Skip to main content
main-content

Über dieses Buch

This book offers the first comprehensive taxonomy for multimodal optimization algorithms, work with its root in topics such as niching, parallel evolutionary algorithms, and global optimization.

The author explains niching in evolutionary algorithms and its benefits; he examines their suitability for use as diagnostic tools for experimental analysis, especially for detecting problem (type) properties; and he measures and compares the performances of niching and canonical EAs using different benchmark test problem sets. His work consolidates the recent successes in this domain, presenting and explaining use cases, algorithms, and performance measures, with a focus throughout on the goals of the optimization processes and a deep understanding of the algorithms used.

The book will be useful for researchers and practitioners in the area of computational intelligence, particularly those engaged with heuristic search, multimodal optimization, evolutionary computing, and experimental analysis.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Introduction: Towards Multimodal Optimization

Abstract
Here we isolate the matter of this work within the large domain of optimization. We introduce a number of basic terms and algorithmic techniques in Sect. 1.1, prior to discussing different possible general aims of multimodal optimization in Sect. 1.2. Next, currently available evolutionary algorithms for multimodal optimization are discussed in Sect. 1.3 with the objective of establishing an improved taxonomy for these methods. Finally, Sect. 1.4 establishes the overall aims of the subsequent chapters.
Mike Preuss

Chapter 2. Experimentation in Evolutionary Computation

Abstract
In which we reflect on the current status of experimentation in evolutionary computation (Sect. 2.1) and beyond (Sect. 2.2). We then argue in favor of a methodology in Sect. 2.3, highlighting the need for a structured approach with well-defined aims, parameter settings, designs, and measures. Finally, Sect. 2.4 deals with the positive aspect of parameters: the possibility of adapting algorithms to concrete needs via new, more suitable parameter settings.
Mike Preuss

Chapter 3. Groundwork for Niching

Abstract
Here we establish a suitable definition of niching in evolutionary computation and approach the question of the potential of niching methods in optimization.
Mike Preuss

Chapter 4. Basin Identification by Means of Nearest-Better Clustering

Abstract
Here we first collect the most important objectives for a basin identification (and thereby clustering) algorithm in the optimization context and then propose a technique for detecting clusters in populations of search points that correspond to basins of attraction. We present this method early and defer literature review and comparison, as it builds the basis for several measurements and algorithms that will be provided in later chapters.
Mike Preuss

Chapter 5. Niching Methods and Multimodal Optimization Performance

Abstract
What is the concrete task of multimodal optimization methods and how can we compare the performance of these algorithms experimentally in a meaningful way? Which niching techniques are actually applied in existing algorithms, and how do they relate to the nearest-better clustering method? Specific benchmark suites and performance measures are still evolving and far from being mature. We review the current state and envision future developments in this respect.
Mike Preuss

Chapter 6. Nearest-Better-Based Niching

Abstract
Here we employ the nearest-better clustering basin identification method derived in a previous chapter for setting up two niching evolutionary algorithms. After doing parameter testing, we investigate how these algorithms perform in comparison to other recent methods for the all-global and one-global use cases by means of available benchmark suites.
Mike Preuss

Chapter 7. Summary and Final Remarks

Abstract
With the materials provided in the six previous chapters, we hope to have extended and deepened knowledge about evolutionary algorithms applied to multimodal blackbox optimization problems. It is now time to summarize the main insights, review what we achieved in comparison to the original goals listed in Sect. 1, and give some hints concerning promising future search directions. Before we turn to the last of these three issues, we will deal with the first (summary) in the context of the second (goals), as each of the chapters is tightly linked to one of the goals. The only exception to this is Sect. 2, which presents our view on the current state of an experimental methodology for research in evolutionary computation and related fields. It may be seen as a summary of the many experiences with experimental research the author has been involved with in the last few years. As the amount of available theory concerning evolutionary computation applied to multimodal problems is small, it is quite important to do experiments in a structured and well-defined way and we make an effort to respect this throughout the whole work.
Mike Preuss

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise