Customer Reviews

Adaptive Filter Theory (3rd Edition)

5 star:		(6)
4 star:		(1)
3 star:		(3)
2 star:		(2)
1 star:		(0)

 

 

Despite the commonly negative opinion against Simon Haykin's book, I find this book to be a very fun reading. It starts off with a very brief review of DSP (more useful just for getting familiar with the notation, really), properties of random processes, and a small section on linear algebra in the middle of the book.

The rest of the book can be viewed as a story of how different approaches and algorithms were developed, and is a little difficult to use as reference due to its lack of structure and over-dependency on the previous chapters, both for technical content and notation.

But there's a lot of hidden treasures within this book that should have been more emphasized. For example, Mold's theorem that states that any discrete stationary process can be decomposed into a deterministic component and a random component, which are uncorrelated to each other. I'm sorry, but a reference to a proof in another book is not enough to really motivate me. This is a very fundamental theorem if you're interested in stochastic signal processing. Sure, you don't cover the Fundamental Theorem of Calculus in your very first calculus class, but then again this is supposed to be a fairly advanced book.

So if you're interested in learning certain things quickly, this is NOT the book to get. Consider Munson Hayes' book instead. Save this one when you feel like investing a little time to hear Haykin's story on stochastic signal processing.

The book is sound, but I have to agree with others here. Formulas, procedures are presented without an intuitive sense of why things turn out the way they are, or even from the beginning of derivations. Good to implement mechanically all those algorithms without proper understanding.

As it is usually the case, in this very important subject, one has to learn from many sources.

This book looks very impressive, but if you try to understand it you'll find it very mechanical. There is not much motivation behind the many pages of formulas and derivations. I'm not even sure how many people actually read those derivations becuase even in its 4th edition the book and its solution manual both have many typos (see, for example, equations 8.11 and 12.5). Even the problems are more focused on derivations than on numerical examples. This is a good cookbook if you just want to implement an algorithm or find some pointers to the original research papers. Like many other reviewers, I beleive that engineering textbooks are losing their depth and becoming more and more like instruction manuals.

Despite the commonly negative opinion against Simon Haykin's book, I find this book to be a very fun reading. It starts off with a very brief review of DSP (more useful just for getting familiar with the notation, really), properties of random processes, and a small section on linear algebra in the middle of the book.

The rest of the book can be viewed as a story of how different approaches and algorithms were developed, and is a little difficult to use as reference due to its lack of structure and over-dependency on the previous chapters, both for technical content and notation. I have to admit that the notation used in this book is very, very poor and can be a source of frustration. The dependency is also a pain because you always have to keep flipping 100 pages back because Mr. Haykin prefers to say "Eqn. (4.24)" instead of "an AR model".

But there's a lot of hidden treasures within this book that should have been more emphasized. For example, Mold's theorem that states that any discrete stationary process can be decomposed into a deterministic component and a random component, which are uncorrelated to each other. I'm sorry, but a reference to a proof in another book is not enough to really motivate me. This is a very fundamental theorem if you're interested in stochastic signal processing. Sure, you don't cover the Fundamental Theorem of Calculus in your very first calculus class, but then again this is supposed to be a fairly advanced book.

So if you're interested in learning certain things quickly, this is NOT the book to get. Consider Munson Hayes' book instead. Save this one when you feel like investing a little time to hear Haykin's story on stochastic signal processing.

I was introduced to this text in a graduate course. I was not too thrilled about learning from another Haykin book due to a previous experience with his Communication Systems text in an undergraduate course (Horribly confusing... Proakis's text is infinitely better). To my surprise, the book was very detailed and easy to read. The math is very clear and detailed (great for the self learner). Also, the second chapter, which serves as a review of stationary processes and properties, was written much better than most random process textbooks (I applaud Haykin for this given the section was only a review). In chapter 3 or 4, he shows the derivation of the Levison-Durbin Algorithm step-by-step. I strongly disagree with some of the other reviews stating this text is just the typical engineering manual or cookbook with no explainations.

However, this is either a love or hate text. If you are looking for a text about practical linear predictive filter design, this is NOT the book for you. This text is heavily geared towards understanding the theory behind the design... hence the title Adaptive Filter THEORY. However, it can make a great reference to engineers in the field of DSP.

This book is perfect for professionals!
It has very detailed tables on how to implement the algorithms which makes it really easy to implement them.
Many people complain on the mathematical nature of the book, which I feel is a strong point and is required for a proper understanding of the algorithms. Though students may find this a drawback as it does not provide an application based approach. It is more for professionals who want a book that helps them quickly implement the algos and provides a lot of mathematical background, which is necessary to diagnose issues and understand proper usages and requirements.
The comparison of algos is also very useful and helps one in making decisions on the right one for the task.

The text is well organized the overview which is given in the beginning of the book is very usefull and it provides an excellent explaination to stochastic signal processing. the background material is a suitable training for those who want to refresh their knowledge in stochastic signals. The second part explains in detail all the prelimanary subjects in statistical signal processing. All the rest of the book is just what any dsp engineer needs as far as theory is concern. What is a bit missing is an interface to the practical world (maybe a few examples of the algorithms which are introduced in the book implemented on of the shelf dsp cores) but anyone can achieve staff like this in manufacturer benchmarks. In short great book

I have always wondered why many people have negative opinions about books by Simon Haykin, whether it is 'Communication Systems' or 'Adaptive Filter Theory'. Particularly, this book 'Adaptive Filter Theory', in my opinion, is one of the bestbooks on this subject. As Julius Kusuma correctly mentioned, this book is indeed an "adventure ride" into the field of Adaptive Filter Theory.

I discovered this book when I was doing a class project on Self-Orthogonalizing algorithms for Adaptive Beamforming and I felt that all the relevant information that I needed was present in this book. I did'nt really feel the neccesity to refer anything outside this book.

Apart from that, this book contains everything that a graduate student needs to know about this exciting field of adaptive filters. The author assumes some background on Random Signal Theory... I'd suggest to look up Sam Shanmugan et al's, "Random Signals: Detection, Estimation and Data Analysis" before beginning to read (enjoy) this "adventure ride" on Adaptive Filters.

Simon Haykin's Adaptive Filter Theory is a perfect example of everything which is wrong with modern engineering text books. It begs the question of why great educators like Van Trees and Kailath write out of print books, and Sallys like Haykin, Kay and Papoulis can make it to third editions. As a cookbook, one can't do better, but then again, what is science without proper motivation? Even when Haykin attempts to illustrate the theory, he adopts the uninsightful but straightforward mechanics of undergraduate linear algebra, and then again, why did it take hundreds of years to develop these ideas if Haykin has a 3 line proof. The book should be re-titled "Why I'm so smart" By Simon Haykin.

Before start reading this book, read "Uderstanding Digital Signal Processing" by Lyons first.