This book is a collection of papers on the general subject of evolutionary multiobjective optimization. The work presented is mostly the result of a 2006 workshop by the same title. This area of research has been around for quite some time, but has received renewed attention in recent years due to its ability to address complex optimization problems in multiple application areas, and for its use as a way of uncovering new design principles and understanding the true structure of complex problems.
The first chapter of the book provides an overview of the entire field from a historical and technical perspective, focusing primarily on multiobjective optimization problems, solutions, and new directions. It also includes a detailed list of what is covered in the other chapters, providing the glue that connects the separate contributions. This is a very readable introduction that even casual readers will find valuable. The rest of the book gets into the technical details of the approaches and solutions presented and should be of interest to more sophisticated readers.
The contributed chapters in the book are divided into four parts. Part 1 covers multiobjective optimization from a problem-solving perspective. This part has the largest number of chapters. It includes coverage of coevolutionary algorithms, constrained optimization, and problems such as minimum spanning tree and single-source shortest path, where a multiobjective optimization approach based on evolutionary algorithms is demonstrated to be superior to a single-objective evolutionary optimization algorithm.
Part 2 focuses on machine learning with multiple objectives. It includes contributions on supervised learning, genetic programming, and rule mining.
Part 3 deals with the use of multiple objectives in design and engineering. The applications covered are engineering design of devices such as cranes, truss design, and welded-beam design. Various forms of visualization are also covered.
The last part of the book is on scalability. It focuses on topics such as methods for fitness assignment, modeling regularity, and handling a large number of objective functions.