Liger: A cross-platform open-source integrated optimization and decision-making environment
Liger: A cross-platform open-source integrated optimization and decision-making environment
复制标题
DOI：
10.1016/j.asoc.2020.106851
复制
发表时间：
2020-11
期刊：
影响因子：
--
通讯作者：
João A. Duro;Yiming Yan;I. Giagkiozis;S. Giagkiozis;Shaul Salomon;Shaul Salomon;Daniel C. Oara;
中科院分区：
文献类型：
--
作者：
João A. Duro;Yiming Yan;I. Giagkiozis;S. Giagkiozis;Shaul Salomon;Shaul Salomon;Daniel C. Oara;
涉及多个冲突目标的现实世界优化问题通常使用多目标优化来最好地解决，因为这为决策者提供了一系列权衡解决方案。然而，使用多目标优化算法的复杂性往往阻碍了优化过程。要知道哪种优化算法最适合给定的问题，甚至要选择哪种设置参数，需要有人成为优化专家。缺乏现成、易于使用和透明的支持软件可能导致设计时间延长和成本增加。为了应对这些挑战，Liger被提出。Liger的设计目的是让非优化专家在工业中易于使用。用户通过可视化编程语言与Liger交互，以创建优化工作流，使用户能够解决优化问题。Liger包含一个称为Tigon的新优化库。该库利用设计模式的概念，通过使用简单的可重用操作符节点来实现优化算法的组合。该库提供了各种各样的多目标进化算法，涵盖了进化计算中的不同范式;支持各种各样的问题类型，包括支持同时使用多种编程语言来实现优化模型。此外，Liger功能可以通过插件轻松扩展，这些插件目前提供对最先进的可视化工具的访问，并负责管理图形用户界面。最后，新的用户驱动的交互功能，以促进决策过程中，并证明了控制工程优化问题。
Real-world optimization problems involving multiple conflicting objectives are commonly best solved using multi-objective optimization as this provides decision-makers with a family of trade-off solutions. However, the complexity of using multi-objective optimization algorithms often impedes the optimization process. Knowing which optimization algorithm is the most suitable for the given problem, or even which setup parameters to pick, requires someone to be an optimization specialist. The lack of supporting software that is readily available, easy to use and transparent can lead to increased design times and increased cost. To address these challenges, Liger is presented. Liger has been designed for ease of use in industry by non-specialists in optimization. The user interacts with Liger via a visual programming language to create an optimization workflow, enabling the user to solve an optimization problem. Liger contains a novel optimization library known as Tigon. The library utilises the concept of design patterns to enable the composition of optimization algorithms by making use of simple reusable operator nodes. The library offers a varied range of multi-objective evolutionary algorithms which cover different paradigms in evolutionary computation; supports a wide variety of problem types, including support for using more than one programming language at the time to implement the optimization model. Additionally, Liger functionality can be easily extended by plugins that currently provide access to state-of-the-art visualisation tools, and are responsible for managing the graphical user interface. Lastly, new user-driven interactive capabilities are shown to facilitate the decision-making process, and are demonstrated on a control engineering optimization problem.