为有效。