A survey of diversity-oriented optimization

The concept of diversity plays a crucial role in many optimization approaches: On the one hand, diversity can be formulated as an essential goal, such as in level set approximation or multiobjective optimization where the aim is to find a diverse set of alternative feasible or, respectively, Pareto optimal solutions. On the other hand, diversity maintenance can play an important role in algorithms that ultimately search for single optimal solutions. Examples are dynamical optimization and global optimization of multimodal objective functions, where diversity maintenance during a population-based search process increases the robustness of optimization algorithms or heuristics. While the motivations to study diversity and its maintenance can be various, some salient issues reoccur such as effective strategies for diversity maintenance, indicators used to measure diversity, and the study of dynamic processes on set-valued state spaces. Although there is a growing attention to methods that address diversity as a search objective, the research is so far spread out across various disciplines and research schools who have developed independently terminologies and classification schemes, making it difficult to find relations between different works. Although there is an increasing interest in diversity-oriented search a broad survey of this topic is missing so far and to provide it will be the goal of this work. This survey is intended to develop an integrated view of diversity-oriented optimization algorithms. Rather than going into details of implementations and adding new methods, the aim is to develop a systematic classification scheme. To integrate the various layers of algorithm design and various terminologies and methods that were developed in different research schools/areas, an ontology will be developed with the intention to ease the classification of existing and future algorithms in this field and identify overlapping and related areas of research. This work is structured in two parts: The first part provides a review of research in diversity-oriented optimization, looking at it from different angles. In the second part, an ontology that integrates these views is developed and existing results are discussed in the light of this ontology.