|
|
|
|
|
|
|
10. Interim and Prospectus |
|
|
|
|
|
|
|
|
Adaptation in Natural and Artificial Systems,after a seven-year gestation, made its appearance in 1975. It is now 1991, and much has happened in the interim. Topics that were speculative in 1975 have been carefully explored; extensions, applications, and new areas of investigation abound. More than 150 papers were submitted to the 1991 International Conference on Genetic Algorithms (Belew and Booker 1991), and several new books have been written about genetic algorithms (e.g., Davis 1987 or Davis 1991). There is even a textbook (Goldberg 1989). Most of this new research has been reported in the published proceedings of the genetic algorithm conferences of 1985, 1987, 1989, and 1991 (Grefenstette 1985, Grefenstette 1987, Schaffer 1989, and Belew and Booker 1991) and is readily accessible there, so I will not attempt to review it herethe review would be, at best, little more than an annotated listing. Instead, I'll follow the pattern of the rest of the book, using this new chapter to report on lines of research I've pursued since 1975. A new edition also provides an opportunity to correct errors in the original edition. Most of these are simple and innocuous, but an error in one proof, discovered and corrected by Dr. Daniel Frantz, is subtle and important. By good fortune, after the correction the theorem involved stands as stated. Finally, a new chapter offers an opportunity to look further into the future; this too I'll attempt. |
|
|
|
|
|
|
|
|
Classifier systems, a specialization of chapter 8's "broadcast language," are a vehicle for using genetic algorithms in studies of machine learning. Classifier systems were introduced in Holland 1976 and were later revised to the current "standard" form in Holland 1980. There is a comprehensive description of the standard form, with examples, in Holland 1986, but there are now many variants (see Belew and Booker 1991). A classifier system is more restricted than the broadcast language in just one |
|
|
|
|
|