My advisor calls this "the big yellow book" of pattern recognition. This book is comprehensive, unabashedly theoretical, and requires a solid foundation in (preferably measure-theoretic) probability. But with that background, what a feast this book represents! A complete proof of Stone's universal consistency for the nearest neighbor rule? Check, it's here. Bounds on classification error, both the Bayes' error and empirical error? Check, it's here. Error rate estimation techniques? Check, it's here. Arbitrarily slow convergence of a classification rule? Check, it's here. I could keep going, but the answer to whether most topics of theoretical interest are covered would be the same.

The book admits in the preface that "some of the material may be a bit dry", but it is certainly worth the effort. You can't throw a stone very far in the pattern recognition / classification literature and not hit a paper by Devroye or one of his coauthors. The writing style is spare and sometimes they say volumes in a single sentence. Furthermore, the book is not intended to be read in a linear fashion. There is a great map for reading the various chapters (> 30) and how they are interrelated, which chapters need to be read before reading a subsequent chapter.

I can't speak highly enough about this book. It's not an easy read, but it is worth any effort you are willing to put in to learn and understand the theory of pattern recognition. This book will not teach you how to build or tune a classifier, but rather how to understand the theoretical factors that one should consider when doing so.

I have taught Pattern Recognition for several years now, and my battered copy of DGL has been a loyal companion. I do not want to duplicate the previous reviews, but simply to add that this is a book on *nonparametric*, distribution-free methods in Pattern Recognition, the best one available by far on these methods. The authors clearly indicate all the strengths (e.g., universal consistency) and limitations (e.g., arbitrary slow convergence of universally consistent rules, Thm 7.2) of nonparametric methods. However, the authors entirely ignore (intentionally, no doubt) the considerable and distinguished body of literature on *parametric* pattern recognition techniques. For example, you will search in vain here for Linear Discriminant Analysis, or the work of very well-known names such as Anderson, Bowker, Sitgreaves, Raudys, McLachlan, etc. I do not feel that this is a "flaw" of the book, so that I still give it five stars. The book is already voluminous as is, and the authors made a choice on the material, to reflect their own interests and research backgrounds. But they selected a title for the book that may be misleading: this is not a book on all of Pattern Recognition, only on nonparametric, distribution-free methodology in Pattern Recognition. With that said, this is required reading for anyone seriously interested in the Pattern Recognition area.

I offers a formal analysis of different topics in machine learning which I haven't found in any other book but this.

This is the best monograph I have seen on statistical learning theory. It is appropriate for someone interested in the theoretical aspects of machine learning (specifically classification), I wouldn't recommend it as a practical guide. I found many of the proofs difficult to read, sometimes skipping more steps than I felt was reasonable.

I am a machine learning/ data mining student and bought this because my advisor recommended it to me and this is my second favorite book after Christopher Bishop's Pattern Recognition and machine learning... !!!

If i were wondering about some inequality, this book has it.

if i want some sort of risk bounds this book has it. wow. i just skimmed through the contents lists and read parts but i am already a fan.

probably not used in many courses (spanning: CMU, MIT, STANFORD, etc where machine learning is big) because this is more like a reference than a straight learning textbook.

but THIS IS A GEM and a must have.

BUY THIS!!!

The book was damaged as received.