Oppostition-based evolutionary algorithms: toward solving high-dimensional optimization problems efficiently

科学或工程问题的建模通常会导致优化问题。优化为各种各样的应用程序（从工程设计到面向服务的应用程序，如医疗保健、金融和交通）中的决策制定提供了正式的基础。在本提案中，进化算法（EAs）将被研究，以解决这些优化问题，否则难以或不可能解决的经典方法。解决混合变量、动态环境、多目标约束函数和非分析函数的问题是ea突出能力的例子。但是目前，ea由于其进化性质，在计算上是昂贵的。此外，ea还存在维度问题。这意味着随着搜索空间维数的增加，它们的性能会迅速下降。在研究和设计过程中，快速解决问题是非常可取的。因此，ea的加速是科学界在更短的时间间隔内解决大规模问题的一个重要问题。本研究开发了新的ea加速方案，以及高维问题特别需要的智能采样方法。将这些方案和采样方法结合起来，将产生更有效和鲁棒的方法来加速已知的进化算法并有效地解决高维问题。对于耗时的优化问题，对函数求值的简化要求很高。求解高维昂贵优化问题是一个具有挑战性的研究领域。这一建议旨在对这一领域作出重大贡献。它将为高素质人才提供一个非常好的培训环境，他们可以成为加拿大未来在科学和工程应用优化技术方面的领导者和先驱研究人员。这些应用依赖于在产品设计或解决科学问题期间成功地找到100个参数的最佳值。

Modeling of a scientific or engineering problem often leads to an optimization problem. Optimization provides a formal basis for decision making in a wide variety of applications, ranging from engineering design to service oriented applications such as healthcare, finance, and transportation. In this proposal, Evolutionary Algorithms (EAs) will be investigated to solve those optimization problems, which are otherwise difficult or impossible to solve by classical methods. Solving problems with mixed-type variables, dynamic environments, multi-objective constrained functions, and non-analytical functions are examples that highlight the outstanding capabilities of EAs. But currently, EAs are computationally expensive because of their evolutionary nature. Furthermore, EAs suffer from the problem of dimensionality. This means that their performance deteriorates quickly as the dimensionality of the search space increases. Rapid solutions of the problems are highly desirable during research and design processes. As a consequence, acceleration of EAs is a significant issue of importance for the scientific community to solve large-scale problems in a smaller time interval. This research develops new acceleration schemes for EAs, and also smart sampling methods, which are needed particularly for high-dimensional problems. Combining these proposed schemes and sampling methods would result in more efficient and robust approaches to accelerate well-known evolutionary algorithms and effectively solve high-dimensional problems. Reduction of function evaluations for time-consuming optimization problems is highly demanding. Solving high-dimensional expensive optimization problems is a challenging research area. This proposal aims to make significant contributions to this field. It would provide a very good training environment for highly qualified personnel, who can become Canada's future leaders and pioneering researchers in optimization techniques for applications in science and engineering. These applications depend on successfully finding optimum values for a hundred parameters during design of a product or solving a scientific problem.