This book is fairly mathematically demanding, and when you are tackling such a book you have to rely on the text being accurate. Quite early on, in equation 2.62, there is an error that wasted me two days as I tried to figure out exactly what it was I had missed. I ended up having to refer to the article the material was based on, Geman et al.
From that point on, the clarity of the explanation degraded to the point that I was taking so long understanding it (and with my faith in the accuracy of the information now severely impaired), that I gave up and bought "Introduction to the Theory of Neural Computation", recommended in the comp.ai.neural-nets faq, instead.
The material was not so bad up to that point, and perhaps if you have a tutor to help you when you get stuck it is worth using, but for self-study I would not recommend it. If like me you really want to fully understand the material you will end up wasting a lot of time. If not, there are clearer, simpler books available.