An Introduction to Genetic Algorithms (Complex Adaptive Systems)

Great.

Melanie Mitchell’s book “an introduction to Genetic Algorithms” explains what Genetic Algorithms are and how they work. It is somewhat outdated by now. However, that does not matter a whole lot since the book is focused on the foundations and the theory behind genetic algorithms and is academic in nature. This book is not a “cook-book” for Genetic Algorithms, and it does not have any practical examples or code that you can “borrow”. Being an academic book it goes into the theoretical foundations of Genetic Algorithms, it uses a fair amount of mathematics, and it backs up claims and discussions with references to research articles. At times the mathematics gets a little bit complex and you better know something about probability, functions, matrix algebra, vector/matrix notation, and infinite series.For those who do not know; Genetic Algorithms imitate aspects of the evolutionary process observed in nature to solve engineering problems and scientific problems. As such it can also shed light on the natural evolutionary processes (punctuated equilibria, the Baldwin effect, etc). In Genetic Algorithms you have a genetic representation, for example, a “chromosome” (bit string) and you simulate cross over (chromosome blending), mutations, fitness criteria, etc. Melanie Mitchell describes the use of Genetic Algorithms in scientific models, and how they can be used to simulate and explain evolution in nature, she describes different approaches to Genetic Algorithms, automatic programming, using Genetic Algorithms for prediction, and she explains how to use them to solve problems in Artificial Intelligence/Computer Science, and she also describes how to use them with evolving Neural Networks.What I liked about the book is that despite the fact that it is only 200 pages it covers a lot of ground. The book is well organized, well written, interesting, and concise. I read this book because I was interested in finding out whether I could use some form of Genetic Algorithm to solve some optimization problems at work. Therefore it might not have been exactly the right book for me. At the same time I found the book to be quite interesting and I liked the learning experience. You should understand Genetic Algorithms before you use them anyway. The book is 16 years old by now, perhaps too academic for some people’s taste, and I believe I found a term that was left out in a derivation, so I’ll give a four star rating, but I enjoyed reading it.

This book primarily deals with the theoretical side of genetic algorithms. If you are looking for practical knowledge of how to implement a GA you should look elsewhere. For all intents and purposes this is a textbook. It's heavy on theory and proofs, but doesn't always explain everything in depth (that's what class time is for). There are problems at the end of each chapter that can be assigned to students.There are case studies of many academic projects that seem to drone on forever and aren't really that useful in helping you learn how to write your own GA. Chapter 1 gives an overview and provides all of the appropriate terminology. Chapter 5 gives an high-level overview of how to implement a GA. Those are the 2 must-read chapters, all of the others can be used as torture for CS students.To recap, if you're teaching a class in artificial intelligence this book is good. If you're trying to figure out how to implement a GA to solve a practical problem not so good. That evens out to 3 stars for my rating. I recommend searching the web, there are a few good sites on GA programming.

First it must be said that the book is not an introduction that the non-scientist will easily understand. Some knowledge of computer programming is assumed. It acknowledges this in the last paragraph of the preface. Many of the notations in the book are unfamiliar to business or financial readers. There is no mathematics beyond algebra so the aforementioned prerequisites are the main hills to climb.Mitchell's book is an overview of genetic algorithm analysis techniques as of 1996. The author gives a history of pre-computer evolutionary strategies and a summary of John Holland's pioneering work. A description of the basic terminology is presented and examples of problems solved using a GA (such as the prisoner's dilemma). The second chapter discusses evolving programs in Lisp and cellular automata. Also included in this chapter is a discussion of predicting dynamical systems. This was the section that has the most interest for me. Also interesting was the summary in this chapter about putting GAs into a neural network so that the ANNs could evolve.The fifth chapter discusses when to employ a GA for maximum success. I appreciate the clearly thought out discussion of when to choose a GA for a problem. Sometimes authors of these types of books mimic the man with a hammer that thinks everything looks like a nail.

This book is brief and to the point. You won't find here pages of source code that you could have easily ftp'd yourself. What you will find is solid theory in a mere 224 pages. This is the quickest and best way to get up to speed on GA's there is. Which is why it is a standard textbook in the field.