 |
|
Ballard's new book provides a unique approach to several relevant issues in
natural computation, emphasizing algorithms in relative detriment of neural approaches. This book
presents many virtues and a few secondary drawbacks which can be easily overcomed in future editions.
|
|
|
|
On the positive side, we have the careful presentation of a series of concepts
and techniques in a logical manner and according to a progressive level of complexity. The many appendices
in Part I cover in an effective fashion many of the principal basic concepts needed for proper comprehension
of the respective chapters. Although aimed at undergraduates and beginning graduate students, it is felt
that some sound knowledge on calculus is an essential pre-requisite for the proper assimilation of the
presented material. Another nice feature in Ballard's book is the constant preocupation in illustrating
the introduced concepts and techniques in terms of real, if simple, examples. The chapters also include
some notes and exercises for the industrious reader (no solutions provided). A brief abstract initiates e
very chapter, outlining some of the most important aspects to be covered.
|
|
|
|
Perhaps the principal shortcoming in Ballard's new book stems from something that
is often also one of its virtues, namely the adopted minimalist approach. The problem is that, in some
cases, the concise way in which the concepts are present will likely not be enough to properly acquainting
the novice reader. This was felt especially in Chapters 6, 7, and 11. In addition, though many of the most
promising and modern issues are covered in this book, some important topics have been overlooked, including
multiscale representation and analysis, linear and non-linear regularization, Markov fields, Fourier and
other transforms, invariants, symmetry perception, chaos, and artificial life. It would also be nice if
one or two more complex and complete case-examples (such as some artificial intelligent systems), involving
a number of the presented concepts, could be included and discussed in some detail.
|
|
|
|
All in all, Ballard's new book is an accomplishment in itself and should constitute an important addition to the library of most people interested in any of the related areas. It will also provide an excellent textbook to be used in courses on related areas. It is expected that new editions of this book will keep coming and providing up to dated follow-ups of developments in the covered areas as well as presenting and discussing future trends.
|
|
|
|