• Contribute
• What's New
• Submit News

9.5

Author: Gary William Flake Excerpt: In this book [Flake] develops in depth the simple idea that recurrent rules can produce rich and complicated behaviours...Flake argues that it is the computational properties of interactions that account for much of what we think of as "beautiful" and "interesting"...Flake explores what he considers to be today's four most interesting computational topics: fractal, chaos, complex systems and adaption. .: Buy at Amazon.com :. .: Buy at Amazon.co.uk :.

When a book tries to cover as much as The Computational Beauty of Nature (CBN) it often manages to cover too little of too much, but somehow Flake managed to pull CBN off beautifully. This book is an excellent read: it can be followed sequentially chapter-by-chapter, or can be read randomly a chapter at a time. The book covers topics such as fractals, l-systems, strange attactors, predator-prey relationships, coevolution, the iterated prisoner dilemma, genetic algorithms, perceptrons, neural networks, cellular automata, artificial life and a plethora of other topics brilliantly.

The book starts by just looking at numbers and their various properties, and quickly progresses to the idea of incomputability. From there on to fractals and some of the most interesting chapters in the book. I'm a sucker for fractals: I find their concept mind-boggling - such infinite complexity from such a simple algorithm. CBN covers simple fractals such as the Cantor Set and the Koch Curve nicely, before moving on to L-Systems (this chapter prompted me to update the Generation5 L-Systems program) and finally on to complicated fractal systems like IFS, and the Mandelbrot and Julia sets.

What really stuck out about CBN was how Flake backed it up with programs that he'd written. The book doesn't come with a disk or CD-ROM, but allows you to download the binaries/source code from the Internet if you wish. Flake wrote nearly 40 programs to go with CBN that allowed the user to interactively explore virtually every topic covered. While I couldn't get the code to compile under Windows 9x (using VC5 or VC6), I'm sure with a little tweaking it wouldn't be too hard to port over.

The book covers more complexity and strange attractors before moving on to what we're interested in - Artificial Life and the concepts of Artificial Intelligence. The cellular automata chapter looks at Wolfram's 1D CA (see CA Creator), Conway's Life and natural occurences of CAs. Flake then goes on to look at self-organization and emergent behaviour such as artificial termites, Langton's Ants, and the famous Boids example. Interestingly, Flake added a fourth boid rule that sometimes caused the boids exhibit the V-shaped flying pattern one can see in nature. In the following chapter, Flake goes into some considerable depth with cooperations and competition using the IPD as the example. From there, there is a short chapter covering simple perceptrons and associative memory.

It is in the "Adaption" section that most of the AI stuff comes out: genetic algorithms, classifier systems and one of the best introductions to neural networks I've ever seen. All the information presented in these chapters are very well written and organized and will definitely have little "gold nuggets" of information for every level of AI-enthusiast.

Augment this excellent book with the code and a very comprehensive glossary at the end, Flake definitely has a winner. I'd say the only downpoints of the book are trivial two points: the rather strange size of the book (almost square) and one or two rather complicated chapters (a derivation of OGY chaotic prediction using calculus springs to mind) that didn't fit the level of the book. This aside, CBN was a delight to read and will definitely serve as a reference book for years to come.