Format: Paperback

Winston's book is really terrible. I mean truly repellently, malignantly bad. "Can it really be as bad as all that?" you wonder. Yes!! It's that bad!! For starters, the book is poorly organized. Topics that logically belong together are often several chapters apart. There is no overall structure to the book. It seems like a collection of topics in AI that were hastily assembled without concern for thematic organization or flow. For example, the forward and backward chaining algorithms are presented in a chapter (Ch. 7) on rule-based systems, but are not even mentioned in the chapter (Ch. 13) on logic! Perceptron training is presented AFTER backpropagation! Contrast this with the much better book by Russell and Norvig, which uses the theme of intelligent agents as a continuing motivation throughout, and which groups related topics into logically arranged chapters. The examples in Winston are atrocious. The main example in the backpropagation chapter is some kind of classification network with a bizarre topography. This example is so trivial and weird that it totally fails to illustrate the strengths of backpropagation. The explanations of generalization and overfitting in backprop training are awful. The only chapter of this book that is not an unmitigated pedagogical disaster is the chapter on genetic algorithms, although better introductions exist (e.g. Melanie Mitchell). A further annoyance is the placement of all the exercises at the end of the book instead of the end of the chapters to which they correspond. Avoid this book. It is truly horrible, and vastly superior books on AI are readily available at comparable prices.

See the article here:

Artificial Intelligence: Winston P. Henry: 9780201855043 ...