Illustrating Evolutionary Computation with Mathematica (The Morgan Kaufmann Series in... at Dominant Systems Book Store

Living organisms manage to solve all kinds of deviously complex problems with a natural simplicity that leaves programmers speechless. Incorporating techniques based on principles elaborated by Darwin and his intellectual descendents, a new generation of hackers has tackled hairy challenges with surprising success. Christian Jacob introduces interested programmers and scientists to these tools in Illustrating Evolutionary Computation with Mathematica, translated from German by the author. The basics of biological evolution through mutation and adaptation are covered quickly before they are adapted themselves to the purposes of computer-aided problem solving. Jacob then explores the fundamentals of evolutionary computing through well-illustrated examples and a good balance of text, formulae, and code. Genetic algorithms, evolutionary strategies, and finite state automata each get their share of attention and integration with Evolvica, Jacob's Mathematica-based genetic programming system. The system and Web enhancements to the book are available through the University of Calgary's site and are essential for getting the most from the text. The last few chapters cover advanced applications like the classic "hungry ants" programs, cellular automata, and artificial plant evolution, suggesting further possibilities for this programming frontier. Illustrating Evolutionary Computation with Mathematica is an excellent introduction and handbook for those wishing to harness the power of this vigorous new hybrid. --Rob Lightner

Customer Reviews & Comments
Like the previous reader, I must applaud this book. While I know Mathematica well, I am no mathematician and I have no patience for books that do not explain the foundations of their programs or functions. The author here, Jacob, does an excellent job of introducing the reader gradually to the different concepts of simulating evolution. As you can download the Mathematica notebooks and run them on your own computer, this quickly becomes a fun and interactive book. The book starts with simple selection processes for reproduction. Select shapes, colors or features and see a next generation evolve! This can be a fun game. See breeding and mutation be used to search for an optimum of a three-dimensional function, where the reader knows the global optimum, while different "populations" try to find it by evolutionary methods-mutating or breeding to a different spot, which they evaluate and according to its height be successful in the passing of their genes or not. Other fun chapters include evolutionary production of mobiles and flowers. The culmination is in the evolution of algorithms. This evolves small programs for searching for food in a maze. The successful programs "breed," "mutate," and reproduce, while the unsuccessful ones starve and die. The result is a complex path toward better algorithms for searching for food. Part of the value of this book for me is that it really shows the limits of evolutionary analysis. You can simulate the successes--the butterflies that do manage to change colors to avoid falling easy prey when the environment changes; the evolutionary mechanisms that find the global optimum of a function-but there is no concrete way to determine or describe their efficiency ex ante. This is a major failure of evolutionary analysis generally, rather than a drawback of the book. If anything, the book deserves credit for making this failure understandable, although Jacob does not spend time exploring or solving the problem of determining evolutionary fitness. [To put it in an example, suppose there are two evolutionary mechanisms. An organism can evolve by mutation or by reproduction. Mutation is the random change of some individuals in the population, and the change makes them either more or less successful in their environment. Reproduction means parents producing an offspring by mixing their features, and the different offspring will have different degrees of success in their environment. We can simulate their operation in a hypothetical environment, by for example, saying that the background foliage changes color and organisms have different probabilities of being eaten by predators depending on their color. We run the simulation and see which evolutionary mechanism adapts to the new environment faster and better. Nevertheless, we cannot conclude that the evolutionary mechanism that won this test will win every test. Needless to say, when designing evolutionary systems this conclusion is crucially necessary. If we are designing a computer search program, should we have it "mutate" or "reproduce"? Since we do not know the challenges it will face (the changes in the environment that it must overcome) we cannot evaluate its success ex ante.] With the caveat of not exploring measurements of the success (fitness) of different evolutionary mechanisms, this is a spectacular book. It is worth comparing it with the books of the various biologists, who simply offer examples of evolutionary changes from the past or hypotheses of evolutionary explanations for various phenomena. Those are speculations of amateurs compared to the experimentation and verification that Jacob's approach offers. That the field is not ready for rigorous conclusions is unfortunate, but something that is no fault of this author.

Back To Top