Customer Reviews

Independent Component Analysis - Theory and Applications

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

 

 

This book is not what you expect at all because :

(1) It is merely a collection of papers for the same author lacking proper organization;

(2) Blank pages are deliberately left between chapters, also figures are placed on separate pages this can tell you a lot regarding the quality you expect;

(3) The book is expensive ,I think the author should make it the half or less ;

(4) A researcher in this field hardly sees the book as a reference;

(5) Other topics are ignored ex:Tensoral methods;ICA is not only about Infomax;

(6) The first pages compose the climax of the book ,the rest is just loose and even abscent concepts;

(7) Finally,I think that the book was published too early ,it seems a lot of maturity could have been witnessed if the author waited instead.

Anyone new will be presented to the name of the subject but not the subject itself .The book by Hyvarinen should be available ... ,go for it.I believe life will be a lot easier .The latter is divided into four parts which clearly puts the reader in the right place to start and are : (I) Mathematical Preliminaries (II) Basic Independent Component Analysis (III) Extensions and related Methods (IV) Applications of ICA ,also after reading the sample chapter and contents I think you will not be disappointed . ....

The first part of the book is the best part, it deals with ICA in the information theoretic framework and shows how the (ML) and (infomax) are closely related. However, it is not for beginners since the background material is abbreviated as well as the mathematical exposition of this book assumes the preknowledge in ICA theory.The overall impression one gets is that the book is too short, knowing that the book is more or less a collection of the author's papers, this should not be surprising at all.

I rate it 4/5 because of its expensive price.

Presents clearly the problems and the method. The introductory part was helpful in understanding the ICA theoretical model although more detail on kurtosis description would have been beneficial. I liked the infomax algorithm and the way it was presented. On the down side: some minor erroneus explanations found. I have the feeling that ICA is more than just infomax approach and that the title "Independent Component Analysis - an Infomax Approach" would have been more appropiate. On the application section, very good presentation of the signal separation but very succint explanation on natural images for example. Being the first book I see in the field, I think, a thorough presentation would have been helpful.

If the newcomer is expecting a textbook which gives a thorough and rigorous introduction to the subject, he will not find it here. Essentially, Independent Component Analysis by T. Lee is a compendium of Lee's work on the subject, being, for the most part, a regurgitation of his papers. This in itself is no cause for distress; however, I feel that perhaps some more detail and work could have gone into other researchers' avenues to the problem. For instance, cumulant based methods hardly make it into the book. The derivation of the most important formulas for multiple decorrelation algorithms are omitted. The Fixed-Point Method of Hyvarinen is omitted. A paragraph is given to algorithms which work entirely in the frequency domain. Short shrift is given to JADE.

The applications side is dominated nearly entirely by the biomedical applications to which Lee is associated with, with a small foray into the world of feature extraction.

The introduction and conclusions are well written, though more detail could have helped. There are a few errata throughout though this is normal for a first book. All in all a book with a rather narrow focus.