Modern optimization strategies such as evolutionary algorithms, ant colony algorithms, Bayesian optimization techniques, etc. come with several parameters that steer their behavior during the optimization process. To obtain high-performing algorithm instances, automated algorithm configuration techniques have been developed. One of the most popular tools is irace, which evaluates configurations in sequential races, making use of iterated statistical tests to discard poorly performing configurations. At the end of the race, a set of elite configurations are selected from those survivor configurations that were not discarded, using greedy truncation selection. We study two alternative selection methods: one keeps the best survivor and selects the remaining configurations uniformly at random from the set of survivors, while the other applies entropy to maximize the diversity of the elites. These methods are tested for tuning ant colony optimization algorithms for traveling salesperson problems and the quadratic assignment problem and tuning an exact tree search solver for satisfiability problems. The experimental results show improvement on the tested benchmarks compared to the default selection of irace. In addition, the obtained results indicate that non-elitist can obtain diverse algorithm configurations, which encourages us to explore a wider range of solutions to understand the behavior of algorithms.
翻译:现代优化策略, 如进化算法、 蚁群算法、 巴耶斯优化技术等, 带有指导优化过程中行为的若干参数。 为了获得高性能算法实例, 已经开发了自动算法配置技术。 最受欢迎的工具之一是 irace, 用来评估连续种族的配置, 使用迭代统计测试来丢弃不良的配置。 在比赛结束时, 利用贪婪的拖网选择, 从未丢弃的幸存者配置中, 选择了一组精英配置。 我们研究两种替代选择方法: 一种是让最优秀的幸存者, 从幸存者组群中随机选择其余的配置, 而另一种是应用昆虫来尽量扩大精英的多样化。 这些方法经过测试, 用于调整流动销售人员问题和二次分配问题的蚂蚁群优化算法, 并调整可容纳性问题的精确树搜索解答器。 实验结果显示, 与默认选择 irace 相比, 测试的基准有所改善 。 此外, 获得的结果显示, 非精英 能够获取多种的算法配置, 从而鼓励我们探索范围 。