Abstract
The choice of which objective functions, or benchmark problems, should be used to test an optimization algorithm is a crucial part of the algorithm selection framework. Benchmark suites that are often used in the literature have been shown to exhibit poor coverage of the problem space. Exploratory landscape analysis can be used to quantify characteristics of objective functions. However, exploratory landscape analysis measures are based on samples of the objective function, and there is a lack of work on the appropriate choice of sample size needed to produce reliable measures. This study presents an approach to determine the minimum sample size needed to obtain robust exploratory landscape analysis measures. Based on reliable exploratory landscape analysis measures, a self-organizing feature map is used to cluster a comprehensive set of benchmark functions. From this, a benchmark suite that has better coverage of the single-objective, boundary-constrained problem space is proposed.
Highlights
The field of computational intelligence is inordinately powered by empirical analysis.Without theoretical derivations for the performance of optimization algorithms, one must compare said algorithms against one another by analyzing their performance on a collection of benchmark problems
The measures which have platykurtic distributions, i.e., wide and flat distributions, are the dispersion and level-set feature sets. These platykurtic distributions indicate that the number of sample sizes needed to produce robust exploratory landscape analysis (ELA) measures often differs for different benchmark functions
The larger the sample size used in exploratory landscape analysis, the more accurate the characterization of the objective function
Summary
The field of computational intelligence is inordinately powered by empirical analysis.Without theoretical derivations for the performance of optimization algorithms, one must compare said algorithms against one another by analyzing their performance on a collection of benchmark problems. The problem of selecting which algorithm to use is non-trivial and is limited to computational intelligence. The algorithm selection problem was formalized by Rice in 1976 [1]. Rice’s framework defines four components, namely the problem space, the feature space, the algorithm space, and the performance measure space. The problem space contains all the possible problems that can exist for a particular problem type. The feature space consists of all possible measures that can describe the characteristics of problems found in the problem space. The algorithm space consists of all possible algorithms that can be used to solve the problems found in the problem space. Benchmark problems and benchmark suites in single-objective, continuousvalued, boundary-constrained optimization are discussed. Algorithms 2021, 14, 78 mark suites is discussed.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have