You are here

Stochastic Global Optimization

Anatoly Zhigljavsky and Antanas Zilinskas
Publisher: 
Springer
Publication Date: 
2008
Number of Pages: 
262
Format: 
Hardcover
Series: 
Springer Optimization and Its Applications 9
Price: 
79.95
ISBN: 
978-0-387-74022-5
Category: 
Monograph
[Reviewed by
Tom Schulte
, on
02/25/2008
]

This excellent book is written for researchers interested in global optimization. It is not a gateway to the topic for unguided self-study. Readers of this book seeking to maximize the value of their time should come with a practical and theoretical graduate-level understanding of global optimization and statistical methods. However, the approach of carrying through from basic ideas to the most recent techniques will make this a valuable resource for the initiated. Population-based methods, random multistart, statistical inference about a minimum, and stochastic models about objective function are among the topics explored in this survey of stochastic global optimization.

Gathering together contemporary methods and theoretical developments in stochastic global optimization, this text presents four chapters. An initial chapter of basic principles and ideas gives the foundations of global optimization and stochastic methods. Chapters two and three build from an introduction to global random search. Also detailed are a variety of algorithms, statistical inference in random search, Markov random search methods and more. The concluding chapter presents statistical models and methods, including detailed sections on one- and multi-dimensional algorithms. These chapters are bolstered by over 300 references.

Aiming to be a point of departure for further exploration of various stochastic methods, this book strives only to explain clearly the main methodological principles and features of many methods. Detailed examples, implementation details, case studies, etc. are foregone in order to complete the taxonomy of the stochastic methods in a concise volume. This gives room for sophisticated techniques such as random and semi-random coverings, stratified sampling schemes, population based algorithms, and more. Also helpful is insight given into determining the efficiency of algorithms and the specific difficulties that can occur in applied global optimization.


Tom Schulte is a PhD candidate at Oakland University specializing in constraint programming.