Machine learning and big data are all the craze these days, seen as key technologies to handle the deluge of data expected to flow from the upcoming Internet of Things (IoT). Multimedia information, transmitted by video encoders or still cameras, will form a significant share of this deluge, and coming up with adequate tools to handle it is of paramount importance. Camastra and Vinciarelli’s extended edition of their 2008 book intends to fully cover this rapidly evolving field.

The book is, interestingly, structured in four main parts. The first one describes, in about 80 pages, how our analog world is captured by our senses and digitized in sounds, images, and videos. Part 2, which is the main focus of the book and amounts to 280 pages, digs into the various families of machine learning techniques used to try to classify and, more generally, automatically interpret multimedia digital data. Bayesian theory, clustering techniques, neural networks, hidden Markov models, and kernel methods, such as support vector machines (SVMs) or principal component analysis, are all there. Once the basics are covered, Part 3 illustrates, in 100 pages, their practical use in a few application cases, from speech recognition to automatic personality perception. The final part of the book is a series of appendices that summarize the mathematical prerequisites in, for example, statistics, signal processing, and kernel methods. A detailed index closes the book.

Normally, my review would now give a more detailed account of the book chapters, but I’m afraid this would be a waste of time. Indeed, even though the authors show a clear and deep mastery of the subject, made particularly obvious by the long lists of bibliographic references provided at the end of each chapter (199, for instance, for chapter 9 that covers the kernel techniques used in SVMs), I’m having a very hard time finding why I should recommend this book to anyone. It suffers from so many typographical errors, English lexical and grammatical errors, and mathematical inconsistencies in notations that I don’t think there is one error-free page. I don’t understand how this can happen with a reputable editor such as Springer, and even more so in a second edition of a book, unless, maybe, if it has been completely rewritten. One can even find a couple of Italian words, stemming from what I assume was a draft written in the authors’ native language.

This is really unfortunate since, with proper proofreading and editing, I’m confident this book could have been an interesting addition to the literature on such an important subject; it covers a large variety of issues and provides a wide survey of the literature. But, unless you’re ready to endure approximate English and have a pen and eraser at hand to fix many mathematical formulas, I don’t recommend you read this book.

More reviews about this item: Amazon