【论文】黑箱连续优化问题的算法选择：方法与挑战的综述

Algorithm selection for black-box continuous optimization problems: A survey on methods and challenges

黑箱连续优化问题的算法选择：方法与挑战的综述

Abstract 

摘要

        Selecting the most appropriate algorithm to use when attempting to solve a black-box continuous optimization problem is a challenging task. Such problems typically lack algebraic expressions, it is not possible to calculate derivative information, and the problem may exhibit uncertainty or noise. In many cases, the input and output variables are analyzed without considering the internal details of the problem. Algorithm selection requires expert knowledge of search algorithm efficacy and skills in algorithm engineering and statistics. Even with the necessary knowledge and skills, success is not guaranteed.
        在试图解决黑盒连续优化问题时，选择最合适的算法是一项具有挑战性的任务。这些问题通常缺乏代数表达式，不可能计算导数信息，并且问题可能表现出不确定性或噪声。在许多情况下，分析输入和输出变量时没有考虑问题的内部细节。算法选择需要搜索算法功效的专家知识以及算法工程和统计学方面的技能。即使有必要的知识和技能，也不能保证成功。

In this paper, we present a survey of methods for algorithm selection in the black-box continuous optimization domain. We start the review by presenting Rice’s (1976) selection framework. We describe each of the four component spaces – problem, algorithm, performance and characteristic – in terms of requirements for black-box continuous optimization problems. This is followed by an examination of exploratory landscape analysis methods that can be used to effectively extract the problem characteristics. Subsequently, we propose a classification of the landscape analysis methods based on their order, neighborhood structure and computational complexity. We then discuss applications of the algorithm selection framework and the relationship between it and algorithm portfolios, hybrid meta-heuristics, and hyper-heuristics. The paper concludes with the identification of key challenges and proposes future research directions.
在本文中，我们提出了黑箱连续优化领域的算法选择方法的调查。我们首先介绍Rice（1976）的选择框架。我们描述了四个组成部分的空间-问题，算法，性能和特点-在黑盒连续优化问题的要求。其次是探索性的景观分析方法，可以用来有效地提取问题的特征进行检查。随后，我们提出了一个分类的景观分析方法的顺序，邻域结构和计算复杂性的基础上。然后，我们讨论了算法选择框架的应用程序和它与算法组合，混合元算法，超算法之间的关系。本文最后指出了关键挑战，并提出了未来的研究方向。

Keywords 关键词

Algorithm selection 算法选择

Black-box continuous optimization 黑箱连续优化

Empirical performance models 经验性能模型

Exploratory landscape analysis 探索性景观分析

Performance prediction 性能预测

Problem hardness measures 问题难度度量

https://www.sciencedirect.com/science/article/pii/S0020025515003680

1. Introduction 

1.介绍

        The objective of an optimization problem is to improve a measure of performance or cost – the output variable – by adjusting the values of the input variables. Typically, the optimization problem is represented as a function that maps the inputs to the output, subject to constraints. When both the input and output variables are real numbers, the problem is referred to as a continuous optimization problem. Such problems are common in science, engineering, finance, and other fields [81].
        优化问题的目标是通过调整输入变量的值来改善性能或成本的度量-输出变量。

        通常，优化问题被表示为将输入映射到输出的函数，受到约束。

        当输入和输出变量都是真实的数时，该问题被称为连续优化问题。

        这些问题在科学、工程、金融和其他领域都很常见[81]。

Many continuous optimization problems lack algebraic expressions and may not even have a precise goal. Topologically, the problem may present local and global optima or discontinuities where it is not possible to calculate derivative information. The problems frequently incorporate dependencies