Evolutionary Computation. A Unified Approach. Kenneth A. De Jong. (2006, MIT Press.) £32.95, $50.00, 256 pages

Abstract

Evolutionary Computation. A Unified Approach. Kenneth A. De Jong. (2006, MIT Press.) £32.95, $50.00, 256 pages. Evolutionary computation (EC) is nowadays a very dynamic research field. EC arose as a research area from the union of separate efforts on the 1960s [1, 3, 5]. Even in our time, the necessity to unify concepts, ideas, methods, schemes, paradigms, and the like is still one of the key issues in the area, and Kenneth De Jong addresses it in his book. De Jong proposes to merge ‘‘subspecies’’ (as he calls them) such as genetic algorithms (GAs), evolution strategies (ESs), and evolutionary programming (EP) into one more general class of evolutionary algorithms (EAs). The book is conveniently organized in such a way that the reader is able to get a concise view of an EA, its main components, and their corresponding effects on the algorithm’s behavior. Furthermore, some theoretical results complement the knowledge about how an EA works. Finally, some paths of current and future research are examined. In the following, we provide a more detailed review of each chapter in the book. Chapter 1 presents the evolutionary process, based on a simple natural-based example, which is immediately translated into a very simple evolutionary system called EV. EV is used through the chapter to show its behavior and performance when solving a simple one-dimensional landscape problem. Using an incremental approach, a more complicated problem is used to test EV now with additional features. A chronological set of landmark events in EC history is presented in Chapter 2. The 1960s are presented as the years when the first EAs were proposed, all of them based on the same biological inspiration, natural evolution. After describing the 1970s as the time when the first empirical and theoretical studies emerged and also the 1980s, of which the application of EAs was the main feature, the book identifies the 1990s as the time when researchers of different groups started to interact with each other, providing more complete and robust studies about an area known as EC. Finally, the twenty-first century is seen as the time of maturity and also the time to tackle even more complex problems and to solidify our knowledge about EAs. Although an EA is always based on the idea of natural evolution and the survival of the fittest, three main paradigms are identified, evolutionary programming [1], evolution strategies [6], and genetic algorithms [2]. Similarities and differences are discussed in Chapter 3. A small performance comparison among these paradigms helps to understand how different the corresponding performance can be. Again, the simplicity of the example leads to a very clear explanation and understanding of each EA. Chapter 4 focuses on describing the elements of an EA and how they interact. The role of the parent and offspring populations is explained. After that, the two types of selection mechanisms (parent and survival selection) are presented and some schemes are enumerated. To end this chapter, two issues are addressed: (1) the influence of the level of representation of solutions, and (2) the role of recombination and mutation as the way to generate new solutions. Chapter 5 deals with EAs now as problem solvers. An EA is presented as a problem-independent parallel adaptive search mechanism that has to be designed for the problem to be solved. Then,