|
|
|
|
|
|
|
Preface to the 1992 Edition |
|
|
|
|
|
|
|
|
When this book was originally published I was very optimistic, envisioning extensive reviews and a kind of "best seller" in the realm of monographs. Alas! That did not happen. After five years I did regain some optimism because the book did not "die," as is usual with monographs, but kept on selling at 100-200 copies a year. Still, research in the area was confined almost entirely to my students and their colleagues, and it did not fit into anyone's categories. "It is certainly not a part of artificial intelligence'' and "Why would somebody study learning by imitating a process that takes billions of years?" are typical of comments made by those less inclined to look at the work. |
|
|
|
|
|
|
|
|
Five more years saw the beginnings of a rapid increase in interest. Partly, this interest resulted from a change of focus in artificial intelligence. Learning, after several decades at the periphery of artificial intelligence, was again regarded as pivotal in the study of intelligence. A more important factor, I think, was an increasing recognition that genetic algorithms provide a tool in areas that do not yield readily to standard approaches. Comparative studies began to appear, pointing up the usefulness of genetic algorithms in areas ranging from the design of integrated circuits and communication networks to the design of stock market portfolios and aircraft turbines. Finally, and quite important for future studies, genetic algorithms began to be seen as a theoretical tool for investigating the phenomena generated by complex adaptive systemsacollective designation for nonlinear systems defined by the interaction of large numbers of adaptive agents (economies, political systems, ecologies, immune systems, developing embryos, brains, and the like). |
|
|
|
|
|
|
|
|
The last five years have seen the number of researchers studying genetic algorithms increase from dozens to hundreds. There are two recent innovations that will strongly affect these studies. The first is the increasing availability of massively parallel machines. Genetic algorithms work with populations, so they are intrinsically suited to execution on computers with large numbers of processors, using a processor for each individual in the population. The second innovation is a unique interdisci- |
|
|
|
|
|