Vol. 4  Iss. 5
The Chemical Educator 
1999 Springer-Verlag New York, Inc. 
     

ISSN 1430-4171
http://journals.springer-ny.com/chedr
S 1430-4171(99) 04320-4 

Book Review  

Evolutionary Design by Computers
by Peter Bentley

 Reviewed by
Andrew Tuson
Department of Computing,
City University, Northampton Square, London EC1V OHB.
andrewt@soi.city.ac.uk  


Evolutionary Design by Computers, by Peter Bentley. Academic Press: Sidcup, Kent, England. ISBN 0 1208 9040 4. Includes CD-ROM containing programs and source code. 44.94.


Overview

The use of artificial intelligence (AI) methods in chemistry has become more common of late [1]. One subdiscipline of this, evolutionary computation (EC), concerns itself with applying the metaphor of natural evolution by natural selection to problem solving, specifically optimization.

Evolutionary Design by Computers is edited by Peter Bentley, an internationally known research fellow at University College London. Its focus is on "evolutionary design" the optimization and discovery of high-quality and possibly novel solutions to problems in design.

Relevance to AI/Chemistry Researchers

This book is aimed squarely at the EC research community, though in a manner that those with merely an interest in EC will find accessible. The sense of excitement currently in the field is conveyed very well by Bentley. However, though he does provide a sound overview of evolutionary computation, I do feel that the reader would do well to also consult one of the excellent textbooks available [2].

This book is a collection of contributions from the leading researchers in the evolutionary-design community. Each chapter either examines issues behind evolution and the creative design progress (e.g., David Goldberg's chapter: "The Race, the Hurdle, and the Sweet Spot"), or case studies on the applications of evolutionary design methods.

It is in the latter area where this book will be of greatest value to AI / chemistry researchers. Though none of the case studies concerns chemistry problems, they do constitute a repository of best practice in using evolutionary methods to solve difficult real-world design problems. As such, the book can be used as a source of ideas and inspiration for those trying to apply evolutionary methods to chemistry.

In summary, this collection provides an excellent overview of the state of the art in evolutionary design, and therefore it represents a highly useful resource for those with an interest in using evolutionary methods to solve design problems (chemists included), as well as those interested in evolutionary methods in general.

Relevance to Chemical Education

Given that this book is squarely aimed at the research community, it is not surprising that its relevance to chemical education is quite limited. In fact, given that the AI coverage in most chemistry courses is likely to be broad rather than deep in nature, a general AI text [3] would place EC in a much stronger setting, and would deal better with the important question of when evolutionary methods should be used instead of other AI approaches.

However, this does not mean that the book has no utility in the teaching of evolutionary and AI methods in chemistry. It could, for example, be used in coursework—an assignment based on a case study described in the text and how it could be applied to a chemical problem is a distinct possibility. In addition, the case studies could be used as the basis of student projects in the application of AI/evolutionary methods in chemistry.

Summary

This book is an excellent compendium of current work in the use of evolutionary methods for design. Even though it does not contain any material specific to chemistry, it does provide an excellent source of ideas. One example is the graph representation used in the design of robot morphology and controllers by Karl Sims; such graph representations could clearly be transferred to the representation and optimization of chemical structures.

In the education of chemists, the book can play at best a supplementary role as a source of case studies and project suggestions rather than be used as a course's primary textbook.

That said, I strongly recommend this book to anyone who has an interest in the application of evolutionary methods to demanding real-world problems in design.

References

[1] H. M. Cartwright, Applications of Artificial Intelligence in Chemistry; Oxford University Press: Oxford, UK, 1993.

[2] Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, 3rd ed.; Springer-Verlag: New York, 1996.

[3] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach; Prentice Hall: Englewood Cliffs, NJ,1996.