27 of 29 people found the following review helpful
Good not great,
This review is from: Numerical Recipes 3rd Edition: The Art of Scientific Computing (Hardcover)
This book is collection of 15 years work and is widely used and highly respected in scientific computing. It seems churlish to criticize it.
However, Numerical Recipes is not without its faults. In my experience (optimisation, MCMC sampling) the algorithms given do not adequately represent the ones available in the field. There is only one global optimiser (simulated annealing), no non-linear contrained optimisers and no mention of slice sampling for instance. This incompleteness would be helped by including a wide ranging bibliography for each group of algorithms. However, I found the references quite limited.
The book describes itself as a cookbook for cooks. Although this is a worthy aim, it cannot compare to reading the original papers or reviews of algorithms available in journals. In essence, this further reading is what someone needs to do in order to alter an algorithm for their own needs.
The shortcomings could be forgiven if the book provided a way of getting something , relatively simple, working quite quickly. However, you have to type the code in yourself or pay extra to get it in electronic form. Using either of these methods, the licencing terms are restrictive and are for personal use only. This made the book an expensive disappointment for me, especially since free alternatives like the GNU scientific library exist.
On the plus side the descriptions of the available algorithms are excellent given the limited space available to describe them. The authors also include tips based on their experience and mention why a particular algorithm may be more popular despite being no better than some of the others. To my knowledge, Numerical Recipes has no decent competition when it comes to the description of algorithms.