The Pattern Recognition Basis of Artificial Intelligence

Rating:
Summary: Somewhere between a general audience book and a text
Review: I found this book quite interesting due to the way it was structured (from low-level tasks requiring pattern recognition and neural networks to high-level tasks such as game playing and natural language), and due to the author's use of pattern recogniton as a theme for intelligence. In many sections towards the end of each chapter, he also hints at how it may be possible to merge the connectionist and symbolic approaches to AI. For example, with respect to game playing, he discusses possible ways to include backpropagation as tool for evolving a good game evaluation function for backgammon. I also found his intuitive discussions on neural networks (chapters 2 and 3) quite engaging, with an emphasis on possible applications more so than a rigorous mathematical treatment. One main problem I have with the book, however, is its tendency to oversimplify or omit important aspects of the theory. For example, no where does he mention Bayesian classification methods in the pattern recognition chapters. And in general he does not seem to acknowledge the importance of probability in AI. I plan to use the text in a 5 week summer course in AI, because I believe that, with such short time, a terse but engaging book along with good lecture notes seems most appropriate. However, had the couse been 10 or 16 weeks, I would have opted for a more standard text that has 3x as many pages and definitions (e.g. Nilsson's good book). In closing, I recommend this book to anyone who is interested in learning some AI , but does not want to be overburdened with math, or to an expert who prefers to browse the discussions to gain a possibly new perspective on the material.

Rating:
Summary: The book provides an alternative approach to learning AI.
Review: This is an interesting book that introduces AI from a pattern recognition point of view.The author attempted and solved many complex problems by neurocomputing approach. He thus emphasizes the need for neural approach in Artificial Intelligence. He however did not eliminate the scope of all conventional approaches. Some comparative study of the conventioanl and neural algorithms are apparent from the text. The writing style is beautiful. Mature readers will find the book like a novel. I strongly recommend the book for those who want to see the contrast between the "symbol processing AI" and "the neural AI".

Rating:
Summary: The book provides an alternative approach to learning AI.
Review: This is an interesting book that introduces AI from a pattern recognition point of view.The author attempted and solved many complex problems by neurocomputing approach. He thus emphasizes the need for neural approach in Artificial Intelligence. He however did not eliminate the scope of all conventional approaches. Some comparative study of the conventioanl and neural algorithms are apparent from the text. The writing style is beautiful. Mature readers will find the book like a novel. I strongly recommend the book for those who want to see the contrast between the "symbol processing AI" and "the neural AI".