ABSTRACT

This paper presents a C++ application programming interface for a co-evolutionary algorithm for solution and scenario generation in stochastic problems. Based on a two-space biased random-key genetic algorithm, it involves two types of populations that are mutually impacted by the fitness calculations. In the solution population, high-quality solutions evolve, representing first-stage decisions evaluated by their performance in the face of the scenario population. The scenario population ultimately generates a diverse set of scenarios regarding their impact on the solutions. This application allows the straightforward implementation of this algorithm, where the user needs only to define the problem-dependent decoding procedure and may adjust the risk profile of the decision-maker. This paper presents the co-evolutionary algorithm and structures the interface. We also present some experiments that validate the impact of relevant features of the application.

摘要

本文提出了一个 C++ 应用程序编程接口，用于在随机问题中用于解决方案和场景生成的协同进化算法。基于二空间偏向随机密钥遗传算法，它涉及受适应度计算相互影响的两类种群。在解决方案群体中，高质量的解决方案不断发展，代表了第一阶段的决策，通过它们在场景群体中的表现来评估。场景群体最终会生成一组关于它们对解决方案的影响的不同场景。该应用程序允许该算法的直接实施，其中用户只需要定义与问题相关的解码过程，并且可以调整决策者的风险状况。本文介绍了协同进化算法并构造了接口。我们还提供了一些实验来验证应用程序相关功能的影响。